@comchatx/cx-sponsored-ads-playbook 0.0.0-stage → 0.1.0

package/research/beyond/goals-and-economics.md

@@ -0,0 +1,1084 @@
+# Goals and unit economics
+
+Read 2026-11-09 by the research seat.
+
+Tags: [official] is Amazon's own words: an Amazon Ads guide, help, news or API page, an OpenAPI model, a Selling Partner API document or model, a sell.amazon.com or aboutamazon.com page, or an Amazon staff post in the Seller Forums. Each is cited by page name and URL and was read 2026-11-09 unless another date is given. [opinion] is a named practitioner, agency or paid tool, with the person or firm given; a practitioner article hosted on an Amazon site is still [opinion]. [research] is a published or working paper. [government] is a U.S. government statement (the FTC's press release): neither Amazon's words nor research. [seat] is this seat's own reasoning: allowed in the file, never a position. Figures (percentages, dollar amounts, day counts, thresholds, dates that may change) are recorded with page and date **for human readers only**; a skill reads the live figure at run time from the place named in *What cx must read at run time*.
+
+## What this covers
+
+This file covers the money side of a seller's ad work: what an ad sale is worth to the seller, what the seller is trying to get from advertising, and which limits outside the ad console (fees, margins, stock) bound what advertising can sensibly do. It covers:
+
+- **Break-even ACOS and break-even ROAS**, as Amazon defines them, and the margin before ad cost they rest on;
+- **how that margin is computed** from Amazon's fee schedule, the Selling Partner API's fee estimates and the seller's own costs, and which Amazon costs a per-unit fee estimate does not carry;
+- **target ACOS and ROAS**, and how practitioners derive a target from a profit goal;
+- **TACoS** (ad spend over total sales) and how it relates to ACOS and to profit after ads;
+- **the economics of a click**: CPC, conversion rate, order value, cost per acquisition, and what a click actually costs under Amazon's auction (including the reserve-price facts that became public in the August 2026 FTC complaint and Amazon's response);
+- **marginal against average returns**: why the last dollar of spend can lose money while the campaign's average ACOS looks fine;
+- **goals**: profit, growth, launch, defence and liquidation as Amazon and practitioners frame them; the console's and the API's goal, KPI and bid-strategy fields; Amazon's guidance on choosing an objective;
+- **lifetime value and repeat purchase**: new-to-brand metrics, Amazon's long-term sales metric, Subscribe & Save, the Repeat Purchase report, Amazon's "affordable CPA";
+- **inventory as a limit on ad spend**: out-of-stock pausing, FBA capacity limits and the IPI, storage and aged-inventory costs, days of supply and weeks of cover, and the reports that carry them;
+- **dated fee changes** (2026 and 2027 US letters, the fuel surcharge, holiday peak fees), **returns** as a margin input and what attributed sales deduct, **Amazon Points** on Amazon.co.jp as a seller-funded price lever, and **price against advertising** as levers on the same sales (sections C, O, P, Q; added in the gap fill);
+- **practitioners' frameworks**, named, and where they disagree;
+- **research** on advertising elasticity and optimal ad spend (Dorfman–Steiner, the advertising-elasticity meta-analyses, TV and paid-search ROI studies, LTV of search-acquired customers, paid-to-organic spillover), on why a seller's own regression does not identify the elasticity and what experiments do, and on price elasticity.
+
+Why a senior Amazon advertising lead cares. Every other decision in the pack (a bid, a budget, a harvest, a pause, a season plan) is judged against a number: a target ACOS, a ROAS floor, a TACoS ceiling, a cost per new customer. If that number is wrong, every downstream change is wrong in the same direction. The commonest failures a lead sees are:
+
+- a break-even computed from cost of goods alone, leaving out referral and FBA fees;
+- one target for a whole catalog with very different margins;
+- an average ACOS used where the marginal one decides;
+- attributed sales treated as incremental sales;
+- a launch judged against a mature product's target;
+- spend scaled into a stock-out.
+
+This file gives the model the definitions, Amazon's own rules and the measured effects needed to set and read those numbers. It sets none of them.
+
+It stays in its lane. Topics owned elsewhere are cross-referenced:
+
+- attribution windows, the exact definitions of attributed sales, halo, and which metric exists where: `beyond/metrics-and-attribution.md` (this file uses those metrics, and names fields only where a goal or a margin calculation needs them);
+- organic rank and the ads-against-organic question (incrementality in the account, cannibalisation tests): `areas/ads-organic-and-total.md`;
+- the launch case itself (programs, credits, timeline): `areas/launch.md`;
+- bid strategies, placement adjustments and dynamic bidding mechanics: `areas/bids-and-placements.md`;
+- budgets, budget rules and pacing: `areas/budgets-and-pacing.md`;
+- events, deals and peak-season CPCs: `areas/seasons-and-deals.md`;
+- branded search economics and defence: `beyond/brand-defense-and-competition.md`;
+- Amazon's own automation (Ads Agent, recommendations, rules): `beyond/automation-and-amazon-tools.md`;
+- which ad products a seller may use: `beyond/ad-products-and-eligibility.md`;
+- endpoints and headers: `method/ads-api-surface.md`; report lifecycles: `method/reporting.md`; before/after windows: `method/change-measurement.md`.
+
+Out of scope: a seller's tax and accounting treatment of ad spend; DSP fee structures beyond noting they exist; vendor (1P) margin, which is set by Amazon's purchase price rather than by fees (the Vendor Net Pure Product Margin report is named once in *What cx must read at run time*).
+
+## Amazon's own rules and guidance
+
+Each claim is marked **RULE** (Amazon states it as a requirement or definition, an API enforces or exposes it, or it is a fee or limit) or **GUIDANCE** (Amazon recommends or illustrates it).
+
+### A. ACOS, ROAS and break-even, as Amazon defines them
+
+- **RULE (definition).** "Amazon ACOS is calculated by dividing ad spend by ad revenue, then converting it to a percentage." "ACOS = (ad spend ÷ ad revenue) x 100". "ROAS = ad revenue ÷ ad spend". "Return on ad spend (ROAS) is the inverse of Amazon ACOS". Example: "if you spent $50 on an ad campaign and earned $100 from it, your Amazon ACOS would be 50%" and "the ROAS would be 2". [official] *What is advertising cost of sales (ACOS)? Calculation and tips | Amazon Ads*, https://advertising.amazon.com/library/guides/acos-advertising-cost-of-sales
+- **RULE (definition, other page).** The budget guide says the same with "ad-attributed revenue": "You calculate ACOS by dividing your ad spend by your ad-attributed revenue"; "A lower ACOS indicates a better performance when you are measuring success based on ad-attributed sales." [official] *Advertising budget: Definition, examples, measurement*, https://advertising.amazon.com/library/guides/advertising-budget
+- **RULE (break-even).** "Break-even ACOS is directly linked to your profit margin. In order to maintain a profit, your Amazon ACOS needs to be lower than your profit margin. Otherwise, you will be spending more on ads than you're earning." The same page defines profit margin loosely: "it's the amount earned beyond the cost of product production and any other expenses your brand may have." [official] *What is advertising cost of sales (ACOS)?*, as above.
+- **GUIDANCE (target).** "Each brand is different and will have a different target ACOS. However, the first step should be achieving break-even ACOS and comparing it to your profit margin. From there, you can determine what goal is most important to your brands and campaigns: Is it increasing sales? Is it raising brand awareness?" [official] *What is advertising cost of sales (ACOS)?*
+- **GUIDANCE (no universal number).** "There isn't a definitive number for a good Amazon ACOS. It's dependent on your industry, company size, and campaign frequency, among other variables." [official] same page.
+- **GUIDANCE (newness).** "new campaigns tend to have a high ACOS, merely because of their newness. But that doesn't make a new campaign a failure; it means you should consider different KPIs." [official] same page, section *Why you shouldn't focus too much on ACOS*.
+- **RULE (API definitions).** The v1 reporting metric `metric.roas` is "Return on ad spend, calculated as attributed sales per unit of ad spend. ROAS = sales / total cost", eligible for Sponsored Products, Sponsored Brands, Sponsored Display, Sponsored TV, DSP and Full-Funnel. `metric.totalCost` is "The total amount spent on running the campaign. This excludes the third-party fees paid by the agency." [official] *Purchases (Amazon retail conversions)*, https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/ads-v1/metrics/amazon-retail-conversions/purchases.md; *Costs (Costs and fees)*, https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/ads-v1/metrics/costs-and-fees/costs.md
+- **RULE (API, v3).** Reporting v3 carries a precomputed `acosClicks7d` ("Advertising cost of sales based on purchases made within 7 days of an ad click") and `acosClicks14d`, `roasClicks7d` and `roasClicks14d`, in the `spTargeting`, `spSearchTerm` and `spAdvertisedProduct` report types (among others); `cost` is "Total cost of ad clicks"; `sales7d` is "Total value of sales occurring within 7 days of an ad click". The v3 page also splits same-SKU from other-SKU sales: `attributedSalesSameSku7d` ("where the purchased `SKU` was the same as the `SKU` advertised") and `salesOtherSku7d` ("where the purchased `SKU` was different"). [official] *Reporting columns (version 3)*, https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/v3/columns.md. The same page opens with the banner "Reporting API v3 will be deprecated on June 30, 2027." and points to the v1 reporting API (date for human readers only). Which window applies to which ad product is `beyond/metrics-and-attribution.md`'s.
+
+[seat] Amazon's break-even rule compares ACOS, an attributed-sales ratio, with a margin. Two consequences follow from the definitions, not from any Amazon page:
+
+1. If attributed sales include other SKUs (`salesOtherSku7d`), the margin that matters is the margin of the mix that sold, not of the advertised SKU.
+2. If some attributed sales would have happened without the ad, an ACOS under break-even does not by itself mean the ad paid for itself (section N; *Published research*; `areas/ads-organic-and-total.md`).
+
+### B. Ads Math: break-even ROAS, the click-to-sale chain, and lifetime value
+
+Amazon's *Ads Math* guide is the most complete statement of unit economics on any Amazon Ads page read. It is "a chapter in the Ads Strategy Guidebook for holiday marketing 2025". [official] *Practical calculations for holiday marketing success | Amazon Ads* (Ads Math), https://advertising.amazon.com/library/guides/ads-math
+
+- **GUIDANCE (no threshold).** "There is no general threshold for good or bad return on ad spend (ROAS). A "good" ROAS is one that is profitable for you. Start by identifying your break-even ROAS based on your profit margin. This tells you the minimum ROAS you need to not lose money."
+- **GUIDANCE (ROAS is preliminary).** "ROAS alone is a preliminary metric. It should be considered in the context of your profit margin. A high ROAS may still mean you're losing money if your margin is too low. To answer questions like "is this good or bad?", preliminary metrics should be interpreted in context and paired with deeper performance indicators—such as conversions, customer lifetime value (LTV), profit, or incremental lift."
+- **RULE-shaped formulas (Amazon's).**
+  - "(Revenue – Cost of goods sold) ÷ Revenue = Gross profit margin"
+  - "1 ÷ Gross profit margin = Break-even ROAS"
+  - "A break-even ROAS of 2.5 means you need $2.50 in attributed sales for every $1.00 spent on ads to break even."
+  - "If campaign ROAS greater than or equal to break-even ROAS, then the campaign is profitable (or at least not losing money)."
+- **GUIDANCE (worked table, figures for human readers only).** Advertiser A: ROAS 3, margin 20%, revenue $3,000, gross profit $600, ad spend $1,000, net −$400. Advertiser B: ROAS 2, margin 60%, revenue $2,000, gross profit $1,200, ad spend $1,000, net +$200. "Advertiser A has a higher ROAS, but because their margin is small, they still lose money."
+- **GUIDANCE (bidding up).** "Seasonal events may lead to higher cost per click (CPC) but often increase conversion and average order value (AOV) too. It's essential to model the full equation to see if the trade-off is worth it." Example (figures for human readers only): "CPC is projected to increase by 25%. ROAS is expected to drop from 3.2 to 2.8. Your break-even ROAS is 2.0." Verdict: "The campaign is still profitable because a ROAS of 2.8 is greater than 2.0 (break-even point). Therefore, you can afford to bid up." Key insight: "Go in with a plan, not fear. Test and scale where the math supports it."
+- **RULE-shaped formulas (the click chain).** "Don't judge CPC alone as it is a preliminary metric. Link it to conversion rate and profit per order to find your effective cost per acquisition (CPA)". The formulas:
+  - "CPC ÷ Conversion rate = CPA"
+  - "Average order value × Gross profit margin = Profit per sale"
+  - "Profit – CPA = Net profit or loss per sale"
+  - and: "If the net profit or loss per sale result is positive, then the campaign is profitable. If it's negative, then the campaign is unprofitable, and you either need to lower your CPC or increase your conversion rate."
+  - The page limits this calculator: it "is meant to determine if your CPC is too high in relation to CPA for one-off purchases and does not take into account customers who continue to purchase from your brand."
+- **RULE-shaped formulas (lifetime value).**
+  - "To set a profitable ad budget, you first need to calculate customer lifetime value (LTV). Then, using your profit margin, you can determine your affordable cost per acquisition (CPA)—the maximum you can spend to acquire a customer and still break even over their lifetime."
+  - "Average order value × Average orders = Customer LTV"
+  - "LTV × Profit margin = Affordable CPA"
+  - "Once you've identified your affordable CPA, you can use it as the upper limit for your ad spend per new customer to ensure you break even over their lifetime."
+  - "It helps you avoid common pitfalls like bidding too low on valuable audiences or scaling back campaigns that are profitable in the long run, even if their initial return seems low."
+
+[seat] Three things about Ads Math a skill should hold precisely:
+
+1. Its "gross profit margin" is revenue minus cost of goods sold, over revenue. For a seller on Amazon, Amazon's own selling and fulfillment fees (section C) are costs that scale with each unit sold. A margin that leaves them out overstates break-even ACOS by the fee share of price. The page does not say whether "cost of goods sold" is meant to include them.
+2. The LTV formula multiplies order value by order count and then by "profit margin". It does not discount future orders, and it does not deduct a repeat order's own fees or a Subscribe & Save discount. Those are a seller's own choices to add.
+3. The bid-up example compares the *projected average* ROAS after the change with break-even. The ROAS of the *added* spend alone can be lower than the new average (section N; *Published research*, Shapiro, Hitsch and Tuchman).
+
+**Algebra that follows from Amazon's own definitions** [seat, an identity, not a claim of Amazon's]:
+
+- ACOS = spend ÷ sales = (clicks × CPC) ÷ (orders × AOV) = CPC ÷ (CVR × AOV), where CVR = orders ÷ clicks and AOV = sales ÷ orders.
+- Break-even CPC (Amazon's "net profit per sale = 0") = AOV × margin × CVR.
+- Therefore ACOS < margin and "Profit – CPA > 0" are the same test for one-off purchases.
+- "AOV" here is attributed sales per attributed order, and "CVR" is attributed orders per click. Both are read from the same report, with the same window.
+
+### C. The costs a margin must carry: Amazon's fee schedule and FBA cost list
+
+- **RULE (referral fee).** "Referral fees vary by product category. For every item sold, you'll pay a percentage of the total price or a minimum amount, whichever is greater." "An item's total price includes its list price, as well as shipping costs and any gift-wrapping charges." "Note: A product's fee category may not be the same as the category that appears to customers in the Amazon store." "In addition to the referral fee, media items incur a closing fee $1.80 per item." [official] *How much does it cost to sell on Amazon?* (the page's title; called *Selling on Amazon fees* in this file), https://sell.amazon.com/pricing. The category table on that page is tiered (examples on the page: "15% for the portion of the total sales price up to $300.00, and 8% for any portion of the total sales price greater than $300.00"; "8% for products with a total sales price of $10.00 or less, and 15% for products with a total sales price greater than $10.00"; minimums such as "$0.30"). The figures are for human readers only; the per-SKU fee comes from section D at run time.
+- **RULE (selling plan).** Individual "$0.99 / item sold"; Professional "$39.99 / month" (figures for human readers only). [official] same page.
+- **RULE (incentives that change margin).** "If you have a new-to-Amazon brand, you can get over $50,000 in automatic referral-fee waivers for your first year—when margins matter the most." [official] same page. Which programs apply to a seller is `areas/launch.md`'s.
+- **RULE (FBA cost list).** Amazon lists the costs FBA can include:
+  - "Fulfillment costs: Includes picking, packing, and shipping orders, as well as handling customer service and returns. These costs are based on the product's price, weight, and dimensions."
+  - "Storage costs: Charged monthly based on the space your inventory occupies in Amazon's fulfillment network (calculated based on your daily average volume in cubic feet)."
+  - "Aged inventory: Charged monthly for all items stored in a fulfillment center for more than 181 days."
+  - "Returns processing: Charged on orders when Amazon provides a customer with free return shipping."
+  - "Removal, disposal, and liquidation orders: You can have Amazon return, dispose of, or liquidate your inventory in a fulfillment center for a per-item charge."
+  - "FBA inbound placement service: ... For standard and large bulky sized products, the cost for this service covers distributing inventory to fulfillment centers close to customers."
+
+  [official] *Amazon FBA (Fulfillment by Amazon)* (the page's title; called *Fulfillment by Amazon (FBA)* in this file), https://sell.amazon.com/fulfillment-by-amazon
+- **GUIDANCE.** "To determine whether FBA is right for you, calculate your margins by taking storage and fulfillment costs into account." "Preview FBA costs. In the Seller Central main menu, hover over Reports , then select Fulfillment , followed by Fee Preview ." [official] same page. The Fee Preview is the report in section D.
+- **RULE (restock limits and storage).** "Restock limits help you determine how much inventory you can ship into our fulfillment network. Storage costs are tied to your usage of space in our fulfillment network. If you exceed your restock limit, you may see an increase in your storage costs." [official] same page. (The page still says "restock limits"; Amazon's staff post in section K describes "capacity limits".)
+- **RULE (Subscribe & Save costs).** "There are no additional selling costs for the Subscribe & Save program. Participants pay standard selling fees and normal costs for FBA." A seller can "fund a 5% or 10% discount on some or all of your enrolled products". [official] *Subscribe & Save | Sell on Amazon*, https://sell.amazon.com/programs/subscribe-and-save. A seller-funded discount lowers the margin on the subscribed unit.
+
+[seat] Per-unit costs that scale with a sale fall into two groups:
+
+- Costs Amazon charges per unit: referral fee, closing fee where it applies, FBA fulfillment fee.
+- Costs that are per unit only after allocation: storage, inbound placement, returns processing, aged inventory, a seller-funded Subscribe & Save discount, coupon and deal fees.
+
+The second group is real but depends on stock levels, sell-through and the seller's choices. That is where practitioners differ on what "margin before ads" includes (*What practitioners add*).
+
+**Fee changes by date (added in the gap fill).** Amazon changes its US fees on announced dates, and two of the changes fall inside the peak advertising season. All figures below are **for human readers only**; a skill reads the live rate at run time (*What cx must read at run time*).
+
+- **RULE (the yearly US fee letter, 2026).** "Near the end of each year, we provide our selling partners with an update on changes to Referral, Fulfillment by Amazon (FBA), and other fees". For 2026: "FBA fees will increase by an average of $0.08 per unit sold, or less than 0.5% of an average item's selling price. This is on top of no increase in US Referral and FBA fees in 2025." "There will be no new FBA fee types in 2026". "Unless otherwise noted, all changes will be effective January 15, 2026." Sellers are pointed to "the Revenue Calculator and Fee and Economics Preview report which are being updated with the 2026 rates" and "the new Profit Analytics dashboard to get a detailed view of your unit economics, including the ability to see how fee changes impact each of your products". [official] *Update to U.S. Referral and Fulfillment by Amazon fees for 2026* (Dharmesh Mehta, dated October 15, 2025), https://sellingpartners.aboutamazon.com/update-to-u-s-referral-and-fulfillment-by-amazon-fees-for-2026; the same letter as an Amazon staff post (`News_Amazon`), *2026 Updates to US Referral and Fulfillment by Amazon Fees*, https://sellercentral.amazon.com/seller-forums/discussions/t/f3fa3211-820b-4e2e-a023-158a9cf55f99. The fee detail page it links (Seller Central help G201411300) is a JavaScript shell to a plain GET (*Unreachable*).
+- **RULE (fuel and logistics surcharge, 2026).** "Starting April 17, 2026, a 3.5% fuel and logistics-related surcharge will be applied to fulfillment fees across Fulfillment by Amazon (FBA) in the US and Canada as well as to Remote Fulfillment with FBA from the US into Canada, Mexico, and Brazil. Starting May 2, 2026, this surcharge will take effect for Buy with Prime in the US and Multi-Channel Fulfillment (MCF) in the US and Canada." "The surcharge will be calculated on your fulfillment fees, not on the sale price of your items. On average, this equates to $0.17 per unit for US FBA". Amazon calls it temporary ("we implement temporary surcharges on our fulfillment fees") and names no end date. [official, Amazon staff in forum] *Fuel and logistics-related surcharge: FBA, MCF, and BWP in US and CA* (poster `News_Amazon`), https://sellercentral.amazon.com/seller-forums/discussions/t/7cbc0233-ee5b-4359-978a-dee7cad5c6f4. Amazon's 2027 letter (below) says "The FBA fuel and logistics surcharge remains unchanged."
+- **RULE (holiday peak fulfillment fees, 2026–27).** "Holiday peak fulfillment fees will apply again from October 15, 2026, to January 14, 2027, for Fulfillment by Amazon (FBA), Remote Fulfillment with FBA, Multi-Channel Fulfillment (MCF), and Buy with Prime – the same dates as last year. Holiday peak fulfillment fees will have the same per unit increase over non-peak rates as last year, averaging $0.32 per unit. The 3.5% fuel and logistics-related surcharge will apply on top of holiday peak fulfillment fees." The same post on storage: "if you send inventory through AWD and enable automatic replenishment to FBA, you'll keep paying the off-peak monthly storage rate through October 31, 2026", which presupposes a peak monthly storage rate; it gives no storage figure. On capacity: "Our fulfillment center teams focus on receiving holiday shipments in September and October, then shift to processing customer orders in November and December. So, you may see lower capacity limits during this time." On deal fees for Best Deals, Lightning Deals and Prime Exclusive Price Discounts: "Upfront fee: $100 per promotion, plus Variable fee: 1.5% of promotional sales (capped at $5,000)". [official, Amazon staff in forum] *Holiday 2026: Same fees, same eligibility, earlier deadlines* (poster `News_Amazon`), https://sellercentral.amazon.com/seller-forums/discussions/t/3e31fbb7-04e0-4ed4-873e-f74b1052e2ff. Deal timing and deal mechanics are `areas/seasons-and-deals.md`'s; this file needs only that the deal fee and the peak fulfillment fee are per-unit or per-sale costs in the margin.
+- **RULE (the yearly US fee letter, 2027, published November 5, 2026).** "On average, FBA fulfillment fees will increase by $0.19 per unit sold, approximately 0.8% of the average item's selling price. The impact will vary by product, size, and price. Fees for some products will increase, and others will decrease". "We're sharing them now so that you have more than two months to plan before most changes take effect on January 15, 2027." On storage: "Effective February 1, 2027, we'll eliminate the storage utilization surcharge, which will make it easier to stock up ahead of high-demand events. We'll also remove the aged inventory surcharge and combine it with monthly storage fees into a single, age-based rate. These changes will result in flat or lower fees for inventory that turns in under six months, higher fees for aged inventory". On the low-inventory-level fee: "Today, the fee is assessed using trailing 30- and 90-day sales. Effective February 1, the fee will apply only when your inventory falls below 28 days of supply, based on actual shipments in both the four weeks before and after the assessment date." The summary page covers "additional updates for small and large standard-size items, items priced above $50, "Buy X, Get Y" promotions, Multichannel Fulfillment, and Amazon Warehousing and Distribution". Sellers are pointed to "the new Fee Change Comparison dashboard in Profit Analytics" and the Revenue Calculator. [official] *Update to US Referral and Fulfillment by Amazon fees for 2027* (Sunny Jain, Vice President, Worldwide Fulfillment by Amazon, dated November 5, 2026), https://sellingpartners.aboutamazon.com/update-to-us-referral-and-fulfillment-by-amazon-fees-for-2027. The MCF page carries the banner "Updates to MCF US fulfillment fees go into effect January 15, 2027." [official] *Multichannel Fulfillment (MCF) - Amazon Supply Chain Services*, https://supplychain.amazon.com/mcf.
+- **RULE (the fee types Amazon itself itemises).** Amazon's Data Kiosk economics schema lists the fee types it reports and previews per product, among them `REFERRAL_FEES`, `CLOSING_FEES`, `PER_ITEM_SELLING_FEES`, `FBA_FULFILLMENT_FEES` ("including base rate and surcharges"), `BASE_FBA_FULFILLMENT_FEE`, `MONTHLY_INVENTORY_STORAGE_FEES`, `BASE_MONTHLY_STORAGE_FEE`, `STORAGE_UTILIZATION_SURCHARGE` ("Monthly Inventory Storage Fee Storage Utilization Surcharge"), `AGED_INVENTORY_SURCHARGE` ("previously known as the Long-term Storage Fee"), `LOW_INVENTORY_LEVEL_FEE`, `FBA_INBOUND_PLACEMENT_SERVICE_FEE`, `RETURN_PROCESSING_FEE` ("also know as Returns Processing Fee for apparel and shoes"), `HIGH_RETURN_RATE_FEE` ("also know as Returns Processing Fee for Non-apparel and Non-shoes"), `DIGITAL_SERVICES_FEE_FBA`, `DIGITAL_SERVICES_FEE_SOA`, `PAN_EU_OVERSIZE_FEE` and `SPONSORED_PRODUCTS_CHARGES` (enum `PreviewFeeType`; Amazon's spelling "know"). When a query's date range spans an Amazon fee change, "the response will split the request date range and display the fee separately", and `amountPerUnitDelta` "is mainly used to indicate the change in amountPerUnit before and after an Amazon fee change". `economicsPreview` takes an end date "no more than 120 days after today". [official] *analytics_economics_2024_03_15.graphql* (Data Kiosk schema), https://github.com/amzn/selling-partner-api-models/blob/main/schemas/data-kiosk/analytics_economics_2024_03_15.graphql
+- **Reported, from Amazon's 2027 help pages, not read here.** PPC Land (Luis Rijo, November 6, 2026) read the seven Seller Central pages that came with the 2027 letter and reports: standard-size FBA rate cards in three price bands, "below $10, $10 to $50 and above $50", with items "priced at exactly $10 and $50" on the $10–$50 rates; low-price rates narrowed ("cheap books, DVDs, computer items and car accessories lose a discount they used to get"); fees "calculated when a shipment leaves a fulfillment center, not when the order is placed"; the returns processing fee, "which applies to products with high return rates", "updated to align with the 2027 size tiers", from January 15, 2027 for apparel and shoes and February 1, 2027 for "all other high-return categories", with the new rates on a separate page "the documents" did not publish; and no 2027 holiday peak rates published yet. It gives 2026 standard-size monthly storage as "$0.78 per cubic foot a month from January to September in 2026 and $2.40 from October to December". [opinion, PPC Land, reporting Amazon's pages] *Amazon sellers face average $0.19 FBA fee rise from January 15, 2027*, https://ppc.land/amazon-sellers-face-average-0-19-fba-fee-rise-from-january-15-2027/. Practitioner storage figures disagree (other vendor pages seen in search give $0.87 for January–September); no Amazon page read here states the US storage rate card.
+- **RULE (seller-fulfilled Prime carries the seller's own delivery and returns cost).** Seller Fulfilled Prime gives "access to Prime branding, showing your commitment to provide fast and free delivery, plus free returns for domestic Prime customers"; enrolment needs a prequalification on "late-shipment, valid-tracking, and cancellation rates" and a 30-day trial. [official] *Seller Fulfilled Prime | Sell on Amazon*, https://sell.amazon.com/programs/seller-fulfilled-prime. For a seller-fulfilled unit, Amazon's own economics schema expects the seller to supply "Cost of fulfilling an order containing the product, including labor, packing material, shipping to customer, customer service etc." (`MfnPerUnitCost.fulfillmentCost`) and "Cost to store a single unit of the product per month" (`storageCost`). [official] *analytics_economics_2024_03_15.graphql*. MCF fulfils "customer orders when they buy on your own website or another sales channel" [official] *Costs for optional selling services | Sell on Amazon*, https://sell.amazon.com/pricing/optional-costs; Amazon's MCF page describes "a single pool of inventory" serving "all your sales channels—Amazon, DTC, social platforms, and more" [official] *Multichannel Fulfillment (MCF)*, as above. An Amazon-store order is fulfilled by FBA at FBA fees, so MCF fees are not a cost of an Amazon ad sale. They enter only through the shared pool (peak storage, capacity).
+
+[seat] What the dated fee changes do to break-even:
+
+- Break-even ACOS is margin before ads over price. A per-unit fee change of Δ moves break-even ACOS by Δ ÷ price, in percentage points. For the same Δ the move is largest on low-priced items.
+- The holiday peak fulfillment fee (October 15 to January 14 in 2026–27) and the peak monthly storage rate coincide with the weeks when CPCs and deal fees peak (`areas/seasons-and-deals.md`). A break-even computed on non-peak fees overstates the peak-season break-even.
+- The 3.5% surcharge is a percentage of the fulfillment fee, so it scales with every fulfillment-fee change, peak included.
+- Storage utilization surcharge and aged-inventory surcharge are holding costs, not per-sale costs, until 2027; from February 1, 2027 Amazon folds the aged surcharge into an age-based monthly rate. For a skill, the per-unit storage share stays an allocation (section K) whatever the rate card.
+- A price change that crosses a fulfillment-fee price band (per PPC Land's reading of the rate cards, $10 and $50) changes the fulfillment fee as well as the referral fee. A price cut is therefore not a pure margin-for-volume trade at the band edges (section Q).
+
+### D. Fee estimates from the Selling Partner API
+
+- **RULE (the API's purpose and caveat).** "You use the Product Fees API to retrieve estimated fees for a product. You can then account for those fees in your pricing." "You must specify your products by ASIN or SKU (not UPC, ISBN, or other identifiers)." "The estimated fees returned by this API are not guaranteed. Actual fees can vary." Availability: "Sellers only". [official] *Product Fees API* (SP-API documentation, page updatedAt 2026-09-09), https://developer-docs.amazon/sp-api/docs/product-fees-api.md
+- **RULE (operations, from the model).** Three operations in Product Fees v0:
+  - `POST /products/fees/v0/listings/{SellerSKU}/feesEstimate` (`getMyFeesEstimateForSKU`);
+  - `POST /products/fees/v0/items/{Asin}/feesEstimate` (`getMyFeesEstimateForASIN`);
+  - `POST /products/fees/v0/feesEstimate` (`getMyFeesEstimates`, a batch; the tutorial is titled "Get product fees for a batched set of up to 20 items").
+
+  The ASIN operation's note: "When using the `getMyFeesEstimateForASIN` operation with an ASIN, the fee estimates might be different. This is because these estimates use the item's catalog size, which might not always match the actual size of the item sent to Amazon." [official] *productFeesV0.json*, https://github.com/amzn/selling-partner-api-models/blob/main/models/product-fees-api-model/productFeesV0.json; *getMyFeesEstimateForSKU* reference, https://developer-docs.amazon/sp-api/reference/getmyfeesestimateforsku.md; SP-API docs index, https://developer-docs.amazon/sp-api/llms.txt
+- **RULE (request shape).** `FeesEstimateRequest` requires `MarketplaceId`, `PriceToEstimateFees` and `Identifier`, and takes `IsAmazonFulfilled` ("When true, the offer is fulfilled by Amazon") and `OptionalFulfillmentProgram` (`FBA_CORE`, the default, "Returns the standard Amazon fulfillment fees for the offer"; `FBA_SNL`, sunset; `FBA_EFN`, "Returns the cross-border European Fulfillment Network fees across EU countries"). `PriceToEstimateFees` requires `ListingPrice` and takes `Shipping` and `Points`. [official] *productFeesV0.json*
+- **RULE (reply shape).** `FeesEstimate` holds `TimeOfFeesEstimation`, `TotalFeesEstimate` ("Total estimated fees for a given item, price, and fulfillment channel") and `FeeDetailList`. Each `FeeDetail` has `FeeType` (a free string: "The type of fee charged to a seller"), `FeeAmount`, `FeePromotion`, `TaxAmount`, `FinalFee` and an `IncludedFeeDetailList`. `FeesEstimateResult.Status` is one of `Success`, `ClientError`, `ServiceError`. The model's sandbox example returns a `TotalFeesEstimate` with an empty `FeeDetailList`. [official] *productFeesV0.json*
+- **RULE (the price is an input).** The estimate is "for an item ... in the marketplace specified in the request body" at the `ListingPrice` the caller supplies. The operation can be called "before the selling partner sets the item's price". [official] same model. [seat] A margin at a deal or coupon price needs an estimate at that price; referral fees are a percentage of price, and FBA fees are documented as "based on the product's price, weight, and dimensions" (section C).
+- **RULE (FBA Fee Preview report).** `GET_FBA_ESTIMATED_FBA_FEES_TXT_DATA`: "Contains the estimated Amazon Selling and Fulfillment Fees for the seller's FBA inventory with active offers. The content is updated at least once every 72 hours." "This report can only be requested once per day per seller." Columns for all stores include `sku`, `fnsku`, `asin`, `your-price`, `sales-price`, dimensions, `estimated-fee-total` and `estimated-referral-fee-per-unit`. US only: `product-size-tier`, `estimated-future-order-handling-fee-per-order`, `expected-future-fulfillment-fee-per-unit`. US and Canada: `estimated-variable-closing-fee`, `expected-fulfillment-fee-per-unit`. US and Mexico: `estimated-future-fee (Current Selling on Amazon + Future Fulfillment fees)`, `estimated-future-pick-pack-fee-per-unit`, `estimated-future-weight-handling-fee-per-unit`. Mexico: `estimated-pick-pack-fee-per-unit`, `estimated-weight-handling-fee-per-unit`. [official] *Fulfillment by Amazon (FBA) Reports* (report type values), https://developer-docs.amazon/sp-api/docs/report-type-values-fba.md
+- **RULE (storage reports).** `GET_FBA_STORAGE_FEE_CHARGES_DATA` carries `estimated_monthly_storage_fee` per ASIN. `GET_FBA_OVERAGE_FEE_CHARGES_DATA` carries `storage_usage_volume`, `storage_limit_volume`, `overage_volume` and `charged_fee_amount`. `GET_FBA_INVENTORY_PLANNING_DATA` carries `estimated-storage-cost-next-month` and the aged-inventory-surcharge columns `quantity-to-be-charged-ais-181-210-days` through `estimated-ais-365-plus-days`. [official] same page.
+- **RULE (actual fees, after the fact).** The Finances API v2024-06-19 has `GET /finances/2024-06-19/transactions` (`listTransactions`): "Returns transactions for the given parameters. Financial events might not include orders from the last 48 hours." The model's schemas include `Transaction`, `Breakdowns` and `Breakdown`. [official] *finances_2024-06-19.json*, https://github.com/amzn/selling-partner-api-models/blob/main/models/finances-api-model/finances_2024-06-19.json
+
+[seat] The SP-API gives three readings of a unit's Amazon cost:
+
+1. an estimate at any price (Product Fees API);
+2. the current estimate per FBA SKU at its current price, with future-fee columns in some stores (Fee Preview);
+3. what was actually charged (Finances).
+
+None of them carries the seller's landed cost of goods, inbound freight to Amazon, or packaging; those are the seller's numbers. The per-unit fee estimate is not documented to include storage, inbound placement, returns processing or aged-inventory charges; those come from the storage and planning reports or from the seller's own allocation.
+
+### E. What a click costs: bids, reserve prices and the 2026 FTC complaint
+
+The CPC in the click chain (section B) is what the seller pays, not what the seller bids. How the two relate became a matter of public record in 2026.
+
+- **RULE (bid is a ceiling).** "Sponsored Products uses a cost-per-click , auction-based pricing model. You set the maximum amount that you are willing to pay when a customer clicks an ad for your product." "A bid is the maximum amount you are willing to pay when a customer clicks an ad." [official] *FAQs - Advertising questions and resources*, https://advertising.amazon.com/resources/faq
+- **RULE (what goes into ad selection).** "When choosing Sponsored Brands and Sponsored Products , Amazon Ads primarily considers the keywords advertisers bid on, the amount advertisers bid, and the customers' expected interest in advertiser offers, based on information about the offer, similarities to other offers, and past customer purchases." "Sponsored ads products do not charge any "demand-side fees"". [official] *Your guide to Amazon Ads pricing transparency*, https://advertising.amazon.com/library/guides/pricing-transparency
+- **RULE (Amazon's description of its pricing, 2026).** Amazon's published response to the FTC lawsuit states:
+  - "Since we first introduced ads in our Store in 2006, we have priced clicks using a form of generalized "second-price" auction".
+  - "we began to test a concept called "soft reserve prices," a real-time minimum value that seeks to better reflect what each placement is actually worth. We also introduced what we call a "hard reserve," which is the minimum a bid must surpass to enter an auction."
+  - "When the winning advertiser's bid exceeds both the hard and soft reserve, they pay the soft reserve, which is less than they were willing to pay. When the winning advertiser's bid exceeds the hard reserve but doesn't meet the soft reserve, we still grant the placement to that advertiser and they pay their bid. In no scenario does an advertiser pay more than their bid."
+  - "reserves are determined in real time and aren't predictable in advance by anyone, including Amazon or the advertiser."
+  - "The Ad Console Campaign Builder ... has clearly stated since 2018 that a bid represents the maximum an advertiser could be charged."
+  - "we updated our Amazon Ads Help content to explicitly explain the use of reserve prices."
+
+  [official] *Amazon's response to the FTC's lawsuit regarding Sponsored Ads*, https://www.aboutamazon.com/news/company-news/amazon-ftc-sponsored-ads-lawsuit-response (the HTML URL returned HTTP 404 to a plain GET and to WebFetch at verification on 2026-11-09; the text was read from Amazon's markdown copy at https://www.aboutamazon.com/md/company-news/amazon-ftc-sponsored-ads-lawsuit-response)
+- **RULE-adjacent (Amazon's figures, for human readers only).** Same page: "in 2024, approximately 92% of selected Sponsored Products ads were not the highest bid"; "The mean winning advertiser's bid is typically about the 128th bid by amount"; "from 2019 to 2024, the average winning bid for Sponsored Products search ads fell 50%"; "for Sponsored Products search ads, from 2019 through 2024, the average cost-per-click to advertisers remained flat when adjusted for inflation"; "Conversion rates ... have increased over 24% for individual Sponsored Products advertisers from 2021 through 2025"; "eighty percent of bid changes on clicked Sponsored Products search ads occurred within one day of a prior change". [official] same page.
+- **The allegation (not a finding).** The FTC press release, dated August 31, 2026, says Amazon told advertisers it ran "second price" auctions where winners would pay "one cent more than the next highest bidder". It alleges that by 2024 advertisers paid their own winning bid about 80% of the time on Sponsored Products ads, up from 30–40% in 2021 and 70% in 2022, after Amazon from 2019 added what it internally called a "soft reserve price". The suit is in the U.S. District Court for the Western District of Washington and covers "Sponsored Product ads," "Sponsored Brands ads," and "Display Ads." [government; not Amazon's words] *FTC, States Sue Amazon over Secret Ad Surcharge Scheme*, https://www.ftc.gov/news-events/news/press-releases/2026/08/ftc-states-sue-amazon-over-secret-ad-surcharge-scheme (read through the WebFetch tool; a plain GET returned no article text). Nothing in the complaint has been tried.
+- **RULE (dynamic bidding raises the ceiling).** In the Ads API v1 `BidStrategy` enum: `SALES_UP_AND_DOWN` "Increases or decreases your bids in real time by a maximum of 100%"; `SALES_DOWN_ONLY` "Bids will never increase beyond your set bid"; `MANUAL` "Uses your exact bid and any placement adjustments you set, and is not subject to dynamic bidding." [official] *Amazon Ads API v1 merged OpenAPI* (`AmazonAdsAPIALLMerged_prod_3p.json`), https://d1y2lf8k3vrkfu.cloudfront.net/openapi/en-us/dest/AmazonAdsAPIALLMerged_prod_3p.json. The exact ceilings per placement, and placement multipliers, are `areas/bids-and-placements.md`'s.
+
+[seat] For the economics, the click chain uses the *paid* CPC from a report (`cost` ÷ `clicks`, or `costPerClick`). Under Amazon's own description:
+
+- the paid CPC equals the bid whenever the bid is under the soft reserve;
+- the soft reserve is unknowable in advance;
+- the "max bid = value per click" rule from second-price theory is therefore not a safe assumption about what a bid will cost;
+- the ratio of paid CPC to the effective bid (after dynamic and placement adjustments) is observable per target, from the seller's own reports.
+
+### F. Goals in Amazon's words: objectives and the KPIs that go with them
+
+- **GUIDANCE (set an objective first).** "It's important to have a specific goal in mind before you launch your campaign. For example: Brand exposure : Get the word out for your company. New product promotion : Build buzz for new items. Inventory clearance : Promote sales of items you're looking to replace or upgrade. Peak season promotion : Maximize products during the holidays , such as Prime Day ." [official] *Best practices for your Sponsored Products ads*, https://advertising.amazon.com/library/guides/sponsored-products-best-practices
+- **GUIDANCE (what SP is for).** "Consider using Sponsored Products for product visibility, new offers, unique selections, offers with low glance views, clearance items, and seasonal promotions." And: "Before you create your first campaign, it's important to know what business goals you want to accomplish through advertising. Establishing your goals up front will help you choose which products to advertise, decide how to structure your campaigns, and better analyze performance." [official] *FAQs*, as above.
+- **GUIDANCE (goals and their KPIs).** "To gauge the success of your campaign, you must define what you want out of it, so you know which performance metrics to focus on." Amazon's measurement guide pairs goals with metrics:
+
+  | Goal (Amazon's words) | Metrics Amazon lists |
+  |---|---|
+  | Awareness: "making your ad more visible to new audiences" | campaign reach, cost per mille (CPM), click-through rate (CTR) |
+  | Consideration: "engaging them during the decision-making process" | CTR, cost-per-click |
+  | Purchase or conversion: "If sales are your immediate priority, your campaign is successful when shoppers buy your product." | CTR, conversion rate, "Cost of acquisition: how much it costs you to convert each new customer", ROAS, "Sales per visitor, sales per order, units per order" |
+  | Loyalty: "engage your existing customers to drive repeat purchases" | CTR, conversion rate, ROAS, "Sales per visitor, sales per order, units per order" |
+
+  The same guide on new customers: "Review the suite of new-to-brand metrics to measure orders and sales from first-time customers of your brand in the Amazon store. Use these metrics to learn how many new customers you've reached, estimate the cost of acquisition, and develop strategies for growing your customer base." And: "For a new product with limited awareness, it may take additional budget to drive sales and improve your ROAS." "After your campaign has been running for at least two weeks, review your advertising reports to understand your performance, in light of your campaign goals." [official] *How to measure and improve your campaigns*, https://advertising.amazon.com/library/guides/measure-improve-campaigns
+- **GUIDANCE (loyalty measured).** "Customer loyalty can be measured through repeat purchases, return on ad spend, percent of new-to-brand customers, number of reviews, follower count, and campaign metrics on Amazon Attribution." [official] *Customer loyalty: Drive brand love and repeat purchases | Amazon Ads*, https://advertising.amazon.com/objectives/customer-loyalty
+- **GUIDANCE (budget from goals).** "To determine your advertising budget, start by deciding what your marketing goals are." "There is no universal formula for calculating a good ad budget." "The best ad budget is one that fits your business's priorities and long-term goals." [official] *Advertising budget*, as above.
+- **GUIDANCE (ROI refinement).** "For a more accurate result, you can refine the formula by subtracting organic sales growth, considering gross or net profit, and including all relevant costs." On timing: "some campaigns aim for long-term goals like building brand loyalty or improving customer relationships. These benefits might not show up in your ROI calculations for months or even years". [official] *What is Marketing ROI? Formula & Examples | Amazon Ads*, https://advertising.amazon.com/library/guides/marketing-roi
+- **GUIDANCE (a benchmark Amazon prints, for human readers only).** "A good benchmark is earning at least $2–4 for every $1 spent on advertising." [official] *How to Help Increase Sales with FBA and Advertising*, https://advertising.amazon.com/library/guides/fba. This conflicts in kind with the ACOS guide's "There isn't a definitive number for a good Amazon ACOS" and Ads Math's "There is no general threshold" (section N).
+- **GUIDANCE (Amazon's product labels as goal hints).** Amazon's campaign-creation page labels products, and its guide maps them to goals. Labels: "New product", "Brand building", "Forecasted demand", "Trending", "Benchmarking", and products with upcoming deals. On "Brand building": "Shoppers who view or purchase these products are more likely to purchase other products in your brand's catalog, visit your store and search for your brand." Amazon's figures (for human readers only): "Products identified as helping grow brand engagement* saw 11x more repurchases when advertised"; "Products with deals or coupons that were advertised with Sponsored Products saw an average 65% increase in sales and 75% increase in units sold compared to when the deal or coupon product was not advertised". [official] *Which products should I add to my campaign?* (product strategy), https://advertising.amazon.com/library/guides/product-strategy
+- **Practitioner on Amazon's site (tagged opinion).** Amazon's "Ad Intel" expert-advice page (dated July 22, 2025) by Jason Landro, "Co-CEO and Co-founder, Nectar" (an Amazon Ads Advanced Partner), argues: "While revenue and return on ad spend (ROAS) matter during peak events, these metrics don't tell the whole story. Define success based on contribution to overall business health with specific, measurable goals incorporating profitability metrics like contribution margin or net profit targets for promoted ASINs. Consider strategic objectives such as new-to-brand (NTB) customer acquisition at sustainable customer acquisition cost (CAC)." Its example (figures for human readers only): a CPG brand "shifted from targeting a 4x ROAS during Prime Day to focusing on contribution margin per unit and specific NTB CAC goals. Though overall ROAS decreased slightly to 3.8x, the focus on higher-margin ASINs yielded 15% higher net profit contribution versus the prior year." [opinion, Nectar, hosted on Amazon Ads] *Ad Intel: 5 Amazon shopping event advertising tips from Nectar*, https://advertising.amazon.com/library/expert-advice/five-advertising-tips-from-nectar
+
+### G. Goals in the console and the API: what a campaign can carry
+
+- **RULE (v1 campaign fields per ad product).** Amazon's v1 campaign entity guide marks `optimizations.goalSettings` as available in the common model, Sponsored Brands and DSP, and **blank for Sponsored Products**. Within `GoalSettings`: `goal` is optional in the common model and **read-only** for Sponsored Brands and DSP; `kpi` is **required** for Sponsored Brands and DSP; `kpiValue` is optional for DSP. `optimizations.bidSettings` is available for SP and SB. `optimizations.budgetSettings` is available for SP (not SB). [official] *The campaign entity in the Amazon Ads API*, https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/campaign-management/entities/campaign.md ("Key: ✓ required | ○ optional | □ read-only | ◐ nested object").
+- **RULE (enums, v1).**
+  - `Goal`: `AWARENESS`, `CONSIDERATION`, `CONVERSIONS`, `ENGAGEMENT` ("customer engagement actions beyond product purchase, such as review requests or follows. Used by Sponsored Actions campaigns").
+  - `KPI` (35 values) includes `ROAS` ("driving improved return of ad spend"), `ROAS_PROMOTED`, `ROAS_COMBINED`, `PURCHASES`, `NEW_TO_BRAND_PURCHASES` ("maximizing new to brand purchases"), `TOP_OF_SEARCH_IMPRESSION_SHARE`, `CLICKS`, `COST_PER_CLICK`, `PAGE_VIEWS`, `SUBSCRIBE`, `REACH`; four values are marked deprecated: `RETURN_ON_AD_SPEND` ("Please use ROAS_PROMOTED"), `TOTAL_RETURN_ON_AD_SPEND` ("Please use ROAS"), `COMBINED_RETURN_ON_AD_SPEND` ("Please use ROAS_COMBINED") and `COST_PER_ACTION` ("Please use COST_PER_CONVERSION_OFF_AMAZON").
+  - `GoalSettings.kpiValue`: "The value of the KPI that the campaign is working to optimize."
+  - `BidStrategy`: `MANUAL`, `NEW_TO_BRAND` ("Optimizes bidding to maximize new-to-brand customer acquisitions"), `PRIORITIZE_KPI_TARGET` ("Optimizes bidding to achieve the KPI target specified"), `RULE_BASED` ("Applies bidding rules defined by the advertiser"), `SALES_DOWN_ONLY`, `SALES_UP_AND_DOWN`, `SPEND_BUDGET_IN_FULL` ("Prioritize spending full budget, while maximizing performance"), `USE_CAMPAIGN_STRATEGY`.
+  - `CampaignGoalFilter.include` takes 1–3 `Goal` values (a campaign query can filter by goal).
+  - `AdProduct` has seven values, including `FULL_FUNNEL` ("Full funnel ad product combining multiple ad types") and `RMS` ("Sponsored ads for external retailers").
+
+  [official] *Amazon Ads API v1 merged OpenAPI*, as above.
+- **RULE (Sponsored Brands goals).** "brands can select one of two campaign goals from the campaign creation process": "Drive page visits" ("The success metric for this goal is "clicks." All previously created Sponsored Brands campaigns will be tagged as "Drive page visits." You will pay by cost-per-click (CPC)") and "Grow brand impression share" ("Your ads will show on top-of-search. The success metric for this goal is "top-of-search impression share." You will pay by cost per 1,000 viewable impressions (vCPM)"). "Combining these two campaign goals in the campaign portfolio, brands can structure campaigns to deliver business short-term sales and long-term sustainable growth." [official] *Sponsored Brands goal-based campaigns*, https://advertising.amazon.com/resources/whats-new/sponsored-brands-goal-based-campaigns (datePublished 2023-10-25). Structural detail is `areas/campaign-structure.md`'s.
+- **RULE (a ROAS guardrail for SP).** "For any existing campaign running for at least 10 days, you can apply a rule with a guardrail of ROAS and Amazon Ads may then adjust your base bids up and down to increase conversions while maintaining your guardrails. Amazon Ads does not guarantee the ROAS guardrail." Eligibility: "running for at least 10 days prior to setting a rule and have a minimum of 10 conversions in the last 30 days", plus a per-marketplace minimum daily budget (table on the page; US 10, JP 600, IN 300 in local currency, for human readers only). "the ROAS guardrail is based on the previous 21 days." "if your campaign is not meeting the guardrail and the return on ad spend (ROAS) of the campaign for the last 20 days drops to less than 30% of your ROAS over a 20-to-40-day period, we will disable the rule on your behalf and will enable your previous bidding strategy and targeting level bids." Amazon's advice: "if you set your ROAS guardrails too aggressively, you may see a decrease in the number of conversions compared to your historical performance." [official] *Sponsored Products rule-based bidding overview*, https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/rules/bidding-rules/overview.md. Operating the rule is `areas/bids-and-placements.md`'s; for this file, it is the one documented place an SP campaign carries a profit-shaped number (a ROAS floor) that Amazon's system acts on.
+- **RULE (goal-led buying, 2026).** Amazon Ads Agent (announced 2026-09-29) "brings three campaign types—sponsored ads, Full-Funnel Campaigns, and Display, Video, and Audio—together". For Full-Funnel Campaigns, "Advertisers provide their budget, products, and creative, and Amazon's AI handles the rest", and reporting shows "new-to-brand and long-term sales metrics". "Advertisers easily launch a DVA+ campaign by starting with business objectives: sales, traffic, or awareness". [official] *Introducing Amazon Ads Agent, your AI-powered ad platform* (datePublished 2026-09-29), https://advertising.amazon.com/library/news/amazon-ads-agent. The unBoxed 2026 page: "You set your goal and AI handles the channel mix, creative, audiences, and optimization. Beta testers saw 67% higher long-term return on ad spend and 29% lower cost to acquire new-to-brand customers" (figures for human readers only). [official] *unBoxed 2026 news: Explore the latest advertising announcements*, https://advertising.amazon.com/library/news/unboxed-2026-news-announcements
+- **Reported, not confirmed by an Amazon page read here.** A practitioner reports an Amazon beta for Sponsored Products named "Automatically optimizing ROAS within the campaign budget": "You set a daily budget only. No base bid, no ROAS target. Amazon controls bid amounts per auction", "rolling out to select US accounts". [opinion, Matthieu, Nova Analytics, 2026-02-18] *Amazon Tests ROAS Auto-Bidding for Sponsored Products*, https://novadata.io/resources/news/amazon-sponsored-products-roas-auto-bidding-2026. No Amazon page read here names it; the v1 `BidStrategy` enum has no value with that name (it has `PRIORITIZE_KPI_TARGET` and `SPEND_BUDGET_IN_FULL`). See *Open questions*.
+
+[seat] For a Sponsored Products campaign, what the campaign is "for" has no field. It shows only through its bid strategy, its rule (if any), its targets, its budget and the seller's naming. A goal such as "profit", "rank", "launch" or "liquidate" is a seller-side label, and has to be kept somewhere else: the seller's own words, a portfolio name, or cx's notes.
+
+### H. Long-term value in Amazon's ad metrics: new-to-brand, long-term sales, Subscribe & Save
+
+- **RULE (new-to-brand defined).** `metric.newToBrandPurchases`: "The number of purchases by shoppers who had not purchased a product from the brand in the last 12 months." Eligible programs: Sponsored Products, Sponsored Brands, Sponsored Display, Sponsored TV, Amazon DSP, Full-Funnel. `metric.newToBrandSales`: "The sales from shoppers who had not purchased a product from the brand in the last 12 months." `metric.costPerNewToBrandPurchase`: "The average cost to acquire a purchase from a shopper who had not purchased a product from the brand in the last 12 months. Cost per purchase (new to brand) = total cost / purchases (new to brand))". `metric.percentOfSalesNewToBrand` = "sales (new to brand) / sales". `metric.newToBrandRoas` ("sales (new to brand) / total cost") is listed for SB, SD, ST, DSP and Full-Funnel, **not SP**. The family has 102 fields, including peer benchmarks (`metric.costPerNewToBrandPurchaseP25`/`P50`/`P75`, "representing the value where half of peer brands fall above and half fall below" for P50). [official] *Purchases (new to brand) (Amazon retail conversions)*, https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/ads-v1/metrics/amazon-retail-conversions/purchases-new-to-brand.md
+- **RULE (v3 differs).** In Reporting v3, `newToBrandPurchases` ("The number of first-time orders for brand products over a one-year lookback window resulting from an ad click or view. Not available for book vendors.") is listed for SB, SD, ST and DSP report types, **not for any `sp*` report type**. [official] *Reporting columns (version 3)*, as above. [seat] A seller reading NTB for Sponsored Products reads it from v1 reporting; in v3 it is not there.
+- **RULE (long-term sales).** `metric.longTermSales`: "Attributed sales plus the estimated 365-day value of new-to-brand shopper engagements attributed to the campaign, based on the brand's historical shopping patterns." `metric.longTermRoas`: "Long-term ROAS = long-term sales / total cost". Horizons `longTermSales30d`, `60d`, `90d`, `180d`. Cohort fields: `metric.cohortLongTermSales*`, and `metric.cohortAccumulatedSales30d` through `365d` ("Sales over 30 days from the cohort of shoppers whose shopping events were used as inputs to long-term sales calculations"). Benchmarks: `metric.longTermRoasP25`/`P50`/`P75`. Program eligibility for all of them: Sponsored Brands, Sponsored Display, Sponsored TV, Amazon DSP, and (for most) Full-Funnel; **Sponsored Products is not listed**. [official] *Long-term sales (Amazon retail conversions)*, https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/ads-v1/metrics/amazon-retail-conversions/long-term-sales.md. The v3 `longTermSales` column: "The 12 month potential value of new-to-brand customer-brand engagements (DPV, Branded Search, ATC, Purchase) based on the average return on engagement of shoppers who completed these same actions 12 months ago with your brand, plus the 14d total sales of the campaign", in `sdCampaigns`, `sbCampaigns`, `stCampaigns`, `dspCampaign`. [official] *Reporting columns (version 3)*.
+- **RULE (Amazon's framing).** "Long-term sales is an ad-attributed metric that captures the complete value your advertising drives—including immediate sales and the projected value from customer engagement over the next 12 months." Campaign reporting covers "proprietary Amazon metrics such as long-term sales, new-to-brand, brand halo, and Subscribe & Save", and "e-commerce goals, which are unique to Amazon ( advertising cost of sale , Subscribe and Save, etc.)". [official] *Campaign reporting: Analyze advertising performance*, https://advertising.amazon.com/measurement-analytics/campaign-reporting
+- **RULE (Subscribe & Save in ad metrics).** `metric.subscribeAndSaveSubscriptions`: "The number of Subscribe & Save subscriptions attributed to an ad interaction." `metric.costPerSubscribeAndSaveSubscription` = "total cost / Subscribe & Save subscriptions". Also `...FromClicks`, `...FromViews` and `...Promoted`. Eligible: SB, SD, ST, DSP, Full-Funnel; **SP not listed**. [official] *Subscribe & Save subscriptions (Amazon retail conversions)*, https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/ads-v1/metrics/amazon-retail-conversions/subscribe-save-subscriptions.md. In v3, `newSubscribeAndSave` ("Number of new Subscribe & Save subscriptions for all products, attributed to an ad view or click. This does not include replenishment subscription orders.") is listed only for DSP report types. [official] *Reporting columns (version 3)*.
+- **RULE (Amazon Retail Purchases in AMC).** The AMC dataset `amazon_retail_purchases` "enables advertisers to analyze long-term (up to 60 months) customer purchase behavior". Its scope is "Entire purchase activity in Amazon store, regardless of advertising"; for Seller Central accounts it "should match with what is shown in the Seller Central Reports > Business Reports > By ASIN Detail report". It carries `purchase_program_name` ("Indicates the purchase program associated with the retail event"). It is a paid-features dataset, but "Starting June 1, 2026, you can query this dataset at no cost through December 31, 2026 in every region where it is available." [official] *Amazon Retail Purchases in AMC*, https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/amazon-marketing-cloud/datasources/amazon_retail_purchase_paid.md; *AMC paid features overview*, https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/amazon-marketing-cloud/datasources/paid_features_overview.md. Running AMC is `beyond/automation-and-amazon-tools.md`'s and `beyond/metrics-and-attribution.md`'s; this file needs only the fact that a multi-year repeat-purchase view exists in Amazon's tooling.
+
+[seat] Several Amazon numbers point at the value of a customer beyond the first order, and they are not the same thing:
+
+- `costPerNewToBrandPurchase` divides *all* cost by NTB purchases. It is not the cost of the NTB purchases alone.
+- Long-term sales is a modelled forecast. It covers SB, SD, ST, DSP and Full-Funnel, but not SP.
+- Ads Math's "affordable CPA" is a seller-side calculation from the seller's own LTV.
+- The Repeat Purchase report and the Replenishment API (section I) are observed seller-side history.
+
+A skill treating any one of these as "LTV" needs to know which one it is reading.
+
+### I. Repeat purchase and Subscribe & Save on the seller side
+
+- **RULE (Repeat Purchase report).** `GET_BRAND_ANALYTICS_REPEAT_PURCHASE_REPORT`: "Contains data on the quantity of repeated purchases of the selling partner's items." Periods `WEEK`, `MONTH`, `QUARTER`. Fields under `dataByAsin`: `startDate`, `endDate`, `asin`, `orders`, `uniqueCustomers`, `repeatCustomersPctTotal`, `repeatPurchaseRevenue` (`amount`, `currencyCode`), `repeatPurchaseRevenuePctTotal`. Availability: "Sellers and vendors who have the Brand Analytics Selling Partner API role and are registered in Amazon Brand Registry", in a listed set of stores. [official] *Analytics Reports* (report type values), https://developer-docs.amazon/sp-api/docs/report-type-values-analytics.md
+- **RULE (Replenishment API).** "Currently, the Subscribe & Save replenishment program is supported." Operations:
+  - `POST /replenishment/2022-11-07/sellingPartners/metrics/search` (`getSellingPartnerMetrics`);
+  - `POST /replenishment/2022-11-07/offers/metrics/search` (`listOfferMetrics`);
+  - `POST /replenishment/2022-11-07/offers/search` (`listOffers`).
+
+  The `Metric` enum includes:
+  - `SUBSCRIBER_LIFETIME_VALUE_BY_CUSTOMER_SEGMENT` ("The total customer spend on your catalog over the past 24 months, including both one-time purchases and Subscribe & Save orders. Subscribers with more than 12 months of tenure are categorized as 'Established', while those with less are classified as 'Growing'.");
+  - `SUBSCRIBER_NON_SUBSCRIBER_AVERAGE_REVENUE` ("over the past 12 months for sellers");
+  - `SUBSCRIBER_NON_SUBSCRIBER_AVERAGE_REORDERS`;
+  - `SUBSCRIBER_RETENTION` ("The percentage of subscriptions retained after 30 and 90 days of subscription creation");
+  - `ACTIVE_SUBSCRIPTIONS`, `TOTAL_SUBSCRIPTIONS_REVENUE`, `SHIPPED_SUBSCRIPTION_UNITS`, `REVENUE_PENETRATION`, `REVENUE_PENETRATION_BY_SELLER_FUNDING`, `SIGNUP_CONVERSION_BY_SELLER_FUNDING`, `SHARE_OF_COUPON_SUBSCRIPTIONS`, `COUPONS_REVENUE_PENETRATION`, `REVENUE_BY_DELIVERIES`;
+  - `NOT_DELIVERED_DUE_TO_OOS` ("The percentage of items that were not shipped out of the total shipped units over a period of time due to being out of stock") and `LOST_REVENUE_DUE_TO_OOS` ("The revenue that would have been generated had there not been out of stock").
+
+  `TimeInterval`: "only data for the trailing 2 years is supported". Offer `EligibilityStatus`: `ELIGIBLE`, `INELIGIBLE` ("existing subscriptions will be cancelled"), `SUSPENDED`, `REPLENISHMENT_ONLY_ORDERING` ("also blocks one-time purchases to preserve inventory for existing subscriptions"). [official] *Replenishment API*, https://developer-docs.amazon/sp-api/docs/replenishment-api.md; *replenishment-2022-11-07.json*, https://github.com/amzn/selling-partner-api-models/blob/main/models/replenishment-api-model/replenishment-2022-11-07.json
+- **RULE (program terms).** "To participate in Subscribe & Save, your products need to be part of a brand enrolled in Amazon Brand Registry". "your brand's products are eligible for Subscribe & Save based on their category, average selling price, sales performance, and fulfillment history and in-stock rate." "To be auto-enrolled in Subscribe & Save, your products must be sold through Fulfillment by Amazon (FBA)". "Products are enrolled with a 0% seller-funded discount that you can increase to 5% or 10%." On Amazon's funding the page says two things: "Amazon funds a 5% discount on Subscribe & Save orders of five items or more" and "Eligible FBA products are automatically enrolled at a 0% seller-funded discount, with Amazon funding a 5% discount." "Keeping products in stock is a key consideration for the Subscribe & Save program." Amazon's figure (for human readers only): "On average, Subscribe & Save products with a 10%-15% discount can drive up to a 1.8x increase in conversion". [official] *Subscribe & Save | Sell on Amazon*, as above.
+
+[seat] For a replenishable product, the seller-side history needed for the LTV in Ads Math ("Average order value × Average orders") exists in two Amazon reads:
+
+- repeat revenue and repeat-customer share per ASIN (Repeat Purchase report);
+- subscriber versus non-subscriber revenue and reorders, plus retention (Replenishment API).
+
+Neither is joined to ad exposure on the seller side. The join exists only inside AMC.
+
+### J. TACoS in Amazon's words
+
+- **GUIDANCE (definition).** "Monitor your Total Advertising Cost of Sales (TACOS) to understand how efficiently your advertising investment is working. TACOS shows your ad spend as a percentage of total sales." [official] *The author's guide to Sponsored Products* (datePublished 2023-09-07, dateModified 2026-10-09), https://advertising.amazon.com/library/guides/authors-guide-to-sponsored-products. It is written for authors (KDP).
+- **GUIDANCE (the wider view, a second Amazon page; restated in the gap fill).** "ACOS only measures the efficiency of ad-attributed sales, not the full picture. That's where total advertising cost of sales (TACOS) comes in. TACOS reveals whether your total sales (organic and paid) are growing in relation to your ad spend." The page then quotes an agency (Adam Mellott, BTR Media, [opinion]): "When ACOS rises during high-traffic events, we don't immediately pull back. Instead, we look at TACOS—and often, we find it dropping." [official, with an [opinion] quote] *Holiday advertising myths: What experts want you to know* (datePublished 2025-09-26, dateModified 2026-10-09), https://advertising.amazon.com/library/guides/common-holiday-advertising-myths
+- So two Amazon library pages name TACOS as ad spend against total sales, total sales being "organic and paid". Neither names the sales field, the ad products counted or the window. No Ads API metric named TACoS was found in the v1 metric families or v3 columns read. This matches the owner file's settled wording (`areas/ads-organic-and-total.md`, P-ads-organic-and-total-6).
+- [seat] TACoS needs total sales from outside the Ads API: Sales and Traffic (`orderedProductSales` from `GET_SALES_AND_TRAFFIC_REPORT`, section *What cx must read at run time*) over the same days and the same marketplace. It also needs ad spend summed across every ad product that advertises the ASIN or brand. Identity: TACoS = ACOS × (attributed sales ÷ total sales). With one margin m for every unit, profit after ads as a share of total sales = m − TACoS. Both identities hold only if every unit carries the same margin.
+
+### K. Inventory as a limit on ad spend
+
+- **RULE (out of stock, no ad).** "Remember that your product must be in stock and priced competitively in order to become the Featured Offer, so take into account product pricing and availability when deciding to advertise. If your product isn't the Featured Offer or is out of stock, your ad will not display." [official] *FAQs*, as above. The pausing mechanics and eligibility reasons are `areas/launch.md`'s and `areas/diagnosis.md`'s.
+- **RULE (FBA capacity limits; Amazon staff post).** An Amazon staff post (poster `Dominic_Amazon`, 2024-03-14) in the Seller Forums:
+  - "FBA capacity limits help us manage unexpected surges in demand and supply chain disruptions so that we can run our fulfillment network efficiently".
+  - "You can view your capacity limits and capacity usage in the Capacity Monitor at the bottom of the FBA dashboard".
+  - "Individual selling accounts have a capacity limit of 15 cubic feet; this limit doesn't change and is not eligible to be increased."
+  - "New professional selling accounts (active in FBA for less than 39 weeks) don't receive FBA capacity limits."
+  - "Professional selling accounts (active in FBA for more than 39 weeks) receive capacity limits based on IPI score and sales performance. Sellers with consistently higher IPI scores receive higher capacity limits, adjusted for sales volume and available capacity. Usage against your FBA capacity limits will consider your inventory on-hand at Amazon and open shipments on the way to our fulfillment centers, calculated in cubic feet."
+  - "you can use Capacity Manager to request an increase to your capacity limits at a reservation fee that you specify." "We'll assess fulfillment center capacity regularly (typically twice per week) and grant requests as capacity is available."
+
+  [official, Amazon staff in forum] *Refresher: FBA Capacity Limits & How They Work*, https://sellercentral.amazon.com/seller-forums/discussions/t/7dd1b421-5863-46c8-a88a-ba2cdcabad94. Amazon's own announcement of the system (dated March 1, 2023; added at verification) says limits are "a single monthly limit to determine how much inventory sellers can send to and store at Amazon", that "Capacity limits for the upcoming month will be announced in the third full week of each month via the Capacity Monitor in Seller Central", and that granted requests' "reservation fees are offset by earning performance credits from the sales they generate using the extra capacity. Performance credits are designed to offset up to 100% of the reservation fee". [official] *Amazon announces new streamlined FBA capacity management system*, https://sell.amazon.com/blog/announcements/fba-capacity. Several practitioner blogs report capacity cuts and IPI-threshold changes for 2026; no Amazon page read here confirms them. The Seller Central help page on capacity limits (GAFNWEYTJUV2GBFC) returned a JavaScript shell to a plain GET (*Unreachable*). The figures are for human readers only.
+- **RULE (IPI; Amazon staff post).** An Amazon staff post (poster `Indy_Amazon`, 2026-08-13):
+  - "Your IPI score (ranging from 0 to 1,000) measures how efficiently you manage your FBA inventory."
+  - "Your IPI score directly influences your FBA capacity limits."
+  - "Items are considered "excess" if you have over 90 days of supply based on forecasted demand."
+  - "Your sell-through rate = units shipped over the past 90 days ÷ average units on hand during that period."
+  - "Stranded inventory = products sitting in a fulfillment center without an active listing."
+  - "If your best-selling, replenishable products keep going out of stock, it signals poor inventory management".
+  - "New ASINs don't affect your IPI for the first 90 days".
+  - The post's own advice: "Maintain enough inventory to cover only 30–60 days of expected sales".
+
+  [official, Amazon staff in forum] *IPI Score Dropping? What's Dragging It Down and How to Recover*, https://sellercentral.amazon.com/seller-forums/discussions/t/a57e02a2-b759-4d73-aa84-ba5f795156bf
+- **RULE (aged inventory and storage, as costs of not selling).** "Aged inventory: Charged monthly for all items stored in a fulfillment center for more than 181 days." Storage is "Charged monthly based on the space your inventory occupies". [official] *Fulfillment by Amazon (FBA)*, as above.
+- **RULE (planning report columns).** `GET_FBA_INVENTORY_PLANNING_DATA` columns include:
+  - `available`, `units-shipped-t7`/`t30`/`t60`/`t90`, `sell-through`;
+  - `days-of-supply`, `estimated-excess-quantity`, `weeks-of-cover-t30`, `weeks-of-cover-t90`, `healthy-inventory-level`;
+  - `recommended-action`, `recommended-sales-price`, `recommended-sale-duration-days`, `recommended-removal-quantity`, `estimated-cost-savings-of-recommended-actions`;
+  - `inv-age-0-to-90-days` through `inv-age-456-plus-days`;
+  - `estimated-storage-cost-next-month`, the aged-inventory-surcharge columns;
+  - `historical-days-of-supply`, `fba-minimum-inventory-level`, `fba-inventory-level-health-status`;
+  - "Exempted from Low-Inventory-Level fee?", "Low-Inventory-Level fee applied in current week?", "Short term historical days of supply", "Long term historical days of supply";
+  - "Total Days of Supply (including units from open shipments)";
+  - `is-seasonal-in-next-3-months`, `season-name`, `season-start-date`, `season-end-date`;
+  - `inbound-quantity`, `inbound-working`, `inbound-shipped`, `inbound-received`.
+
+  "Attributes vary by store." [official] *Fulfillment by Amazon (FBA) Reports*, as above.
+- **RULE (restock report).** `GET_RESTOCK_INVENTORY_RECOMMENDATIONS_REPORT`: "Provides recommendations on products to restock, suggested order quantities, and reorder dates. Content updated in near real-time." Columns include "Total Days of Supply (including units from open shipments)", "Days of Supply at Amazon Fulfillment Network", `Alert`, "Recommended replenishment qty", "Recommended ship date", "Units Sold Last 30 Days", "Total Units", "Inbound". [official] same page.
+- **RULE (stock-outs cost Subscribe & Save).** The Replenishment API reports `NOT_DELIVERED_DUE_TO_OOS` and `LOST_REVENUE_DUE_TO_OOS`. `REPLENISHMENT_ONLY_ORDERING` "blocks one-time purchases to preserve inventory for existing subscriptions." [official] *replenishment-2022-11-07.json*. [seat] When stock is short and S&S subscribers exist, ad-driven one-time demand competes with subscription deliveries for the same units. The second state means Amazon itself may stop one-time purchases, and an ad for that offer would then send shoppers to an offer they cannot buy.
+- **The low-inventory-level fee.** The planning report exposes the fee's flags and the two historical days-of-supply figures. The Seller Central help page that states the fee's rule is a JavaScript shell to a plain GET and was not read here (*Unreachable*). The listing playbook's earlier capture of it (`cx-listing-playbook/research/angles/inventory-and-fulfillment.md`, read 2026-10-14 by a route this seat may not use) is not re-verified here, and no threshold from it is restated.
+
+[seat] Inventory bounds advertising in four ways the model can read:
+
+1. **No stock, no ad.** The ad stops serving.
+2. **Selling faster than stock can be replaced.** This brings forward a stock-out. Its costs are the lost sales and the ranking effects owned by `areas/ads-organic-and-total.md`, and, where the fee applies, the low-inventory-level fee flagged in the planning report.
+3. **Holding too much.** This costs storage, aged-inventory surcharges and IPI, and through IPI it costs capacity. Here ad spend that moves units has a value beyond the sale's margin: the storage and surcharge it avoids.
+4. **Capacity limits.** These cap what can be restocked. Demand created beyond capacity cannot be served from FBA.
+
+The arithmetic links ad-driven units to days of cover: days of supply shrink by the extra units per day. That is what makes "weeks of cover" and an ad budget the same decision. No Amazon page read here states a rule linking ad spend to stock cover; Amazon's own planning report gives the cover, and the ad reports give the units.
+
+### L. Liquidation and clearance in Amazon's words
+
+- **GUIDANCE.** "Inventory clearance : Promote sales of items you're looking to replace or upgrade." [official] *Best practices for your Sponsored Products ads*. "clearance items" are listed among SP uses. [official] *FAQs*.
+- **RULE (alternatives to selling).** "You can have Amazon return, dispose of, or liquidate your inventory in a fulfillment center for a per-item charge." [official] *Fulfillment by Amazon (FBA)*. The planning report's `recommended-action`, `recommended-sales-price`, `recommended-sale-duration-days` and `recommended-removal-quantity` are Amazon's own per-SKU suggestions. [official] *FBA Reports*.
+- [seat] For a SKU being cleared, the comparison is no longer between ad spend and margin. It is between selling through ads and the alternatives: removal or disposal (per-item charges), liquidation, or continued storage with aged-inventory surcharges. An ACOS above the normal margin can still leave the seller better off than those alternatives. Amazon's pages read here give the cost categories, not a formula.
+
+### M. Amazon's figures collected (for human readers only)
+
+| Figure | Page |
+|---|---|
+| 50% ACOS example ($50 spend, $100 sales); ROAS 2 | *What is advertising cost of sales (ACOS)?* |
+| Break-even ROAS 2.5 at 40% margin; Advertiser A/B table; bid-up example 3.2 → 2.8 against 2.0 | *Ads Math* |
+| "at least $2–4 for every $1 spent" | *How to Help Increase Sales with FBA and Advertising* |
+| Referral fees and minimums by category; $1.80 media closing fee; $0.99 per item / $39.99 per month plans; "over $50,000" referral-fee waivers for new brands | *Selling on Amazon fees* |
+| Aged inventory "more than 181 days" | *Fulfillment by Amazon (FBA)* |
+| S&S: 0%/5%/10% seller funding; Amazon funds 5%; "up to a 1.8x increase in conversion" at 10–15% discount | *Subscribe & Save | Sell on Amazon* |
+| 11x repurchases (brand-building products); +65% sales and +75% units for advertised deals and coupons | *Which products should I add to my campaign?* |
+| Rule-based bidding: 10 days, 10 conversions in 30 days, 21-day guardrail basis, 20-day/30%/20–40-day auto-disable; minimum daily budget table | *Sponsored Products rule-based bidding overview* |
+| 92% of selected SP ads not the highest bid; winning bid about the 128th; winning bids −50% 2019–2024; CPC flat in real terms 2019–2024; conversion rate +24% 2021–2025 | Amazon's FTC response |
+| Paid own bid ~80% of the time by 2024 (alleged) | FTC press release |
+| Full-Funnel beta: +67% long-term ROAS, −29% NTB cost | *unBoxed 2026* |
+| Capacity: 15 cubic feet for Individual accounts; no limits under 39 weeks in FBA; Capacity Manager reviewed "typically twice per week" | Amazon staff forum post, 2024-03-14 |
+| IPI 0–1,000; excess over 90 days of supply; new ASINs exempt 90 days; "30–60 days of expected sales" advice | Amazon staff forum post, 2026-08-13 |
+| NTB lookback 12 months; long-term sales horizons 30/60/90/180/365 days | v1 metric pages |
+| 2026 US FBA fees +$0.08 average per unit, effective January 15, 2026 | 2026 fee letter (sellingpartners.aboutamazon.com; `News_Amazon` post) |
+| 3.5% fuel and logistics surcharge on fulfillment fees from April 17, 2026 (MCF and Buy with Prime from May 2); about $0.17 per US FBA unit | `News_Amazon` post |
+| Holiday peak fulfillment fees October 15, 2026 to January 14, 2027, about $0.32 per unit over non-peak; deal fees $100 + 1.5% capped at $5,000 | `News_Amazon` *Holiday 2026* post |
+| 2027 US FBA fees +$0.19 average per unit (about 0.8% of price) from January 15, 2027; storage utilization surcharge removed and aged surcharge merged February 1, 2027; LILF below 28 days of supply on both the four weeks before and after | 2027 fee letter |
+| Data Kiosk returns delay about 45 days; economics preview up to 120 days ahead | Data Kiosk schemas and guide |
+| JP points: 1 point = 1 yen; seller may set up to 50%; default 1% auto-applied from November 1, 2021; double-points weekends add Amazon-funded points up to 5% per item and 5,000 per order | JP `News_Amazon` posts; Amazon.co.jp points terms |
+
+### N. Where Amazon's pages disagree, or are silent
+
+1. **What "profit margin" means.** The ACOS guide's margin is "the amount earned beyond the cost of product production and any other expenses". Ads Math's margin is "(Revenue – Cost of goods sold) ÷ Revenue". The FBA page says to "calculate your margins by taking storage and fulfillment costs into account". The three are not stated as the same quantity.
+2. **Benchmarks.** The ACOS guide ("There isn't a definitive number") and Ads Math ("no general threshold") against the FBA guide ("A good benchmark is earning at least $2–4 for every $1 spent").
+3. **Average against marginal.** Ads Math's bid-up example tests the projected average ROAS. No Amazon page read here discusses the ROAS of incremental spend. Marketing ROI suggests "subtracting organic sales growth", and Ads Math lists "incremental lift" among deeper indicators, but neither gives a method for Sponsored Products. Incrementality tools are `areas/ads-organic-and-total.md`'s.
+4. **Goals for SP.** Sponsored Brands has goal-based campaigns. The v1 entity guide gives SP no goal settings, while the v1 enums and the Ads Agent news describe goal-led buying across products. The status of an SP ROAS-auto-bidding beta is reported only by a practitioner.
+5. **Long-term value metrics skip SP.** NTB purchases are SP-eligible in v1 but absent from v3 SP report types. Long-term sales, NTB ROAS and S&S subscriptions are not SP-eligible in v1.
+6. **TACoS.** Named on two Amazon library pages (the author's guide gives the ratio; the holiday myths page gives the "organic and paid" denominator); neither names fields, and no API metric was found.
+7. **Subscribe & Save funding.** The same sell.amazon.com page states Amazon's 5% two ways (on "orders of five items or more", and as a flat Amazon-funded 5% on auto-enrolled products).
+8. **Auction pricing.** Amazon's description (soft reserve paid when the bid exceeds it, otherwise the bid) and the FTC's allegation (own bid paid ~80% of the time) are both public. The case is untried.
+9. **Returns.** Sponsored-ads attributed sales deduct only payment failures and cancellations within 72 hours (section O); Amazon's economics dataset nets refunds out of product sales. The two Amazon "sales" figures a margin can be built on differ by exactly the returns.
+10. **Points (JP).** Amazon's pricing model subtracts points from the landed price used in its competitive comparisons, while the seller pays for them as a cost on top of the sale (section P). No Amazon page read says whether ad-attributed sales subtract points.
+
+### O. Returns as a margin input (added in the gap fill)
+
+- **RULE (what attributed sales deduct).** Amazon's list of why attributed sales differ from retail sales includes: "Attributed sales don't reflect any discounts applied in the cart." and "Payment failures and purchases that are cancelled within 72 hours of order creation will be deducted from attribution metrics. These deductions may take up to 7 days to process after the cancellation." Returns are not on the list. [official] *Ad campaign attribution* (help GX7KDKHMWQYMJ385), https://advertising.amazon.com/help/GX7KDKHMWQYMJ385 (the body is served inside the page's HTML; the 72-hour and 7-day figures are for human readers only). Whether a sponsored-ads sale returned later is ever removed is not stated on any Amazon page read (owner of the attribution rule: `beyond/metrics-and-attribution.md`, its A6, A7 and open question 5).
+- **RULE (Amazon's own net-of-refund sales figure).** In the Data Kiosk economics dataset, `netProductSales` is "calculated by orderedProductSales - refundedProductSales"; `refundedProductSales` is "The amount of refunded product sales for the selected time period, it's counted by refund request time"; `netUnitsSold` is "unitsOrdered - unitsRefunded". `NetProceeds` is "calculated as sales.netProductSales - sum of fees.charge.totalAmount - sum of ads.charge.totalAmount - (cost of Merchant Fulfilled Network (MFN) products or cost of Fulfillment by Amazon (FBA) products) * sales.netUnitsSold", aggregated by `CHILD_ASIN`, `PARENT_ASIN` or `MSKU` and by `DAY`, `WEEK`, `MONTH` or `RANGE`, with a start date "no more than two years ago". The ads part is `AdSummary` ("Advertisement type name, for example: "Sponsored Products charge""). The per-unit cost fields (`costOfGoodsSold`, `fbaCost.shippingToAmazonCost`, `miscellaneousCost`, MFN costs) are "provided by the seller" in SKU Central or the Revenue Calculator and are null otherwise. [official] *analytics_economics_2024_03_15.graphql*, as above.
+- **RULE (per-ASIN return rate).** The Data Kiosk Seller Analytics dataset has a `returns` group: `returnedUnits` ("Number of units returned by customers with a resolution (refund or replacement), including returnless refunds, mapped to the day each unit was shipped"), `totalRefundAmount` and `returnedRate` ("returned units / shipped units ... Not supported under totals"). "Return data is computed with a delay of about 45 days, so the most recent ~45 days may be empty or incomplete", and `returnedRate` "may appear artificially low for recent dates". A `defects` group gives return-defect topics over "a trailing six-month window". [official] *analytics_sellerAnalytics_2025_03_31.graphql*, https://github.com/amzn/selling-partner-api-models/blob/main/schemas/data-kiosk/analytics_sellerAnalytics_2025_03_31.graphql; *Seller Analytics Dataset Use Case Guide* (updatedAt 2026-09-30), https://developer-docs.amazon/sp-api/docs/seller-analytics-dataset-guide.md
+- **RULE (the returns reports).** `GET_FBA_FULFILLMENT_CUSTOMER_RETURNS_DATA` "Contains customer returned items received at an Amazon fulfillment center, including Return Reason and Disposition. Content updated daily", with `return-date`, `order-id`, `sku`, `asin`, `fnsku`, `quantity`, `detailed-disposition`, `reason`, `status`, `customer-comments`; "This report can only be requested." [official] *Fulfillment by Amazon (FBA) Reports*, as in section K. Seller-fulfilled returns are in `GET_XML_RETURNS_DATA_BY_RETURN_DATE` and `GET_FLAT_FILE_RETURNS_DATA_BY_RETURN_DATE` (fields include `return_reason_code`, `return_quantity`, `resolution`, `refund_amount`, `label_cost`, `label_to_be_paid_by`; "You can request up to 60 days of data in a single report"); "These reports only cover seller-fulfilled order returns. They do not include: FBA returns ... Returnless refunds". [official] *Returns Reports* (updatedAt 2026-10-28), https://developer-docs.amazon/sp-api/docs/report-type-values-returns.md
+- **RULE (Sales and Traffic refunds are by date only).** In the Sales and Traffic report schema, `unitsRefunded` ("The number of units refunded in the selected time period") and `refundRate` sit in `SalesByDate` ("aggregated by date (across the seller's entire catalog of items)"); `SalesByAsin` carries `unitsOrdered`, `orderedProductSales`, `totalOrderItems` and their B2B versions and no refund field. `refundRate`'s description reads "calculated by dividing unitsOrdered by unitsRefunded", the inverse of what a rate of refunds would be (Amazon's wording; the example value 33.33 and the 0–100 range fit refunds ÷ ordered). [official] *sellerSalesAndTrafficReport.json*, https://github.com/amzn/selling-partner-api-models/blob/main/schemas/reports/sellerSalesAndTrafficReport.json
+- **RULE (returns processing fees exist by category).** Amazon itemises a "Returns Processing Fee for apparel and shoes" and a "High Return Rate Fee, also know as Returns Processing Fee for Non-apparel and Non-shoes" (Data Kiosk, above); the FBA page lists "Returns processing: Charged on orders when Amazon provides a customer with free return shipping" (section C). The category thresholds and rates are on Seller Central help G64LS955WNFT6EDP, a JavaScript shell to a plain GET (*Unreachable*); no threshold is stated here.
+- **RULE (category return insights).** The Customer Feedback API "provides insights from customer reviews and returns"; "Return insights are only available at the browse node level". [official] *Customer Feedback API*, https://developer-docs.amazon/sp-api/docs/customer-feedback-api.md
+
+[seat] What returns do to the ad arithmetic:
+
+- Attributed sales are gross of returns, so ACOS read from an ad report is spend over gross sales. With a return rate r on refunded revenue, spend over kept revenue is ACOS ÷ (1 − r). For r = 0.2 a 30% ACOS is 37.5% of kept revenue.
+- A returned FBA unit also costs whatever the seller cannot recover: a returns processing fee where it applies, an unsellable disposition, and fees Amazon does not refund. Margin before ads per kept unit falls on both sides of the ratio.
+- Data Kiosk's return figures lag about 45 days and are mapped to the ship date, so the last six weeks' return rate reads low; the economics dataset counts refunds "by refund request time". A return rate measured in either needs a window that ended at least the lag ago.
+- The return rate that matters for an ASIN is that ASIN's (Seller Analytics `returnedRate`), not the account's (Sales and Traffic `SalesByDate`).
+
+### P. Amazon Points (Amazon.co.jp) as a seller-funded price lever (added in the gap fill)
+
+- **RULE (who sets and who pays).** Amazon Points are granted "according to the number or rate set by Amazon or by the Amazon marketplace seller" (Amazon's terms, in Japanese: "アマゾンまたはAmazonマーケットプレイスの出品者...が定める付与数または付与率に従い付与されます"). One point is used as one yen ("1ポイントを1円相当分として"). The points base excludes gift wrapping, shipping and cash-on-delivery fees. If an order is cancelled or returned before the points become usable, the pending points lapse. [official] *Amazonポイント規約* (Amazon.co.jp Points terms, nodeId 201998170), https://www.amazon.co.jp/gp/help/customer/display.html?nodeId=201998170
+- **RULE (the seller's cost and the tool).** Amazon Japan's 2021 announcement: points are set in "Amazonポイントセントラル" in Seller Central, with an all-listings rate and per-listing rates; from November 1, 2021 a default 1% was auto-applied to every listing not set by the seller, changeable "at any time to any % (maximum 50%)" ("いつでも何度でも任意の％（最大50%）に変更可能"); the seller bears "1 yen per point granted" ("1ポイントの付与に対し1円を出品者様にご負担いただいている"); a 9% discount on the cost of the first 1% ran for one year from November 1, 2021. "Amazon Points are a program unique to Amazon.co.jp". [official, Amazon staff in forum] *【重要】ポイント管理ページ「Amazonポイントセントラル」開設等のお知らせ* (poster `News_Amazon`), https://sellercentral.amazon.co.jp/seller-forums/discussions/t/e5d3ea81f020e093d688303ec86c80f5
+- **RULE (Amazon-funded double points).** During weekly "Double Points" weekends (excluding sale events such as Prime Day and Black Friday), Amazon adds points equal to those the seller grants, up to "5% points per item and 5,000 points per order" funded by Amazon; listings with 1% or more show a "ダブルポイント" badge in search results and on the detail page. Eligibility: 1% or more points on at least 80% of buyable listings by the Sunday deadline, held through the event. [official, Amazon staff in forum] *ポイントを付与してダブルポイント祭りで売り上げを伸ばしましょう* (poster `News_Amazon`), https://sellercentral.amazon.co.jp/seller-forums/discussions/t/4c59309f-66c5-49d8-8ce4-8fe25694a560 (posted about three years before 2026-11-09; whether the program runs today was not confirmed on a current page).
+- **RULE (points in Amazon's price comparisons).** In the SP-API Product Pricing model, `LandedPrice` is "calculated by adding ListingPrice + Shipping - Points"; `CompetitivePriceThreshold`: "The offer may be ineligible for the Buy Box if the seller's price + shipping (minus Amazon Points) is greater than this competitive price"; `SuggestedLowerPricePlusShipping` is "including shipping and Amazon Points"; `Points` holds `PointsNumber` and `PointsMonetaryValue`. [official] *productPricingV0.json*, https://github.com/amzn/selling-partner-api-models/blob/main/models/product-pricing-api-model/productPricingV0.json. The Product Fees estimate takes `PriceToEstimateFees.Points`; the model says "The Amazon Points program is only available in Japan." [official] *productFeesV0.json*. Since an ad does not display without the Featured Offer (section K), points enter ad serving through the Featured Offer comparison.
+- **Seller-forum reading (not Amazon's).** A seller asked whether a 10% discount and 10% points differ for the seller; other sellers replied that the customer pays the same and only the form differs. [opinion, unnamed sellers] *値引きとAmazonポイント付与の違いについて*, https://sellercentral.amazon.co.jp/seller-forums/discussions/t/9464c5e4-d862-43db-9c14-c5630f2fb2c0. No Amazon reply is on the thread.
+
+[seat] Points in the margin: a points rate p costs the seller p × price per unit sold, on top of fees, while the listed price (and, per Amazon's attribution list, ad-attributed sales, which deduct only cart discounts) stays at the full price. Break-even ACOS on the full price falls by p. Whether referral fees are computed on the price before points is not stated on a page read; the Product Fees estimate accepts points as an input, so an estimate with and without points shows it for a given item. Whether Sponsored Products ads display the points line, and whether JP deal-price rules count points, were not found on any Amazon page read (*Open questions*).
+
+### Q. Price and advertising as two levers on the same sales (added in the gap fill)
+
+- **GUIDANCE (Amazon: price and stock gate the ad).** "your product must be in stock and priced competitively in order to become the Featured Offer, so take into account product pricing and availability when deciding to advertise." [official] *FAQs*, section K. The Featured Offer comparison uses landed price (section P).
+- **GUIDANCE (Amazon: deals and ads together).** "Products with deals or coupons that were advertised with Sponsored Products saw an average 65% increase in sales and 75% increase in units sold compared to when the deal or coupon product was not advertised" (figures for human readers only). [official] *Which products should I add to my campaign?*, section F. Amazon's comparison is advertised against not advertised for a deal product, not a price cut against a bid raise.
+- **RULE (coupons and attributed sales).** "Attributed sales don't reflect any discounts applied in the cart." [official] *Ad campaign attribution*, section O. [seat] A discount applied in the cart leaves attributed sales at the pre-discount price, so ACOS read during a cart-applied promotion is spend over a revenue the seller did not receive.
+- **RULE (Amazon tests prices with switchbacks).** Amazon's pricing scientists write that "all visitors to the Amazon Store at the same time see the same prices for all products — we need to apply experimental treatments to product prices over time"; random-days switchbacks reduce standard error "by 60%", and carryover arises because "if treatment increases a product's sales, recommendation algorithms may recommend that product more often". [research, Amazon-authored] Cooprider and Nassiri, *The science of price experiments in the Amazon Store* (Amazon Science, April 14, 2023), https://www.amazon.science/blog/the-science-of-price-experiments-in-the-amazon-store
+- [seat] The Dorfman–Steiner condition already sets the two levers against each other (*Published research*): with the price at its own optimum (margin = 1 ÷ |price elasticity|), ad spend over sales = advertising elasticity ÷ |price elasticity|. The ratio of the two elasticities, not either alone, says how much of a sales goal is cheaper to buy with ads than with price. A price cut also moves the referral fee, can move the FBA fee band (section C), changes the Featured Offer comparison, and changes the conversion rate in the click chain; a bid raise changes CPC and placement. Both move sales rank where rank follows sales (`areas/ads-organic-and-total.md`).
+
+## What practitioners add
+
+All [opinion]. Each is the named firm's or person's claim, as written on the page read 2026-11-09. Thresholds and bands below are theirs and are recorded for human readers only.
+
+### Break-even: what goes into "margin before ads"
+
+- **My Amazon Guy** (Ken Zhou, Chief Operating Officer; published 2026-10-13, page metadata modified 2026-10-31), *How to Find the Ideal True ACoS on Amazon PPC and Set Campaign Targets*, https://myamazonguy.com/advertising/how-to-find-the-ideal-true-acos-on-amazon-ppc/:
+  - "Start with your selling price, then subtract your landed product cost, the referral fee, the FBA fee, and an allowance for returns and storage. Leave ad spend out of this step".
+  - "Break-even ACoS is the point where an ad sale earns zero profit, and it equals your margin before ad costs."
+  - It names as a common mistake "using gross margin instead of contribution margin".
+  - "Peak events raise both click costs and storage fees, so rerun your break-even ACoS with those higher numbers before you increase budgets."
+  - "Rufus Sponsored Prompts became billable under existing CPC bids on March 25, 2026 , so that spend already belongs in your true ACoS total. Add Amazon DSP spend at the brand level too".
+- **Canopy Management** (Chuck Kessler, 2025-10-10, "Last Updated: October 2025"; page metadata modified 2026-08-20), *The Ultimate Guide to Amazon ACoS and TACoS*, https://canopymanagement.com/ultimate-guide-to-acos-and-tacos/: "Your product sells for $50. After product costs ($20) and Amazon fees ($12), your unit margin is $18." ... "Any ACoS below 36% makes you money." (the page shows "Break-even ACoS = (18 / 50) x 100 = 36%" between the two) "Keep high-margin products in separate campaigns from low-margin products. This lets you set appropriate ACoS targets for each".
+- **EightX** (CFO firm), *Average Amazon TACOS Benchmarks by Category (2026)*, https://eightx.co/blog/average-tacos-benchmarks: "Break-even TACOS equals your pre-ad gross margin. For a $30 product with $8 COGS, a $4.50 referral fee, and a $5.50 FBA fee, break-even TACOS is 40%. Your sustainable target is that number minus your net-margin goal."
+- **Where they differ.** All three subtract referral and FBA fees, which Amazon's Ads Math formula does not name. They differ on returns, storage and other allocations: My Amazon Guy includes an allowance; EightX's example does not.
+
+### From a profit goal to a campaign target
+
+- **My Amazon Guy** (as above) builds the target in three steps:
+  1. margin before ads;
+  2. minus a target net margin ("A net margin of 15% to 30% is a typical range for Amazon FBA sellers"), giving an "ideal true ACoS" (their name for TACoS: "Is true ACoS the same as TACoS? Yes");
+  3. "Divide your ideal true ACoS by the share of total sales that comes from ads."
+
+  Example: a 38.3% margin before ads minus a 28% target net margin leaves about 10.3%. "The same product supports a 29.5% campaign ACoS when ads drive 35% of sales, but only 17.2% when ads drive 60%." Identity: "Your profit after ads equals your margin before ads minus your true ACoS". Per-ASIN targets: "Set a target for each ASIN based on its own margin and stage, then combine them into one brand number weighted by sales."
+- **Perpetua** (help centre; author Ashley Chung per the page's metadata), *Sponsored Products Goal Setup: Targeting Types, Target ACoS, and Budget*, https://help.perpetua.io/en/articles/4859772-sponsored-products-goal-setup-targeting-types-target-acos-and-budget. Perpetua sets targets from history, not from margin:
+  - "Look at your historical ACoS and budget over the past 30 days to set your initial targets."
+  - For brand-based goals: "Your brand name is less expensive, so it should have a lower budget and target ACoS than the unbranded segment."
+  - For advanced goals: "Competitor keywords and ASINs are more expensive so they should have a higher ACoS than the branded & category segments."
+  - "Minimum $10 daily budget".
+- **Where they differ.** My Amazon Guy derives the target from margin and profit goal. Perpetua anchors it on the realised ACoS of the last 30 days and varies it by segment (brand, competitor, category). The two can give very different numbers for the same product.
+
+### TACoS bands and stage
+
+- **Canopy Management** (as above) gives TACoS bands:
+  - "High TACoS (15–25%): Usually indicates new products, aggressive growth phases, or heavy dependence on paid traffic."
+  - "Moderate TACoS (10–15%): Typical for growing brands with decent organic presence."
+  - "Low TACoS (5–10%): Strong organic presence ... This is the goal for mature, established products."
+  - "Very low TACoS (<5%): Either your organic presence is dominant, or you're underinvesting in advertising".
+
+  On launches: "New product launches often require higher ACoS (sometimes 40–50%) to build initial momentum and reviews." "Use aggressive ACoS targets during launches (40–50%) to build velocity, then gradually lower ACoS as organic sales increase." And: "These are guidelines, not rules."
+- **EightX** (as above):
+  - "A healthy Amazon TACOS ... is 10-15% of total revenue for an established seller. New launches run 25-40% for the first 1-3 months".
+  - "That is intentional rank investment, not a red flag, as long as the number starts falling by weeks 6-10 and is clearly below the launch peak by month 4."
+  - "Category sets your floor. Electronics and grocery brands run 7-13% TACOS at maturity; supplements and beauty structurally carry 14-20%".
+  - Its sources, by its own note: "Autron 2026 Amazon Advertising Benchmarks (CPC, CVR, ACOS); implied TACOS ranges from Titan Network 2026. No single public source reports both ACOS and TACOS by category".
+- **My Amazon Guy** (as above): a stage table of "Starting True ACoS": Launch 15% to 25% ("Build rank"), Scaling 10% to 15% ("Grow organic share"), Mature 7% to 10% ("Protect margin"), Harvest "Under 7%" ("Keep sales moving"). "Treat these ranges as starting points, not rules".
+- **Jungle Scout** (Brian Connolly, published 2023-05-24, page metadata modified 2026-07-22), *What Is Amazon TACoS? Total Advertising Cost of Sale in 2023*, https://www.junglescout.com/blog/amazon-tacos/: TACoS is "ad spend relative to an Amazon seller's total sales revenue, including ad sales and organic sales". "A decrease in TACoS means you spend a smaller proportion of your total sales on advertising costs ... and/or your organic sales are increasing." Its rank claim, stated as illustration: "The more sales your ads bring in, the higher your Amazon Best Sellers Rank and organic keyword ranking will be for your product." Whether that holds is `areas/ads-organic-and-total.md`'s.
+- **Where they differ.** For a launch, Canopy's 15–25% "high" TACoS band, EightX's 25–40% (and "up to 50%") and My Amazon Guy's 15–25% "true ACoS" do not agree. None of the three publishes its underlying data; EightX names its sources as vendor benchmarks.
+
+### The click chain as practitioners use it
+
+- **Ad Badger** (video transcript, presenter "Mike"), *Calculating a Bid Strategy for Amazon Ad Campaigns*, https://www.adbadger.com/blog/calculating-amazon-bid-strategy/: "It is average order value times your conversion rate, divided by one over your Target ACoS, equals what you should be paying every single click." Example: "$19 times 0.10, over 1 over 0.3, giving us 57 cents". It also notes the drift: "the next day your conversion rate is going to change", and "the two-day conversion delay that Amazon sponsored products have".
+- **AdLabs** (help centre; author Stephen Noch per the page's metadata, and the text answers a reader addressing "Stephen"), *What is Revenue Per Click (RPC) Bidding?*, https://help.adlabs.app/en/articles/1-what-is-revenue-per-click-rpc-bidding: "Target CPC = RPC × Target ACOS%". "You can still bid aggressively and rank with RPC. You only need to set a higher Target ACOS for ranking keywords." Its warning on bids: "If you only decrease your bids WITHOUT increasing campaign bidding adjustments for Top of Search, you will destroy your ads by forcing them to worse performing placements".
+- **Ad Badger on what a bid costs** (byline "The Badger", 2026-09-10), *Amazon Says Its Ad Auction Was Second-Price. The FTC Says You Paid Your Own Bid Anyway.*, https://www.adbadger.com/blog/amazon-says-its-ad-auction-was-second-price-the-ftc-says-you-paid-your-own-bid-anyway/:
+  - The two public statistics measure different things: ""Not the highest bid" is about who wins ... "Paying your own bid" is about what you are charged after you win."
+  - "you cannot infer your auction mechanics from Amazon's aggregate statistics, and you cannot infer them from the FTC's either. You can only infer them from your own account."
+  - Its reading: "divide average CPC by your current max bid. Ratios clustering at 0.90–1.00 mean you are effectively in a first-price auction."
+  - Its test: "Cut bids by 15%. Wait 14 days. Check what happened to CPC."
+- [seat] Ad Badger's and AdLabs' bid formulas set the *bid* equal to target ACOS × revenue per click. The target is hit only if paid CPC equals the bid. Under a second-price outcome the paid CPC is lower, so the realised ACOS undershoots the target. If the bid is paid in full (Amazon's own description when the bid is under the soft reserve), the target is met. Under `SALES_UP_AND_DOWN` the effective bid can be above the base bid. The formulas are an identity on paid CPC, not on bid.
+
+### Profit, not ACOS
+
+- **Optmyzr** (Ronia, Content Strategist; 2026-03-16), *Why Your Amazon ACoS Looks Fine—But Profit Is Still Declining (And How to Diagnose It)*, https://www.optmyzr.com/blog/why-amazon-acos-looks-good-but-profit-is-declining/: "ACoS only measures spend versus sales.. It doesn't tell you whether those sales are actually improving your profitability." It lists "Rising CPCs, shifting ad placements, over-reliance on branded terms, or budget overlap between campaigns" as causes, and "Account-level ACoS often hides product-level problems." On branded terms: "Many shoppers who search for your brand already know what they want or were influenced earlier through organic listings, reviews, or external traffic."
+- **Nectar** (Jason Landro, on Amazon's site; section F): contribution margin per unit and NTB CAC instead of a ROAS target.
+
+### Inventory-aware spend
+
+- **Helium 10** (Lauren Stair, 2026-04-01), *Sync Ad Spend with Stock Levels Across Marketplaces*, https://www.helium10.com/blog/how-to-sync-ad-spend-with-stock-levels-across-marketplaces/:
+  - "Inventory-aware advertising means your spend decisions change based on what you can reliably fulfill, not only what looks efficient in-platform."
+  - It defines inventory states: "Healthy: safe to scale demand generation"; "Watch: scale cautiously and prep substitutes"; "Low: throttle acquisition and shift budget to in-stock alternatives"; "Critical: pause demand generation and protect only essential traffic"; "Overstock: increase demand generation with controlled offers".
+  - "The exact thresholds will vary by lead times and velocity".
+  - "The move that saves profit is not pausing ads when you stock out. The move is redirecting demand before the stockout".
+  - "Treat stockouts and forced discounting as costs, not as operations problems."
+  - "If your catalog is constrained by storage or inbound rules, those constraints should influence your thresholds, especially under restock limits."
+- **Teikametrics** (help centre), *Inventory Optimization Overview*, https://help.teikametrics.com/en/articles/8613831-inventory-optimization-overview: the tool layers "inventory data with sales data" for "3P FBA and WFS held products". It offers a "Demand Forecast detailing anticipated remaining days of inventory remaining", "Excess Inventory detailing products that have no sales in the past 90 days", and out-of-stock cards with "estimated missed sales" and "active linked campaigns".
+- **Where they differ.** Helium 10 frames inventory as a state machine that drives budget. Teikametrics' help page frames it as insight cards beside the campaigns. Neither publishes thresholds as universal.
+
+### Returns in the margin (added in the gap fill)
+
+- **Nova Analytics** (Antoine, CEO; published April 3, 2026, updated October 8, 2026), *Amazon Break-Even ACoS*, https://novadata.io/resources/blog/amazon-break-even-acos-guide:
+  - "Amazon's Campaign Manager shows you ACoS, but it doesn't show you your break-even point. It doesn't know your COGS, your actual FBA fees per unit, your return rate, or the dozen other costs eating into your margin."
+  - Its worked example takes "All Other Fees Storage, returns, fuel surcharge, etc." off the price before ads, giving a 22.4% break-even ACoS on a $24.99 product.
+  - A second example itemises "Returns Processing (est.) $0.48 Based on 8% return rate, $6 per return" and "Refund Administration $0.30 Amazon keeps 20% of referral on returns". The 20% figure is Nova's; no Amazon page read here states it (*Open questions*).
+  - A category table, "based on Feedvisor marketplace data and Amazon's published fee schedules", pairs average return rates with break-even ACoS, from "Health & Supplements 8% 3-5% 45-55%" to "Clothing & Apparel 17% 15-25% 10-20%".
+- **FiveX** (FiveX Marketplace Intelligence Team, 2026-07-10), *Break-even ACOS for Amazon Ads: the margin line your campaigns must respect*, https://fivex.com/blog/amazon-ads-break-even-acos/: a "Return allowance (5% rate, $2.50 avg processing) −$0.13 Return rate × processing cost per unit" line in the pre-ad margin. In its peak case, "Peak FBA surcharge adds $0.40 per unit (Oct 15 – Jan 14)", "Storage climbs from $0.78 to $2.57 per cubic foot", "Return rates rise in Q4, increasing return allowance", and the recalculated margin is "−$1.04 per unit ... On a campaign with 28% ACoS that looked fine in July." Its figures differ from Amazon's announced $0.32 average peak increase (section C).
+- **EightX** (Ash Kagali, Senior Financial Analyst, June 30, 2026), *Average ACOS by Amazon Category: 2026 Benchmarks*, https://eightx.co/blog/average-acos-by-amazon-category: "Clothing and Apparel runs 42% on AdBadger's Sponsored Products data partly because more than one in five units come back as returns (Autron's separate Sponsored Products dataset puts Apparel at approximately 57%, reflecting its different account composition)".
+- **What the pages read do and do not do.** All three put returns on the margin side, as a per-unit allowance. None of them restates ACOS or TACoS on revenue net of returns. None says attributed sales are gross of returns, which is Amazon's rule (section O). No practitioner page read computes an ACOS on kept revenue.
+
+### Price, coupons and bids as levers (added in the gap fill)
+
+- **Repricer.com** (Colin Palin; published July 29, 2025, last updated September 1, 2026), *Amazon Repricing and PPC: How the Two Work Together to Win More Buy Boxes*, https://www.repricer.com/blog/amazon-ppc-vs-repricing-buy-box-impact/: "Winning the Amazon Buy Box through repricing places your Sponsored Products ads in the highest-converting ad position on the listing page. Higher CTR at the same bid means more conversions per dollar of ad spend, which lowers ACoS." "Repricing and PPC are the same margin equation from two directions." Campaign Manager "does not show is how much of that performance depends on the Buy Box position your repricing is producing." A repricing vendor's claim, with no data shown.
+- **AMALYZE** (unsigned, "Sponsored Success series"), *Price, Cart and Repricers: How They Quietly Wreck Your Amazon PPC*, https://amalyze.com/blog/price-cart-repricer-ppc-impact: "If you have an ACOS target set in your PPC tool, but your repricer is fluctuating the product price between €22.50 and €25.50 to win the Buy Box or match competitors, your ACOS target becomes a moving goalpost." It gives the bid formula "Target Bid = Target ACOS × Conversion Rate × Product Price", so price enters the bid twice: through price and through conversion rate.
+- **Ad Badger** (byline "The Badger", published December 12, 2024, modified September 8, 2026), *How to Use Amazon Coupons to Promote Your Products*, https://www.adbadger.com/blog/amazon-coupons/: the coupon badge is "often more powerful than a simple price drop"; "Coupons all seem to nudge up conversions by about 5 – 10% no matter how big or small the discount is"; "aside from the margin hit you take from offering coupons, you really don't lose anything from keeping coupons on all of your products." No data source is named for the 5–10%.
+- **Quartile** (Sandie Shin, May 26, 2026), *Why the Smartest Brands on Amazon Are Rethinking Promotions in 2026*, https://www.quartile.com/blog/why-the-smartest-brands-on-amazon-are-rethinking-promotions-in-2026: it defines its own TPCS (Total Promotion Cost of Sales), which "accounts for both advertising spend and promotional cost against total revenue". It claims that with ads and promotions coordinated "brands ... see a 15% to 20% reduction in TACoS while growing sales by 10%", with no method stated.
+- **What was not found.** The task named Canopy Management, Helium 10 and Pacvue. No page of theirs read here sets a price cut against a bid raise for the same sales goal. The pages read treat price as an input to conversion rate and Featured Offer eligibility. None estimates a price elasticity or an advertising elasticity for the seller, and none compares the two levers' cost per incremental unit.
+
+### Disagreements a skill should know exist
+
+1. **What margin before ads includes.** Returns and storage allowances: My Amazon Guy yes, EightX's worked example no. Amazon's Ads Math names only cost of goods.
+2. **Launch tolerance.** Launch ACOS or TACoS bands differ by firm (15–25% to 25–40%, "up to 50%", ACoS "40–50%").
+3. **Target source.** Margin-derived (My Amazon Guy, EightX) or history-derived (Perpetua).
+4. **Whether rank justifies over-break-even spend.** Canopy, EightX, AdLabs and My Amazon Guy say a launch or ranking phase justifies it. Amazon's pages say new campaigns have high ACOS "merely because of their newness" (section A) and do not tie it to rank. The rank mechanism is `areas/ads-organic-and-total.md`'s.
+5. **Where returns go.** Practitioners read put returns into margin before ads as an allowance (Nova, FiveX, My Amazon Guy). Amazon leaves them in attributed sales (section O). [seat] Break-even ACOS on gross attributed sales is right only if the allowance covers what a return loses (the refunded price less what is recovered, plus fees and processing), not the processing cost alone.
+
+## Published research
+
+All [research]. For each: what was measured, on what data, and how far it carries to an Amazon seller's ads. The transfer notes are [seat].
+
+### Optimal advertising: the Dorfman–Steiner condition
+
+- **What it says.** "The optimal level of advertising for a firm is found where the ratio of advertising to sales equals the price-cost margin times the advertising elasticity of demand. The obvious result is that the greater the degree of sensitivity of quantity demanded to advertising and the greater the margin on the extra output then the higher the level of advertising." Source: Dorfman, Robert, and Peter O. Steiner (1954), "Optimal Advertising and Optimal Quality", *American Economic Review* 44, 826–36. [research] *Dorfman–Steiner theorem* (Wikipedia, last edited 2023-08-23), https://en.wikipedia.org/wiki/Dorfman%E2%80%93Steiner_theorem. The original article was not read (JSTOR not tried).
+- **Transfer** [seat]. Written with Amazon's own quantities, the condition is: ad spend ÷ total sales = m × ε. Here m is the margin before ads (price minus marginal cost, over price) and ε is the elasticity of *total* sales with respect to ad spend. Three consequences follow by algebra:
+  1. The left side is TACoS, not ACOS: the condition is about all sales the spend moves.
+  2. The condition is a marginal one. It is equivalent to m × (dS/dA) = 1: the last dollar of spend returns exactly 1/m in total sales, which is Amazon's break-even ROAS applied to the *marginal* dollar.
+  3. Because measured advertising elasticities are well below 1 (next subsection), m × ε is a small fraction of m. A TACoS equal to the margin (all margin spent on ads) is far above the profit-maximising level whenever ε is small.
+
+  Limits: the model is static (one period, no carry-over); it takes the elasticity as known and constant; and it ignores ranking dynamics, capacity, and the auction.
+
+- **Why a seller's own regression does not give ε (added in the gap fill).** The condition needs the *causal* elasticity. A seller regressing sales on its own spend gets a biased one, because spend is set with knowledge of demand. Budgets are raised for peaks, cut at stock-outs, and Amazon's own dynamic bids move "in proportion to the likelihood of a conversion" ([official] *Bidding strategies for Sponsored Products*, https://advertising.amazon.com/help/GCU2BUWJH2W3A8Z7, re-read 2026-11-09). The research says this in so many words:
+  - Shapiro, Hitsch and Tuchman (2021): "In general, advertising is not randomly assigned, and thus, in the presence of unmeasured confounders, the estimated advertising effects do not have a causal interpretation." Their identification "relies on the specific institutions of the ad buying process". "As much as 80% of advertising is purchased in an upfront market well in advance of the ads being aired." They assume "brands cannot observe" local, transient demand shocks "at the time of purchasing an ad spot". As a robustness check they use "changes in advertising at TV market borders", which "requires that consumers who are exposed to advertising do not cross the border and purchase in the neighboring DMA", and find the border results "nearly identical" to the baseline. With that identification "the mean and median of the distribution of estimated long-run own-advertising elasticities are 0.023 and 0.014, respectively, and two thirds of the elasticity estimates are not statistically different from zero." [research] the article's PDF, *Econometrica* 89(4) 1855–1879, read via a public copy, https://www.gwern.net/doc/economics/advertising/2021-shapiro.pdf (ligatures are lost in the text extraction; quotes restored).
+  - Lewis and Rao (2015): "selection bias, due to the targeted nature of advertising, is a crippling concern for widely employed observational methods." [research] as below; abstract via OpenAlex, https://api.openalex.org/works/doi:10.1093/qje/qjv023
+  - Gordon, Zettelmeyer, Bhargava and Chapsky (2019), "A Comparison of Approaches to Advertising Measurement: Evidence from Big Field Experiments at Facebook", *Marketing Science* 38(2) 193–225: "Unobservable factors make exposure endogenous". Over "15 U.S. advertising experiments at Facebook comprising 500 million user-experiment observations", "The observational methods often fail to produce the same effects as the randomized experiments, even after conditioning on extensive demographic and behavioral variables." [research] EconPapers record, https://econpapers.repec.org/RePEc:inm:ormksc:v:38:y:2019:i:2:p:193-225
+  - Gordon, Moakler and Zettelmeyer (2022/2023), "Close Enough? A Large-Scale Exploration of Non-Experimental Approaches to Advertising Measurement": across "663 large-scale experiments at Facebook" with "over 5,000 user-level features", double/debiased machine learning and stratified propensity score matching both fail: "The median RCT lifts are 29%, 18%, and 5% for the upper, middle, and lower funnel outcomes, respectively. Using DML (SPSM), the median lift by funnel is 83% (173%), 58% (176%), and 24% (64%)"; "we are unable to reliably estimate an ad campaign's causal effect." [research] arXiv 2201.07055, https://arxiv.org/abs/2201.07055
+- **The experimental route.** Randomising the spend removes the selection problem:
+  - On a single seller's account, switching ads on and off by day. Jonathan Hershaff (a seller writing on his own Substack, February 13, 2026) did it "alternating my ads "On" and "Off" daily for about a month (post-Christmas late Dec 2025 – Jan 2026)". He reports a "Platform ROAS: ~2.35" against an "Incremental ROAS: 1.25" from the on/off difference in total sales, and calls 1.25 "likely my performance floor" because "A daily switchback assumes marketing has zero memory". He then fits an MMM (Google Meridian) to his switchback data for carryover. [opinion, one seller's account, figures for human readers only] *Estimating Amazon Ad Incrementality for Sellers: How to Combine Switchback Tests with MMM*, https://whatstheimpact.substack.com/p/estimating-amazon-ad-incrementality. Owner of the method: `method/change-measurement.md`; owner of the incrementality reading: `areas/ads-organic-and-total.md`.
+  - Amazon's experimenters use the same design family. Masoero, Missault, Delbé, Richardson and Imbens (2025), "Regular balanced switchback designs for robust multi-unit online experimentation": "a fixed fraction of items receive treatment at each time step, and all items receive treatment equally" over time; "40-60% reduction in standard errors compared to simple item randomization and 20-30% improvement over unbalanced switchback approaches", "unbiased estimation of average treatment effects under carryover effects". [research, Amazon-authored] https://www.amazon.science/publications/regular-balanced-switchback-designs-for-robust-multi-unit-online-experimentation
+  - Jain, Hut, Islam and Pan (2023), "Cross-unit spillovers in A/B testing: Empirical evidence from ads": "in the context of advertiser experiments, a treatment may affect the outcome of control advertisers who compete with treated advertisers in ad auctions. Such spillover effects lead to biased treatment effect estimates under simple advertiser randomization due to the violation of the required Stable Unit Treatment Value Assumption (SUTVA)." [research, Amazon-authored] https://www.amazon.science/publications/cross-unit-spillovers-in-a-b-testing-empirical-evidence-from-ads. [seat] For one seller's own on/off test, the analogue is the seller's own campaigns competing with each other and with the seller's organic listing, so the measured lift is the account's, not a campaign's.
+  - Lewis and Wong ("Incrementality Bidding and Attribution", below) put randomisation inside bidding itself.
+- **Transfer** [seat]. ε in the condition is a causal, total-sales, marginal elasticity. A regression of a seller's sales on its own spend estimates none of the three: spend follows demand, attributed sales are not total sales, and an average over a period is not a margin. Small ε also means Lewis and Rao's noise point applies: an on/off test on one account measures the lift with a wide band.
+
+### How large advertising effects are: elasticity meta-analyses and large-scale ROI studies
+
+- **Sethuraman, Tellis and Briesch (2011).** "How Well Does Advertising Work? Generalizations from Meta-Analysis of Brand Advertising Elasticities", *Journal of Marketing Research* 48(3). Meta-analysis of 751 short-term and 402 long-term brand advertising elasticities from 56 studies, 1960–2008. Findings:
+  - mean short-term elasticity 0.12 (against 0.22 in the prior meta-analysis);
+  - a decline over time;
+  - elasticity higher for durable than non-durable goods, higher "in the early phase of the life cycle rather than in maturity", higher with annual than quarterly data, and higher when advertising is measured in GRPs than in money;
+  - mean long-term elasticity 0.24 (prior 0.41).
+
+  [research] IDEAS/RePEc record hal-02055301, https://ideas.repec.org/p/hal/journl/hal-02055301.html. That record is the French-language version of the article ("À quel point la publicité est-elle efficace? ...", *Recherche et Applications en Marketing*, 2011, DOI 10.1177/076737011102600204); its abstract carries the figures above. The JMR article's own page was not read (see *Unreachable*). Transfer [seat]: mostly offline brand advertising on aggregate sales, not search ads on a marketplace. The life-cycle finding is the one an Amazon launch argument can cite: the same spend moves sales more for a new product than a mature one.
+- **Shapiro, Hitsch and Tuchman (2021).** "TV Advertising Effectiveness and Profitability: Generalizable Results From 288 Brands", *Econometrica* 89(4), 1855–1879. Abstract: "Our results reveal substantially smaller advertising elasticities compared to the results documented in the literature, as well as a sizable percentage of statistically insignificant or negative estimates." "The ROI analysis shows negative ROIs at the margin for more than 80% of brands, implying over‐investment in advertising by most firms. Further, the overall ROI of the observed advertising schedule is only positive for one third of all brands." [research] EconPapers record, https://econpapers.repec.org/RePEc:wly:emetrp:v:89:y:2021:i:4:p:1855-1879. Transfer [seat]: TV, not search. But the study measures the gap this file cares about: ROI *at the margin* against ROI of the *whole schedule*. A schedule can be profitable on average while its last increment loses money. The 80%/one-third split shows the gap is common, not theoretical.
+- **Blake, Nosko and Tadelis (2015).** "Consumer Heterogeneity and Paid Search Effectiveness: A Large-Scale Field Experiment", *Econometrica* 83, 155–174 (NBER w20171). Abstract: "returns from paid search are a fraction of conventional non-experimental estimates. As an extreme case, we show that brand-keyword ads have no measurable short-term benefits. For non-brand keywords we find that new and infrequent users are positively influenced by ads but that more frequent users whose purchasing behavior is not influenced by ads account for most of the advertising expenses, resulting in average returns that are negative." [research] NBER, https://www.nber.org/papers/w20171. Transfer [seat]: eBay on Google and Bing, for a famous brand. The parts that carry over are mechanisms, not magnitudes:
+  - attributed is not incremental;
+  - the incremental part sits with new and infrequent buyers, which is the economic reason a new-to-brand metric matters;
+  - average return can be negative while some segments are positive.
+
+  Measuring this in a seller's account is `areas/ads-organic-and-total.md`'s and `method/change-measurement.md`'s.
+- **Lewis and Rao (2015).** "The Unfavorable Economics of Measuring the Returns to Advertising", *Quarterly Journal of Economics* 130(4), 1941–1973. Twenty-five field experiments: "The median confidence interval on return on investment is over 100 percentage points wide"; "a coefficient of variation of 10 is common"; "informative advertising experiments can easily require more than 10 million person-weeks"; "selection bias, due to the targeted nature of advertising, is a crippling concern for widely employed observational methods." [research] EconPapers record, https://econpapers.repec.org/RePEc:oup:qjecon:v:130:y:2015:i:4:p:1941-1973. Transfer [seat]: a single seller's estimate of the return on a spend change is noisy. A target ACOS is itself an estimate with a wide band, which is relevant to how hard a model can lean on a small difference from target.
+- **Simonov, Nosko and Rao (2018).** "Competition and Crowd-Out for Brand Keywords in Sponsored Search", *Marketing Science* 37(2), 200–215. On Bing: "When no competitors are present, we find a positive, statistically significant impact of brand ads of 1%–4%, with larger brands having a smaller causal effect. In this case, the effective "cost per incremental click" is significantly higher than what focal brands typically pay on other keywords." When the brand does not advertise and competitors do, competitors "steal" 18%–42% of clicks, "Under such position effects, we find the return on investment on defensive advertising to be strongly positive." [research] EconPapers record, https://econpapers.repec.org/RePEc:inm:ormksc:v:37:y:2018:i:2:p:200-215. Transfer [seat]: the economics of a branded term depend on whether competitors bid on it. The "cost per incremental click" idea (spend ÷ incremental clicks, not ÷ all clicks) is the general tool. Branded-term decisions are `beyond/brand-defense-and-competition.md`'s.
+
+### Price elasticity, the other lever (added in the gap fill)
+
+- **Bijmolt, Van Heerde and Pieters (2005).** "New Empirical Generalizations on the Determinants of Price Elasticity", *Journal of Marketing Research* 42(2) 141–156. "Across a set of 1851 price elasticities based on 81 studies, the average price elasticity is −2.62." "over the past four decades, sales elasticities have significantly increased in magnitude"; "accommodating price endogeneity has a strong (magnitude-increasing) impact on price elasticities." [research] Crossref record with abstract, https://api.crossref.org/works/10.1509/jmkr.42.2.141.62296. Transfer [seat]: set against the advertising elasticities above (means of 0.12 short-term in the meta-analysis, 0.023 in Shapiro et al.), the price elasticity is one to two orders of magnitude larger. In the Dorfman–Steiner ratio (section Q), that is why the profit-maximising ad-to-sales ratio is small.
+- **Chevalier and Goolsbee (2003; NBER w9085, 2002).** "Measuring Prices and Price Competition Online: Amazon and Barnes and Noble". From sales ranks of about 20,000 books: "demand at Barnes and Noble is much more price-elastic than is demand at Amazon." In the paper: "the own price elasticity of demand is close to -4 and the cross-price elasticity very high" at BN.com, "than at Amazon where the price elasticity is around -0.6"; with one parameter choice "At Amazon ... it is actually less than one in absolute value, at -0.45", which the authors call "important and puzzling". [research] NBER, https://www.nber.org/papers/w9085 (abstract; the paper's PDF read for the figures). Transfer [seat]: books in 2001, Amazon as the retailer, not a third-party seller's single offer among rivals. It shows that Amazon-store demand can be far less price-elastic than a rival store's. That does not hold for every offer.
+- **Bach, Chernozhukov, Klaassen, Spindler, Teichert-Kluge and Vijaykumar (2024).** "Adventures in Demand Analysis Using AI", arXiv 2501.00382. On "toy cars on Amazon.com", text, image and tabular embeddings "lead to more credible causal estimates of price elasticity"; "we uncover strong heterogeneity in price elasticity driven by these product-specific features." [research] https://arxiv.org/abs/2501.00382. Transfer [seat]: price elasticity on Amazon varies by product within one category, so a category figure is not an ASIN's.
+- **Amazon's own price testing.** Amazon runs price experiments as switchbacks because one product has one price at a time (section Q; Cooprider and Nassiri, Amazon Science, 2023).
+- **Transfer** [seat]. No paper read here measures, for one Amazon listing, both its price elasticity and its advertising elasticity. Those are the two numbers the price-against-bid comparison needs.
+
+### Value beyond the first order
+
+- **Chan, Wu and Xie (2011).** "Measuring the Lifetime Value of Customers Acquired from Google Search Advertising", *Marketing Science* 30(5), 837–850. "Results show that customers acquired through Google search advertising in our data have a higher transaction rate than customers acquired from other channels. After accounting for future purchases and spillover to off-line channels, the calculated value of new customers using our approach is much higher than the value obtained using conventional method." Data: one small US firm, individual-level panel. [research] EconPapers record, https://econpapers.repec.org/RePEc:inm:ormksc:v:30:y:2011:i:5:p:837-850. Transfer [seat]: supports Amazon's Ads Math point that judging acquisition by first-order return undervalues it when customers repeat. The size of the effect is one firm's.
+
+### Paid-to-organic spillover through rankings
+
+- **Ju, Zhao and Aral (2025).** "Advertising Spillovers in Mobile Apps: Evidence from Ad Shutoffs and Store Rankings", arXiv 2504.16151. A global ad shutoff by a mobile game developer "decreased organic installs by 20-30%"; "every $100 spent is associated with 32 paid installs and 2.2 organic installs"; "paid installs boost app store category rankings, thereby increasing organic visibility"; "the relationship between organic installs and ad spend disappears once these rankings are factored in". [research] https://arxiv.org/abs/2504.16151. Transfer [seat]: a marketplace whose ranking uses sales could show the same mechanism. If it does, the return on ad spend includes organic sales the ads produce through rank, and that changes the margin-only break-even. Whether and how much Amazon's search behaves this way is `areas/ads-organic-and-total.md`'s.
+
+### Evidence from Amazon's own data
+
+- **Pauwels, Peran, Shah, Schnaidt and Vercamer (2023).** "Sponsored brands video rings up clicks and sales in the short and long run", *Applied Marketing Analytics*; listed on Amazon Science. Using Amazon.com data and causal-inference methods: "brands that invested in Sponsored Brands video (SBv) ads in addition to sponsored ads static ads had a 25% higher click-through rate (CTR) and 10% higher year-over-year sales growth". [research] Amazon Science, https://www.amazon.science/publications/sponsored-brands-video-rings-up-clicks-and-sales-in-the-short-and-long-run
+  - **Identification (read in the gap fill).** The paper is observational. Brands chose whether to adopt SBv; nothing was randomised. For the month horizon the authors "use a two-stage Gaussian process ... (our version of causal multi-task Gaussian processes)" on "a matched sample of 25,364 advertisers in North America and Europe", finding that "brands that launched an SBv campaign for the first time obtained an average 21.3% increase in sales the following month compared with those that did not". For the year horizon "we applied propensity score stratification (Austin 2011) to compare sales growth and CTR". The adjustment uses "the 50+ features we used to account for confounding. The top confounders include the vertical (product category) of the advertiser, the number of enabled campaigns, trailing retail performance metrics, country of origin, and inventory position." The CTR comparison restricts to "consumers who each had both static and video impressions during the same month". Their limitations: "we did not analyze the role of frequency of video ad exposure, nor the impact on advertising efficiency, i.e. Return on Advertising Investment (ROAS)". [research] the paper's PDF, https://cdn.amazon.science/d1/66/f222c8524c4d87fc1b0b342345b7/sponsored-brands-video-rings-up-clicks-and-sales-in-the-short-and-long-run.pdf
+  - **Transfer** [seat]. The estimates are causal only if adoption is as good as random once the 50+ observed features are matched. Brands that start video may differ in ways the features miss, such as a new product, a creative budget, or an agency. Gordon et al. (2019; 2023, above) found matching and propensity methods far from experimental results on Facebook ads even with thousands of features. The figures are therefore associations under selection on observables, from Amazon-authored work, about format mix and sales growth. They are not a ROAS and not an incrementality measurement.
+- **Dash, Ghosh, Mukherjee, Chakraborty and Gummadi (2024).** "Sponsored is the New Organic: Implications of Sponsored Results on Quality of Search Results in the Amazon Marketplace", arXiv 2407.19099. From 4,800 simulated searches in four countries: "items with poor organic ranks (beyond 100th position) appear as sponsored results even before the top organic results on the first page"; "in majority of the cases, these top sponsored results are costlier and are of poorer quality than the top organic results." [research] https://arxiv.org/abs/2407.19099. Transfer [seat]: confirms that ad position is available to products with no organic standing. That is the economic premise of launch spend.
+
+### Bidding, diminishing returns and incrementality
+
+- **Kong, Shmakov and Yang.** "Do not Waste Money on Advertising Spend: Bid Recommendation via Concavity Changes", arXiv 2212.13923. "The recommended bid is derived based on the turning point from significant increase (i.e. concave downward) to slow increase (convex upward)" of the click-prediction curve. [research] https://arxiv.org/abs/2212.13923. Transfer [seat]: the bid → clicks curve flattens, so each extra cent of bid buys fewer extra clicks. This is the mechanism behind marginal ACOS rising above average ACOS as bids rise.
+- **Lewis and Wong.** "Incrementality Bidding and Attribution", arXiv 2208.12809. They propose a method that unifies "ad buying/bidding/pricing, attribution, and experimentation" so bids reflect "advertising's causal effects". [research] https://arxiv.org/abs/2208.12809. Transfer [seat]: the value per click that should enter a bid is the *incremental* value. Amazon's attributed sales are an upper bound on it, not an estimate of it.
+- **Li, Yuan, Yang, Chen and Song (2026).** "Attributed, But Not Incremental: Cannibalization-Corrected Attribution for Large-Scale Advertising", arXiv 2606.26690. "paid-attributed conversions ... may systematically overstate true incremental growth when paid channels overlap with organic demand, brand-driven traffic, or other acquisition channels"; their experiment-calibrated correction, deployed across "multiple global TikTok markets", "supported budget and traffic strategy adjustments that were followed by an approximately 15-percentage-point reduction in the measured cannibalization rate". [research] https://arxiv.org/abs/2606.26690. Transfer [seat]: a different platform, the same structural point about attributed and incremental.
+
+### What the research does not settle
+
+[seat] No paper read here measures, on Amazon Sponsored Products:
+
+- the elasticity of a seller's total sales to SP spend;
+- the marginal ROAS of a bid increase;
+- the share of attributed SP sales that is incremental.
+
+The research gives mechanisms (diminishing returns, attributed ≠ incremental, value beyond first order, ranking spillover, early-life-cycle elasticity) and orders of magnitude from other channels. The numbers for a given seller can only come from that seller's own data and tests.
+
+## Stable enough for a skill
+
+Each item is an Amazon definition, an Amazon rule or API shape, an identity that follows from Amazon's definitions, or a measured effect with its source. Figures stay out of skill text and are read at run time.
+
+1. ACOS is ad spend over attributed sales; ROAS is its inverse; Amazon's break-even rule is that ACOS must be under the profit margin, and break-even ROAS is 1 ÷ margin. (*What is advertising cost of sales (ACOS)?*; *Ads Math*; v1 *Purchases* `metric.roas`.)
+2. Amazon's click chain: CPA = CPC ÷ conversion rate; profit per sale = AOV × margin; a one-off sale is profitable when profit per sale exceeds CPA. This is the same test as ACOS < margin. (*Ads Math*.)
+3. Amazon's lifetime rule: customer LTV = AOV × average orders; affordable CPA = LTV × margin; Amazon calls affordable CPA "the upper limit for your ad spend per new customer". (*Ads Math*.)
+4. Amazon's guides say there is no universal good ACOS or ROAS, and say new campaigns run high ACOS "merely because of their newness". One Amazon FBA guide nonetheless prints a $2–4 per $1 benchmark. (*ACOS* guide; *Ads Math*; *How to Help Increase Sales with FBA and Advertising*.)
+5. Amazon's fees that scale with a sale are the category referral fee (a percentage with a minimum, on the total price including shipping and gift wrap), a media closing fee, and the FBA fulfillment fee based on price, weight and dimensions. FBA also charges storage, aged inventory over 181 days, returns processing, inbound placement and removal or disposal. (*Selling on Amazon fees*; *Fulfillment by Amazon (FBA)*.)
+6. The SP-API Product Fees API returns an estimate at a price the caller supplies, per SKU or ASIN, with `TotalFeesEstimate` and a `FeeDetailList`. Amazon says the estimates "are not guaranteed", and ASIN estimates use catalog size. The FBA Fee Preview report gives current per-SKU fee estimates, refreshed at least every 72 hours and requestable once a day. Actual charges are in the Finances API. (*Product Fees API*; *productFeesV0.json*; *FBA Reports*; *finances_2024-06-19.json*.)
+7. A bid is the maximum a click can cost. Amazon states that when the bid exceeds the soft reserve the winner pays the soft reserve, and when it does not the winner pays the bid. The FTC alleges advertisers paid their own bid about 80% of the time by 2024. (*FAQs*; Amazon's FTC response; FTC press release.)
+8. In Ads API v1, Sponsored Products campaigns carry no goal settings. Sponsored Brands campaigns carry a read-only goal and a required KPI. The v1 `Goal`, `KPI` and `BidStrategy` enums are read from the model. (*The campaign entity*; v1 merged OpenAPI.)
+9. Sponsored Brands has two goals: drive page visits (clicks, CPC) and grow brand impression share (top-of-search impression share, vCPM). (*Sponsored Brands goal-based campaigns*.)
+10. SP rule-based bidding applies a ROAS guardrail that Amazon does not guarantee and disables under a stated condition; eligibility needs campaign age, recent conversions and a minimum daily budget. (*Sponsored Products rule-based bidding overview*.)
+11. New-to-brand means no purchase from the brand in the prior 12 months. In v1, NTB purchases and sales are SP-eligible, but NTB ROAS is not. In v3, NTB columns appear in no SP report type. `costPerNewToBrandPurchase` divides total cost by NTB purchases. (v1 *Purchases (new to brand)*; v3 columns.)
+12. Long-term sales (attributed sales plus estimated 365-day value of new-to-brand engagements) and long-term ROAS exist for SB, SD, ST, DSP and Full-Funnel, not for SP. S&S subscription metrics are likewise not SP-eligible. (v1 *Long-term sales*; v1 *Subscribe & Save subscriptions*.)
+13. Seller-side repeat-purchase history is in the Brand Analytics Repeat Purchase report, which needs Brand Registry. Subscribe & Save history, including subscriber lifetime value by segment, retention and out-of-stock losses, is in the Replenishment API. (*Analytics Reports*; *Replenishment API*; model.)
+14. An ad does not display when the product is out of stock or not the Featured Offer. (*FAQs*.)
+15. FBA capacity limits depend on IPI and sales; usage counts on-hand plus inbound units; limits can be raised through Capacity Manager at a seller-set reservation fee. The IPI counts excess (over 90 days of supply), sell-through, stranded inventory and in-stock rate of popular products. (Amazon staff forum posts, 2024-03-14 and 2026-08-13.)
+16. Amazon's inventory planning and restock reports give per-SKU days of supply, weeks of cover, excess quantity, aged-inventory exposure, low-inventory-level fee flags, seasonality flags and recommended actions. (*FBA Reports*.)
+17. Brand advertising elasticities average about 0.12 short term and 0.24 long term, and are higher early in the product life cycle. TV ad ROI is negative at the margin for most brands even where the average schedule is profitable for some. (Sethuraman, Tellis and Briesch 2011; Shapiro, Hitsch and Tuchman 2021.)
+18. Paid-search field experiments find attributed returns far above causal returns, with the incremental effect concentrated on new and infrequent buyers. (Blake, Nosko and Tadelis 2015; Lewis and Rao 2015.)
+19. The Dorfman–Steiner condition sets the profit-maximising ratio of ad spend to sales at margin × advertising elasticity. This is equivalent to a marginal ROAS of 1 ÷ margin on total sales. (Dorfman and Steiner 1954, via the Wikipedia statement; the equivalence is algebra.)
+20. Amazon announces US fee changes yearly, with most taking effect January 15. The 2027 letter (November 5, 2026) also has February 1, 2027 changes: the storage utilization surcharge ends, the aged-inventory surcharge merges into an age-based monthly storage rate, and the low-inventory-level fee changes its test. Holiday peak fulfillment fees run October 15 to January 14 (2026–27). A 3.5% fuel and logistics surcharge on fulfillment fees has applied since April 17, 2026. (2026 and 2027 fee letters; `News_Amazon` posts.)
+21. Sponsored-ads attributed sales deduct payment failures and cancellations within 72 hours, and do not reflect cart discounts. No Amazon page read says returns are deducted. (*Ad campaign attribution*.)
+22. Per-ASIN return rates are in Data Kiosk Seller Analytics (`returns.returnedRate`, mapped to ship date, about 45 days late) and can be built from the FBA customer returns report and the seller-fulfilled returns reports. Sales and Traffic gives refunds by date only. (Data Kiosk schemas and guide; *FBA Reports*; *Returns Reports*; Sales and Traffic schema.)
+23. Data Kiosk's economics dataset gives per-ASIN or per-SKU net proceeds after refunds, Amazon fees, ad charges and seller-entered unit costs. It splits fees at an Amazon fee-change date and previews fees up to 120 days ahead. (*analytics_economics_2024_03_15.graphql*.)
+24. On Amazon.co.jp, seller-set points cost the seller one yen per point and are set in Points Central (maximum 50%). Amazon's pricing model subtracts points from landed price in its competitive-price comparisons. (JP `News_Amazon` posts; JP Points terms; Product Pricing model.)
+25. Observational estimates of ad effects (matching, propensity scores, DML), even with thousands of features, often miss experimental results. Spend chosen with knowledge of demand biases a seller's own regression of sales on spend. (Gordon et al. 2019; Gordon, Moakler and Zettelmeyer 2023; Lewis and Rao 2015; Shapiro, Hitsch and Tuchman 2021.)
+26. Measured price elasticities (meta-analytic mean −2.62) are one to two orders of magnitude larger than measured advertising elasticities. Elasticities vary strongly by product on Amazon. (Bijmolt, Van Heerde and Pieters 2005; Bach et al. 2024.)
+
+## What cx must read at run time
+
+Every figure (fee, threshold, minimum budget, window, enum value) is read live. The exact calls belong to the plugins' own references; the operations are named here so a skill knows where each economic input lives.
+
+**Ads API** (through `cx raw amazon-ads`; endpoint details are `method/ads-api-surface.md`'s, report lifecycles `method/reporting.md`'s):
+
+- **Spend and attributed sales per campaign, ad group, target, search term and advertised SKU.**
+  - Reporting v3 `POST /reporting/reports` (which, per the v3 columns page, "will be deprecated on June 30, 2027") with `reportTypeId` `spCampaigns`, `spTargeting`, `spSearchTerm`, `spAdvertisedProduct`, `spPurchasedProduct` (and `sb*`, `sd*` equivalents).
+  - Columns `cost`, `clicks`, `costPerClick`, `purchases7d`, `sales7d`, `attributedSalesSameSku7d`, `salesOtherSku7d`, `unitsSoldSameSku7d`, `acosClicks7d`, `roasClicks7d` (14-day variants where the product uses them). The v3 columns page lists `acosClicks7d` and `roasClicks7d` for `spTargeting`, `spSearchTerm` and `spAdvertisedProduct` (and the prompt and video ad-extension types), not for `spCampaigns` or `spPurchasedProduct`.
+  - Or Ads API v1 reporting (`POST /adsApi/v1/create/reports`, then `POST /adsApi/v1/retrieve/reports`, both in the v1 merged OpenAPI) with `metric.totalCost`, `metric.sales`, `metric.roas`.
+  - [seat] ACOS, CVR, AOV and CPC are best computed from counts and money rather than taken from precomputed rate columns. The listing playbook found two bulksheet rate definitions inverted (*paid-keywords.md*, P-paid-keywords-4, read 2026-10-14). That finding is that file's; this file did not re-read the bulksheets page.
+- **New-to-brand and long-term value.**
+  - v1: `metric.newToBrandPurchases`, `metric.newToBrandSales`, `metric.costPerNewToBrandPurchase`, `metric.percentOfSalesNewToBrand` (SP-eligible).
+  - v1: `metric.longTermSales*`, `metric.longTermRoas`, `metric.cohortAccumulatedSales*`, `metric.subscribeAndSaveSubscriptions` (SB, SD, ST, DSP, Full-Funnel).
+  - v3: `newToBrandPurchases`, `newToBrandSalesClicks`, `longTermSales`, `longTermROAS` in `sb*`, `sd*`, `st*` report types.
+  - Peer benchmarks: `...P25`/`P50`/`P75` fields, which require benchmark dimensions.
+- **What each campaign is set to do.**
+  - The campaign's bid strategy and dynamic-bidding settings (SP v3 `POST /sp/campaigns/list`; v1 `POST /adsApi/v1/query/campaigns` (`QueryCampaign`), with `optimizations.bidSettings.bidStrategy` and, for SB, `optimizations.goalSettings`).
+  - Any rule-based bidding rule and its ROAS guardrail (the rules endpoints are `areas/bids-and-placements.md`'s and `beyond/automation-and-amazon-tools.md`'s).
+  - SB `goal` and `costType`.
+  - Portfolio membership.
+- **Enum values** (`Goal`, `KPI`, `BidStrategy`, `AdProduct`) from the v1 merged OpenAPI (`AmazonAdsAPIALLMerged_prod_3p.json`) and the product models, read when a value matters, since enums gain values (`FULL_FUNNEL`, `ENGAGEMENT`).
+- **Rule-based bidding eligibility and minimum daily budget per marketplace**, from the rule-based bidding overview page.
+
+**SP-API** (through `cx-amazon-sp-api`'s door):
+
+- **Fee estimate at a given price.** Product Fees v0: `getMyFeesEstimateForSKU` (`POST /products/fees/v0/listings/{SellerSKU}/feesEstimate`), `getMyFeesEstimateForASIN`, `getMyFeesEstimates`. Body `FeesEstimateRequest` with `MarketplaceId`, `IsAmazonFulfilled`, `PriceToEstimateFees.ListingPrice` (and `Shipping`), `Identifier`, `OptionalFulfillmentProgram`. Read `TotalFeesEstimate` and each `FeeDetail` (`FeeType`, `FinalFee`).
+- **Current fee estimates for all FBA SKUs.** Report `GET_FBA_ESTIMATED_FBA_FEES_TXT_DATA` (`estimated-fee-total`, `estimated-referral-fee-per-unit`, `expected-fulfillment-fee-per-unit`, `estimated-variable-closing-fee`, future-fee columns where the store has them). Once a day per seller.
+- **Actual fees charged.** Finances v2024-06-19 `listTransactions` (`GET /finances/2024-06-19/transactions`): breakdowns per transaction, lagging up to 48 hours.
+- **Storage and holding costs.** `GET_FBA_STORAGE_FEE_CHARGES_DATA` (`estimated_monthly_storage_fee`), `GET_FBA_OVERAGE_FEE_CHARGES_DATA`, and `GET_FBA_INVENTORY_PLANNING_DATA` (`estimated-storage-cost-next-month`, aged-inventory-surcharge columns).
+- **Total sales for TACoS.** `GET_SALES_AND_TRAFFIC_REPORT` (schema `sellerSalesAndTrafficReport`; `reportOptions` `dateGranularity` `DAY`/`WEEK`/`MONTH`, `asinGranularity` `PARENT`/`CHILD`/`SKU`). Fields `orderedProductSales`, `unitsOrdered`, `totalOrderItems`, `unitSessionPercentage`, `sessions`, `buyBoxPercentage`. Needs the Brand Analytics role; `dataStartTime` no more than two years back.
+- **Stock cover and inventory state.**
+  - `GET_FBA_INVENTORY_PLANNING_DATA`: `available`, `days-of-supply`, `weeks-of-cover-t30`/`t90`, `estimated-excess-quantity`, `sell-through`, `units-shipped-t30`, inventory-age buckets, low-inventory-level fee flags, historical days of supply, `is-seasonal-in-next-3-months`, `recommended-action`, `recommended-sales-price`.
+  - `GET_RESTOCK_INVENTORY_RECOMMENDATIONS_REPORT`: total days of supply including open shipments, alert, recommended replenishment quantity and date.
+  - The FBA Inventory API's inventory summaries.
+  - Capacity limits and the IPI score are shown in Seller Central (Capacity Monitor, Inventory Performance dashboard); no SP-API operation for them was found in the documents read here (*Open questions*).
+- **Repeat purchase and subscription value.**
+  - `GET_BRAND_ANALYTICS_REPEAT_PURCHASE_REPORT` (`orders`, `uniqueCustomers`, `repeatCustomersPctTotal`, `repeatPurchaseRevenue`, `repeatPurchaseRevenuePctTotal`).
+  - Replenishment v2022-11-07 `getSellingPartnerMetrics` / `listOfferMetrics` (`SUBSCRIBER_LIFETIME_VALUE_BY_CUSTOMER_SEGMENT`, `SUBSCRIBER_NON_SUBSCRIBER_AVERAGE_REORDERS`, `SUBSCRIBER_RETENTION`, `REVENUE_PENETRATION`, `LOST_REVENUE_DUE_TO_OOS`) and `listOffers` (offer `EligibilityStatus`, seller funding).
+- **Net proceeds per ASIN, after refunds, fees and ads (added in the gap fill).** Data Kiosk (`createQuery` with a GraphQL query, then `getDocument`) on the economics dataset `analytics_economics_2024_03_15`: `economics` with `sales.orderedProductSales`, `sales.refundedProductSales`, `sales.netProductSales`, `sales.netUnitsSold`, `fees` (by `PreviewFeeType`-style names, with `amountPerUnitDelta` at fee changes), `ads` (`AdSummary`, e.g. "Sponsored Products charge"), `cost` (seller-entered COGS and MFN costs, null when absent) and `netProceeds`, aggregated by `CHILD_ASIN`/`PARENT_ASIN`/`MSKU` and `DAY`/`WEEK`/`MONTH`/`RANGE`, start no more than two years back. `economicsPreview` gives fees for a future window, up to 120 days ahead. Operation names and quotas are `cx-amazon-sp-api`'s.
+- **Return rate per ASIN.** Data Kiosk Seller Analytics `analytics_sellerAnalytics_2025_03_31`, `returns` group (`returnedUnits`, `totalRefundAmount`, `returnedRate`; mapped to ship date; about 45 days late; `returnedRate` not under totals) and `defects` (trailing six months). Event-level: `GET_FBA_FULFILLMENT_CUSTOMER_RETURNS_DATA` (request only; `return-date`, `asin`, `sku`, `quantity`, `detailed-disposition`, `reason`), and for seller-fulfilled orders `GET_FLAT_FILE_RETURNS_DATA_BY_RETURN_DATE` / `GET_XML_RETURNS_DATA_BY_RETURN_DATE` (60 days per request; no FBA returns, no returnless refunds). Account-level refunds by date: Sales and Traffic `salesAndTrafficByDate` `unitsRefunded`, `refundRate`.
+- **Points (Amazon.co.jp only).** Product Pricing v0 offers `Points` (`PointsNumber`, `PointsMonetaryValue`) and `LandedPrice` (price + shipping − points); `CompetitivePriceThreshold` is compared net of points. Product Fees v0 `PriceToEstimateFees.Points` for a fee estimate with points. The seller's points rate per listing is set in Points Central in Seller Central; no SP-API write or read of the rate itself was found in the documents read.
+- **Dated fee changes.** The Fee Preview report's future-fee columns and Data Kiosk's split at fee-change dates show a change per SKU. The Revenue Calculator, Profit Analytics and its Fee Change Comparison dashboard are Seller Central tools with no API found. The yearly letter on sellingpartners.aboutamazon.com and `News_Amazon` forum posts give the dates.
+- **Vendors.** `GET_VENDOR_NET_PURE_PRODUCT_MARGIN_REPORT` is listed among the analytics report types; vendor margin is not this file's scope.
+
+**Keepa** (through `cx-keepa`): price history for the advertised ASIN, so the price behind the margin is the price that actually held over the window (deal and coupon periods change both the referral fee and the margin).
+
+**The seller** (no Amazon source holds these): landed cost of goods per unit, inbound freight to Amazon, packaging, the seller's own allocation of storage, inbound placement and returns, any coupon or deal cost, the points rate on Amazon.co.jp listings where not read from the offer, the seller-funded S&S discount, the profit or growth goal for a product, and its role (launch, core, defend, clear).
+
+**Pages a figure lives on** (read when a figure is needed): *Selling on Amazon fees* (referral table, plans); *Fulfillment by Amazon (FBA)* (cost categories; FBA fee tables are linked from it); *Subscribe & Save | Sell on Amazon* (discount tiers); *Sponsored Products rule-based bidding overview* (minimum budgets); the Seller Central help pages on capacity limits, IPI and the low-inventory-level fee (JavaScript shells to a plain GET; cx-sources' own route to Seller Central help is the plugin's concern).
+
+## Pages read
+
+All read 2026-11-09 by plain GET with a browser user agent, unless the row says otherwise. Cloudfront `.md` and `developer-docs.amazon/....md` rows are Amazon's own markdown copies of its documentation.
+
+| Page name | URL | Type | Date read | What it gave |
+|---|---|---|---|---|
+| What is advertising cost of sales (ACOS)? Calculation and tips | https://advertising.amazon.com/library/guides/acos-advertising-cost-of-sales | official | 2026-11-09 | ACOS and ROAS formulas; break-even ACOS against profit margin; no definitive good ACOS; new campaigns run high ACOS |
+| Practical calculations for holiday marketing success (Ads Math) | https://advertising.amazon.com/library/guides/ads-math | official | 2026-11-09 | Break-even ROAS = 1 ÷ margin; Advertiser A/B table; bid-up example; CPA, profit per sale; LTV and affordable CPA |
+| Advertising budget: Definition, examples, measurement | https://advertising.amazon.com/library/guides/advertising-budget | official | 2026-11-09 | ACOS on ad-attributed revenue; no universal budget formula; start from goals |
+| What is Marketing ROI? Formula & Examples | https://advertising.amazon.com/library/guides/marketing-roi | official | 2026-11-09 | Refine ROI by subtracting organic growth; long-term benefits lag |
+| The author's guide to Sponsored Products | https://advertising.amazon.com/library/guides/authors-guide-to-sponsored-products | official | 2026-11-09 | "TACOS shows your ad spend as a percentage of total sales" (re-read in the gap fill; datePublished 2023-09-07, dateModified 2026-10-09) |
+| Best practices for your Sponsored Products ads | https://advertising.amazon.com/library/guides/sponsored-products-best-practices | official | 2026-11-09 | Objectives list including inventory clearance; group by price range |
+| How to measure and improve your campaigns | https://advertising.amazon.com/library/guides/measure-improve-campaigns | official | 2026-11-09 | Goal-to-KPI pairs; NTB metrics for cost of acquisition; two-week review |
+| Ad Intel: 5 Amazon shopping event advertising tips from Nectar | https://advertising.amazon.com/library/expert-advice/five-advertising-tips-from-nectar | opinion (Nectar, on Amazon's site) | 2026-11-09 | Contribution margin and NTB CAC over a ROAS target; 4x to 3.8x example |
+| How to Help Increase Sales with FBA and Advertising | https://advertising.amazon.com/library/guides/fba | official | 2026-11-09 | "$2–4 for every $1" benchmark |
+| Which products should I add to my campaign? (product strategy) | https://advertising.amazon.com/library/guides/product-strategy | official | 2026-11-09 | Product labels; 11x repurchases; deals and coupons +65%/+75% |
+| Customer loyalty: Drive brand love and repeat purchases | https://advertising.amazon.com/objectives/customer-loyalty | official | 2026-11-09 | How Amazon says loyalty is measured |
+| How to increase sales and conversions for your business (objective page) | https://advertising.amazon.com/objectives/increase-sales | official | 2026-11-09 | Purchases, ROAS and NTB share as the sales-goal measures; little else |
+| Post-event advertising strategy guide | https://advertising.amazon.com/library/guides/post-event-advertising-strategy | official | 2026-11-09 | Quiz-style examples of ROAS falling with broad aggressive bids (datePublished 2025-09-26); not relied on |
+| Campaign reporting: Analyze advertising performance | https://advertising.amazon.com/measurement-analytics/campaign-reporting | official | 2026-11-09 | Long-term sales defined; NTB, halo, S&S named as Amazon metrics |
+| Your guide to Amazon Ads pricing transparency | https://advertising.amazon.com/library/guides/pricing-transparency | official | 2026-11-09 | Bid is a maximum; what ad selection considers; no demand-side fees |
+| FAQs - Advertising questions and resources | https://advertising.amazon.com/resources/faq | official | 2026-11-09 | Bid definition; out of stock or no Featured Offer means no ad; set business goals first; SP uses incl. clearance |
+| Sponsored Brands goal-based campaigns | https://advertising.amazon.com/resources/whats-new/sponsored-brands-goal-based-campaigns | official | 2026-11-09 | Two SB goals, success metrics and cost types (datePublished 2023-10-25) |
+| Introducing Amazon Ads Agent, your AI-powered ad platform | https://advertising.amazon.com/library/news/amazon-ads-agent | official | 2026-11-09 | Full-Funnel and DVA+ goal-led buying (2026-09-29) |
+| unBoxed 2026 news: Explore the latest advertising announcements | https://advertising.amazon.com/library/news/unboxed-2026-news-announcements | official | 2026-11-09 | Full-Funnel beta figures; "one objective, one budget, one campaign" |
+| The campaign entity in the Amazon Ads API | https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/campaign-management/entities/campaign.md | official (API doc) | 2026-11-09 | goalSettings blank for SP; read-only goal and required kpi for SB |
+| Amazon Ads API v1 merged OpenAPI | https://d1y2lf8k3vrkfu.cloudfront.net/openapi/en-us/dest/AmazonAdsAPIALLMerged_prod_3p.json | official (API model) | 2026-11-09 | Goal, KPI, BidStrategy, AdProduct, CostType enums; CreateGoalSettings |
+| Metrics library (v1 reporting) | https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/ads-v1/metrics/library.md | official (API doc) | 2026-11-09 | Index of v1 metric families |
+| Long-term sales (Amazon retail conversions) | https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/ads-v1/metrics/amazon-retail-conversions/long-term-sales.md | official (API doc) | 2026-11-09 | longTermSales, longTermRoas, cohort fields; not SP |
+| Purchases (new to brand) | https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/ads-v1/metrics/amazon-retail-conversions/purchases-new-to-brand.md | official (API doc) | 2026-11-09 | NTB 12-month definition; cost per NTB purchase formula; NTB ROAS not SP |
+| Purchases (Amazon retail conversions) | https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/ads-v1/metrics/amazon-retail-conversions/purchases.md | official (API doc) | 2026-11-09 | metric.roas and metric.sales |
+| Subscribe & Save subscriptions | https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/ads-v1/metrics/amazon-retail-conversions/subscribe-save-subscriptions.md | official (API doc) | 2026-11-09 | S&S ad metrics; not SP |
+| Costs (Costs and fees) | https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/ads-v1/metrics/costs-and-fees/costs.md | official (API doc) | 2026-11-09 | metric.totalCost definition |
+| Reporting columns (version 3) | https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/v3/columns.md | official (API doc) | 2026-11-09 | acosClicks7d, sales7d, same/other SKU, NTB and long-term columns by report type |
+| Sponsored Products rule-based bidding overview | https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/rules/bidding-rules/overview.md | official (API doc) | 2026-11-09 | ROAS guardrail, eligibility, auto-disable, minimum budgets |
+| Amazon Retail Purchases in AMC | https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/amazon-marketing-cloud/datasources/amazon_retail_purchase_paid.md | official (API doc) | 2026-11-09 | 60-month purchase history; free through 2026-12-31 |
+| AMC paid features overview | https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/amazon-marketing-cloud/datasources/paid_features_overview.md | official (API doc) | 2026-11-09 | Paid datasets and the 2026 no-cost window |
+| How much does it cost to sell on Amazon? ("Selling on Amazon fees" in this file) | https://sell.amazon.com/pricing | official | 2026-11-09 | Referral fee rule and tiers, closing fee, plans, fee-category note, new-brand waivers |
+| Amazon FBA (Fulfillment by Amazon) | https://sell.amazon.com/fulfillment-by-amazon | official | 2026-11-09 | FBA cost categories; aged inventory over 181 days; restock limits; Fee Preview path |
+| Subscribe & Save, Sell on Amazon | https://sell.amazon.com/programs/subscribe-and-save | official | 2026-11-09 | Eligibility, discount tiers, Amazon funding, no extra selling cost, 1.8x figure |
+| Amazon's response to the FTC's lawsuit regarding Sponsored Ads | https://www.aboutamazon.com/news/company-news/amazon-ftc-sponsored-ads-lawsuit-response | official | 2026-11-09 | Hard and soft reserves; who pays what; auction figures (read via the /md/ copy at https://www.aboutamazon.com/md/company-news/amazon-ftc-sponsored-ads-lawsuit-response; the HTML URL returned 404 at verification) |
+| FTC, States Sue Amazon over Secret Ad Surcharge Scheme | https://www.ftc.gov/news-events/news/press-releases/2026/08/ftc-states-sue-amazon-over-secret-ad-surcharge-scheme | government (FTC; not Amazon) | 2026-11-09 | Allegations, 80% figure, court, ad products covered (read with WebFetch) |
+| Product Fees API | https://developer-docs.amazon/sp-api/docs/product-fees-api.md | official (SP-API doc) | 2026-11-09 | Purpose; estimates not guaranteed; sellers only |
+| getMyFeesEstimateForSKU (reference) | https://developer-docs.amazon/sp-api/reference/getmyfeesestimateforsku.md | official (SP-API doc) | 2026-11-09 | Operation reference |
+| Get product fee estimates for a specific SKU | https://developer-docs.amazon/sp-api/docs/get-product-fee-estimates-sku.md | official (SP-API doc) | 2026-11-09 | Tutorial; request shape |
+| Get product fee estimates for a specific ASIN | https://developer-docs.amazon/sp-api/docs/get-product-fee-estimates-asin.md | official (SP-API doc) | 2026-11-09 | Tutorial; ASIN variant |
+| Get product fees for a batched set of up to 20 items | https://developer-docs.amazon/sp-api/docs/get-product-fees-batch.md | official (SP-API doc) | 2026-11-09 | Batch limit of 20 |
+| productFeesV0.json | https://github.com/amzn/selling-partner-api-models/blob/main/models/product-fees-api-model/productFeesV0.json | official (SP-API model) | 2026-11-09 | Paths, request and reply schemas, catalog-size note |
+| SP-API documentation index (llms.txt) | https://developer-docs.amazon/sp-api/llms.txt | official (SP-API doc) | 2026-11-09 | Page index; FBA Inventory API and Finances pages located |
+| Fulfillment by Amazon (FBA) Reports | https://developer-docs.amazon/sp-api/docs/report-type-values-fba.md | official (SP-API doc) | 2026-11-09 | Fee Preview, planning, restock, storage and overage report columns |
+| Analytics Reports | https://developer-docs.amazon/sp-api/docs/report-type-values-analytics.md | official (SP-API doc) | 2026-11-09 | Sales and Traffic, Repeat Purchase, Vendor Net PPM report types |
+| sellerSalesAndTrafficReport schema | https://github.com/amzn/selling-partner-api-models/blob/main/schemas/reports/sellerSalesAndTrafficReport.json | official (SP-API model) | 2026-11-09 | orderedProductSales and the traffic fields |
+| Replenishment API | https://developer-docs.amazon/sp-api/docs/replenishment-api.md | official (SP-API doc) | 2026-11-09 | Subscribe & Save scope of the API |
+| replenishment-2022-11-07.json | https://github.com/amzn/selling-partner-api-models/blob/main/models/replenishment-api-model/replenishment-2022-11-07.json | official (SP-API model) | 2026-11-09 | Operations, Metric enum, EligibilityStatus, 2-year limit |
+| finances_2024-06-19.json | https://github.com/amzn/selling-partner-api-models/blob/main/models/finances-api-model/finances_2024-06-19.json | official (SP-API model) | 2026-11-09 | listTransactions path and 48-hour lag |
+| Amazon announces new streamlined FBA capacity management system | https://sell.amazon.com/blog/announcements/fba-capacity | official | 2026-11-09 (verifier) | Monthly capacity limit; announced in the third full week; reservation fee offset by performance credits (dated March 1, 2023) |
+| Refresher: FBA Capacity Limits & How They Work | https://sellercentral.amazon.com/seller-forums/discussions/t/7dd1b421-5863-46c8-a88a-ba2cdcabad94 | official (Amazon staff, forum) | 2026-11-09 | Capacity limit rules; 15 cubic feet; 39 weeks; Capacity Manager (post dated 2024-03-14) |
+| IPI Score Dropping? What's Dragging It Down and How to Recover | https://sellercentral.amazon.com/seller-forums/discussions/t/a57e02a2-b759-4d73-aa84-ba5f795156bf | official (Amazon staff, forum) | 2026-11-09 | IPI range and factors; link to capacity; new-ASIN exemption (post dated 2026-08-13) |
+| What Is Amazon TACoS? Total Advertising Cost of Sale in 2023 | https://www.junglescout.com/blog/amazon-tacos/ | opinion (Jungle Scout, Brian Connolly) | 2026-11-09 | TACoS definition and example; rank claim (2023-05-24) |
+| The Ultimate Guide to Amazon ACoS and TACoS | https://canopymanagement.com/ultimate-guide-to-acos-and-tacos/ | opinion (Canopy Management, Chuck Kessler) | 2026-11-09 | Break-even example; TACoS bands; launch ACoS; separate by margin |
+| How to Find the Ideal True ACoS on Amazon PPC and Set Campaign Targets | https://myamazonguy.com/advertising/how-to-find-the-ideal-true-acos-on-amazon-ppc/ | opinion (My Amazon Guy, Ken Zhou) | 2026-11-09 | Margin before ads; target from net margin and ad share; stage table (2026-10-13) |
+| Why Your Amazon ACoS Looks Fine—But Profit Is Still Declining (And How to Diagnose It) | https://www.optmyzr.com/blog/why-amazon-acos-looks-good-but-profit-is-declining/ | opinion (Optmyzr, Ronia) | 2026-11-09 | Why stable ACoS can hide falling profit (2026-03-16) |
+| Calculating a Bid Strategy for Amazon Ad Campaigns | https://www.adbadger.com/blog/calculating-amazon-bid-strategy/ | opinion (Ad Badger) | 2026-11-09 | Bid = AOV × CVR × target ACoS |
+| Amazon Says Its Ad Auction Was Second-Price. The FTC Says You Paid Your Own Bid Anyway. | https://www.adbadger.com/blog/amazon-says-its-ad-auction-was-second-price-the-ftc-says-you-paid-your-own-bid-anyway/ | opinion (Ad Badger) | 2026-11-09 | CPC ÷ bid reading; bid-cut test (2026-09-10) |
+| What is Revenue Per Click (RPC) Bidding? | https://help.adlabs.app/en/articles/1-what-is-revenue-per-click-rpc-bidding | opinion (AdLabs, Stephen Noch) | 2026-11-09 | Target CPC = RPC × target ACOS; ranking keywords |
+| Sponsored Products Goal Setup: Targeting Types, Target ACoS, and Budget | https://help.perpetua.io/en/articles/4859772-sponsored-products-goal-setup-targeting-types-target-acos-and-budget | opinion (Perpetua, Ashley Chung) | 2026-11-09 | Targets from 30-day history; segment targets |
+| Sync Ad Spend with Stock Levels Across Marketplaces | https://www.helium10.com/blog/how-to-sync-ad-spend-with-stock-levels-across-marketplaces/ | opinion (Helium 10, Lauren Stair) | 2026-11-09 | Inventory states driving spend (2026-04-01) |
+| Inventory Optimization Overview | https://help.teikametrics.com/en/articles/8613831-inventory-optimization-overview | opinion (Teikametrics) | 2026-11-09 | Inventory insight cards beside campaigns |
+| Average Amazon TACOS Benchmarks by Category (2026) | https://eightx.co/blog/average-tacos-benchmarks | opinion (EightX) | 2026-11-09 | Break-even TACoS; launch and category bands; its sources |
+| Amazon Tests ROAS Auto-Bidding for Sponsored Products | https://novadata.io/resources/news/amazon-sponsored-products-roas-auto-bidding-2026 | opinion (Nova Analytics) | 2026-11-09 | Reported SP beta with no bid and no ROAS target (2026-02-18) |
+| Dorfman–Steiner theorem | https://en.wikipedia.org/wiki/Dorfman%E2%80%93Steiner_theorem | research (encyclopedia statement of a paper) | 2026-11-09 | Statement and citation of the 1954 condition |
+| How Well Does Advertising Work? (IDEAS record of the French-language version, RAM 2011) | https://ideas.repec.org/p/hal/journl/hal-02055301.html | research | 2026-11-09 | Meta-analysis sample and elasticities |
+| TV Advertising Effectiveness and Profitability (EconPapers) | https://econpapers.repec.org/RePEc:wly:emetrp:v:89:y:2021:i:4:p:1855-1879 | research | 2026-11-09 | Marginal vs overall ROI across 288 brands |
+| Consumer Heterogeneity and Paid Search Effectiveness (NBER w20171) | https://www.nber.org/papers/w20171 | research | 2026-11-09 | Brand ads no short-term benefit; negative average returns |
+| The Unfavorable Economics of Measuring the Returns to Advertising (EconPapers) | https://econpapers.repec.org/RePEc:oup:qjecon:v:130:y:2015:i:4:p:1941-1973 | research | 2026-11-09 | ROI confidence intervals; selection bias |
+| Competition and Crowd-Out for Brand Keywords in Sponsored Search (EconPapers) | https://econpapers.repec.org/RePEc:inm:ormksc:v:37:y:2018:i:2:p:200-215 | research | 2026-11-09 | Brand-ad effect and competitor stealing |
+| Measuring the Lifetime Value of Customers Acquired from Google Search Advertising (EconPapers) | https://econpapers.repec.org/RePEc:inm:ormksc:v:30:y:2011:i:5:p:837-850 | research | 2026-11-09 | Search-acquired customers' higher value |
+| Advertising Spillovers in Mobile Apps | https://arxiv.org/abs/2504.16151 | research | 2026-11-09 | Paid-to-organic spillover through rankings |
+| Sponsored is the New Organic | https://arxiv.org/abs/2407.19099 | research | 2026-11-09 | Poor organic rank items shown as sponsored on page 1 of Amazon |
+| Do not Waste Money on Advertising Spend | https://arxiv.org/abs/2212.13923 | research | 2026-11-09 | Concave click curve; diminishing returns to bid |
+| Incrementality Bidding and Attribution | https://arxiv.org/abs/2208.12809 | research | 2026-11-09 | Bidding on causal effects |
+| Attributed, But Not Incremental | https://arxiv.org/abs/2606.26690 | research | 2026-11-09 | Attributed conversions overstate incremental |
+| Sponsored brands video rings up clicks and sales in the short and long run | https://www.amazon.science/publications/sponsored-brands-video-rings-up-clicks-and-sales-in-the-short-and-long-run | research (Amazon Science) | 2026-11-09 | SBv +25% CTR, +10% YoY sales growth |
+| Holiday advertising myths: What experts want you to know (gap fill) | https://advertising.amazon.com/library/guides/common-holiday-advertising-myths | official | 2026-11-09 | TACOS against an ACOS spike; total sales "(organic and paid)"; Adam Mellott (BTR Media) quote; datePublished 2025-09-26, dateModified 2026-10-09 |
+| Update to U.S. Referral and Fulfillment by Amazon fees for 2026 (gap fill) | https://sellingpartners.aboutamazon.com/update-to-u-s-referral-and-fulfillment-by-amazon-fees-for-2026 | official | 2026-11-09 | +$0.08 average FBA per unit; no new FBA fee types; January 15, 2026; Revenue Calculator, Fee and Economics Preview, Profit Analytics |
+| 2026 Updates to US Referral and Fulfillment by Amazon Fees (`News_Amazon`, gap fill) | https://sellercentral.amazon.com/seller-forums/discussions/t/f3fa3211-820b-4e2e-a023-158a9cf55f99 | official (Amazon staff in forum) | 2026-11-09 | The 2026 letter as a staff post |
+| Update to US Referral and Fulfillment by Amazon fees for 2027 (gap fill) | https://sellingpartners.aboutamazon.com/update-to-us-referral-and-fulfillment-by-amazon-fees-for-2027 | official | 2026-11-09 | +$0.19 average per unit; January 15, 2027; storage utilization surcharge ends and aged surcharge merges February 1, 2027; new LILF test; surcharge unchanged; Fee Change Comparison dashboard |
+| Fuel and logistics-related surcharge: FBA, MCF, and BWP in US and CA (`News_Amazon`, gap fill) | https://sellercentral.amazon.com/seller-forums/discussions/t/7cbc0233-ee5b-4359-978a-dee7cad5c6f4 | official (Amazon staff in forum) | 2026-11-09 | 3.5% on fulfillment fees from April 17 / May 2, 2026; about $0.17 per US FBA unit |
+| Holiday 2026: Same fees, same eligibility, earlier deadlines (`News_Amazon`, gap fill) | https://sellercentral.amazon.com/seller-forums/discussions/t/3e31fbb7-04e0-4ed4-873e-f74b1052e2ff | official (Amazon staff in forum) | 2026-11-09 | Peak fulfillment fees October 15, 2026 to January 14, 2027, about $0.32 per unit; AWD off-peak storage to October 31; capacity; deal fees |
+| Multichannel Fulfillment (MCF) - Amazon Supply Chain Services (gap fill) | https://supplychain.amazon.com/mcf | official | 2026-11-09 | MCF US fee update banner, January 15, 2027 |
+| Costs for optional selling services (gap fill) | https://sell.amazon.com/pricing/optional-costs | official | 2026-11-09 | MCF serves other sales channels |
+| Seller Fulfilled Prime (gap fill) | https://sell.amazon.com/programs/seller-fulfilled-prime | official | 2026-11-09 | SFP: fast free delivery and free returns at the seller's cost; prequalification and trial |
+| Bidding strategies for Sponsored Products (help GCU2BUWJH2W3A8Z7, gap fill) | https://advertising.amazon.com/help/GCU2BUWJH2W3A8Z7 | official | 2026-11-09 | Up-and-down bids move "in proportion to the likelihood of a conversion" |
+| Ad campaign attribution (help GX7KDKHMWQYMJ385, gap fill) | https://advertising.amazon.com/help/GX7KDKHMWQYMJ385 | official | 2026-11-09 | Attributed sales exclude cart discounts; deduct payment failures and cancellations within 72 hours; returns not listed |
+| analytics_economics_2024_03_15.graphql (gap fill) | https://github.com/amzn/selling-partner-api-models/blob/main/schemas/data-kiosk/analytics_economics_2024_03_15.graphql | official | 2026-11-09 | Net product sales after refunds; net proceeds; fee types incl. SUS, aged, LILF, returns fees; fee-change split; 120-day preview; MFN cost fields |
+| analytics_sellerAnalytics_2025_03_31.graphql (gap fill) | https://github.com/amzn/selling-partner-api-models/blob/main/schemas/data-kiosk/analytics_sellerAnalytics_2025_03_31.graphql | official | 2026-11-09 | `returns` group: returned units, refunds, returned rate; 45-day delay |
+| Seller Analytics Dataset Use Case Guide (gap fill) | https://developer-docs.amazon/sp-api/docs/seller-analytics-dataset-guide.md | official | 2026-11-09 | Returns mapped to ship date; recent rate reads low; defects over six months |
+| Data Kiosk API (gap fill) | https://developer-docs.amazon/sp-api/docs/data-kiosk-api.md | official | 2026-11-09 | The query and document flow |
+| Returns Reports (gap fill) | https://developer-docs.amazon/sp-api/docs/report-type-values-returns.md | official | 2026-11-09 | Seller-fulfilled returns reports; 60 days per request; no FBA returns, no returnless refunds |
+| Fulfillment by Amazon (FBA) Reports (re-read in the gap fill) | https://developer-docs.amazon/sp-api/docs/report-type-values-fba.md | official | 2026-11-09 | `GET_FBA_FULFILLMENT_CUSTOMER_RETURNS_DATA` fields; request only |
+| sellerSalesAndTrafficReport.json (gap fill) | https://github.com/amzn/selling-partner-api-models/blob/main/schemas/reports/sellerSalesAndTrafficReport.json | official | 2026-11-09 | `unitsRefunded` and `refundRate` by date only; no refund field by ASIN |
+| Customer Feedback API (gap fill) | https://developer-docs.amazon/sp-api/docs/customer-feedback-api.md | official | 2026-11-09 | Return insights at browse-node level only |
+| productPricingV0.json (gap fill) | https://github.com/amzn/selling-partner-api-models/blob/main/models/product-pricing-api-model/productPricingV0.json | official | 2026-11-09 | `LandedPrice` = price + shipping − points; Buy Box threshold net of points |
+| productFeesV0.json (re-read in the gap fill) | https://github.com/amzn/selling-partner-api-models/blob/main/models/product-fees-api-model/productFeesV0.json | official | 2026-11-09 | `PriceToEstimateFees.Points`; points only in Japan |
+| Amazonポイント規約 (Amazon.co.jp Points terms, gap fill) | https://www.amazon.co.jp/gp/help/customer/display.html?nodeId=201998170 | official | 2026-11-09 | Points set by Amazon or the seller; 1 point = 1 yen; base excludes wrapping, shipping, COD; pending points lapse on return |
+| 【重要】ポイント管理ページ「Amazonポイントセントラル」開設等のお知らせ (`News_Amazon`, JP, gap fill) | https://sellercentral.amazon.co.jp/seller-forums/discussions/t/e5d3ea81f020e093d688303ec86c80f5 | official (Amazon staff in forum) | 2026-11-09 | Points Central; default 1% from November 1, 2021; up to 50%; seller pays 1 yen per point |
+| ポイントを付与してダブルポイント祭りで売り上げを伸ばしましょう (`News_Amazon`, JP, gap fill) | https://sellercentral.amazon.co.jp/seller-forums/discussions/t/4c59309f-66c5-49d8-8ce4-8fe25694a560 | official (Amazon staff in forum) | 2026-11-09 | Double Points weekends: Amazon-funded match up to 5% per item, 5,000 per order; eligibility |
+| 値引きとAmazonポイント付与の違いについて (JP seller thread, gap fill) | https://sellercentral.amazon.co.jp/seller-forums/discussions/t/9464c5e4-d862-43db-9c14-c5630f2fb2c0 | opinion (sellers) | 2026-11-09 | Sellers on discount against points; no Amazon reply |
+| The science of price experiments in the Amazon Store (gap fill) | https://www.amazon.science/blog/the-science-of-price-experiments-in-the-amazon-store | research (Amazon Science blog) | 2026-11-09 | Price switchbacks; 60% lower standard error; carryover through recommendations |
+| Regular balanced switchback designs for robust multi-unit online experimentation (gap fill) | https://www.amazon.science/publications/regular-balanced-switchback-designs-for-robust-multi-unit-online-experimentation | research (Amazon Science) | 2026-11-09 | Balanced switchbacks; 40–60% lower standard errors; carryover |
+| Cross-unit spillovers in A/B testing: Empirical evidence from ads (gap fill) | https://www.amazon.science/publications/cross-unit-spillovers-in-a-b-testing-empirical-evidence-from-ads | research (Amazon Science) | 2026-11-09 | Auction spillovers violate SUTVA under advertiser randomisation |
+| Sponsored brands video ... (paper PDF, gap fill) | https://cdn.amazon.science/d1/66/f222c8524c4d87fc1b0b342345b7/sponsored-brands-video-rings-up-clicks-and-sales-in-the-short-and-long-run.pdf | research (Amazon-authored) | 2026-11-09 | Matched sample, causal multi-task GP, propensity score stratification; 50+ confounders; no ROAS analysis |
+| TV Advertising Effectiveness and Profitability (article PDF, public copy, gap fill) | https://www.gwern.net/doc/economics/advertising/2021-shapiro.pdf | research | 2026-11-09 | Identification from the ad-buying process; border strategy; elasticities 0.023 mean, 0.014 median |
+| The Unfavorable Economics of Measuring the Returns to Advertising (OpenAlex, gap fill) | https://api.openalex.org/works/doi:10.1093/qje/qjv023 | research | 2026-11-09 | Full abstract incl. selection bias sentence |
+| A Comparison of Approaches to Advertising Measurement (EconPapers, gap fill) | https://econpapers.repec.org/RePEc:inm:ormksc:v:38:y:2019:i:2:p:193-225 | research | 2026-11-09 | Observational methods fail against 15 Facebook RCTs |
+| Close Enough? A Large-Scale Exploration of Non-Experimental Approaches to Advertising Measurement (gap fill) | https://arxiv.org/abs/2201.07055 | research | 2026-11-09 | DML and SPSM against 663 RCTs; median lifts |
+| Incrementality Bidding and Attribution (re-read in the gap fill) | https://arxiv.org/abs/2208.12809 | research | 2026-11-09 | Randomisation inside bidding and attribution |
+| New Empirical Generalizations on the Determinants of Price Elasticity (Crossref, gap fill) | https://api.crossref.org/works/10.1509/jmkr.42.2.141.62296 | research | 2026-11-09 | Mean price elasticity −2.62 over 1,851 estimates; endogeneity raises magnitude |
+| Measuring Prices and Price Competition Online (NBER w9085, gap fill) | https://www.nber.org/papers/w9085 | research | 2026-11-09 | Amazon book demand far less price-elastic than BN.com (about −0.6 against about −4) |
+| Adventures in Demand Analysis Using AI (gap fill) | https://arxiv.org/abs/2501.00382 | research | 2026-11-09 | Heterogeneous price elasticity of Amazon toy cars |
+| Dorfman–Steiner theorem (raw wikitext, re-read in the gap fill) | https://en.wikipedia.org/w/index.php?title=Dorfman%E2%80%93Steiner_theorem&action=raw | research (encyclopedia statement of a paper) | 2026-11-09 | The condition and its two levers (price or advertising) |
+| Estimating Amazon Ad Incrementality for Sellers (gap fill) | https://whatstheimpact.substack.com/p/estimating-amazon-ad-incrementality | opinion (Jonathan Hershaff, a seller) | 2026-11-09 | Daily on/off switchback; platform ROAS 2.35 against incremental 1.25; MMM for carryover |
+| Amazon sellers face average $0.19 FBA fee rise from January 15, 2027 (gap fill) | https://ppc.land/amazon-sellers-face-average-0-19-fba-fee-rise-from-january-15-2027/ | opinion (PPC Land, Luis Rijo, reporting Amazon pages) | 2026-11-09 | 2027 price bands, returns processing fee dates, 2026 storage figures as reported |
+| Amazon Break-Even ACoS (Nova Analytics, gap fill) | https://novadata.io/resources/blog/amazon-break-even-acos-guide | opinion (Antoine, Nova Analytics) | 2026-11-09 | Returns and refund administration in margin; category return rates |
+| Break-even ACOS for Amazon Ads (FiveX, gap fill) | https://fivex.com/blog/amazon-ads-break-even-acos/ | opinion (FiveX team) | 2026-11-09 | Return allowance; peak-season margin |
+| Average ACOS by Amazon Category: 2026 Benchmarks (EightX, gap fill) | https://eightx.co/blog/average-acos-by-amazon-category | opinion (Ash Kagali, EightX) | 2026-11-09 | Apparel ACOS tied to returns |
+| Amazon Repricing and PPC (Repricer.com, gap fill) | https://www.repricer.com/blog/amazon-ppc-vs-repricing-buy-box-impact/ | opinion (Colin Palin, Repricer.com) | 2026-11-09 | Price, Featured Offer and ad CTR as one margin equation |
+| Price, Cart and Repricers (AMALYZE, gap fill) | https://amalyze.com/blog/price-cart-repricer-ppc-impact | opinion (AMALYZE) | 2026-11-09 | Repricing moves the ACOS target; bid formula with price |
+| How to Use Amazon Coupons to Promote Your Products (Ad Badger, gap fill) | https://www.adbadger.com/blog/amazon-coupons/ | opinion (Ad Badger) | 2026-11-09 | Coupon badge and a 5–10% conversion lift claim |
+| Why the Smartest Brands on Amazon Are Rethinking Promotions in 2026 (Quartile, gap fill) | https://www.quartile.com/blog/why-the-smartest-brands-on-amazon-are-rethinking-promotions-in-2026 | opinion (Sandie Shin, Quartile) | 2026-11-09 | TPCS; TACoS reduction claim |
+
+Also consulted, not as sources of a claim: the `cx-amazon-ads` plugin's `skills/amazon-ads/SKILL.md` and `references/door.md` (the door's shape), and the listing playbook's `research/angles/paid-keywords.md`, `research/angles/inventory-and-fulfillment.md` and `research/method/help-page-access.md` (leads only; every page they cite that this file uses was re-opened here, and pages they read by the help hub's POST content call were not reused).
+
+### Unreachable
+
+- **Seller Central help, FBA capacity limits** (`https://sellercentral.amazon.com/help/hub/reference/external/GAFNWEYTJUV2GBFC`): a plain GET returns a JavaScript shell with no article text. The content call is the forbidden POST `.../mons-api/GetHelpTopic`. The staff forum post stands in for it.
+- **Seller Central help, Inventory Performance Index** (G202174810) and **IPI FAQ** (GZJF4DY2W6MERBAL): same reason; not fetched by the forbidden route.
+- **Seller Central help, low-inventory-level fee** (GV43F6S76Y9DHYRH): same reason. The planning report's fee flags are cited instead. The fee's current and February 2027 tests are stated only as Amazon's 2027 letter words them (section C, gap fill).
+- **Seller Central Revenue Calculator**: needs sign-in or a JavaScript app; not tried beyond the help-page shell.
+- **SSRN 1866002** (Sethuraman, Tellis and Briesch): HTTP 403. The IDEAS record was used.
+- **The Sethuraman, Tellis and Briesch PDF** (a public copy): downloaded, but no PDF text extractor would run on this machine (a missing library). Not read. The verifier re-tried a public copy (gwern.net) on 2026-11-09: same failure (poppler cannot load; the PDF's text uses a custom font encoding). No English-language record of the JMR 48(3), 457–471 article was reached (IDEAS and EconPapers JMR handles returned 404).
+- **aboutamazon.com HTML page of Amazon's FTC response** (`/news/company-news/amazon-ftc-sponsored-ads-lawsuit-response`): HTTP 404 at verification; the `/md/` copy is the text read.
+- **EconBiz records** (Sethuraman et al.; Jing and Lewis 2011, "Stockouts in Online Retailing", *JMR* 48(2)): a bot proof-of-work wall. Jing and Lewis is therefore not cited for any claim.
+- **ScienceDirect article S0167624524000374**: a captcha page. Not read; not cited.
+- **LandingCube, true break-even ACoS article**: a bot verification page (10 words).
+- **Teikametrics product release blog (2022)**: HTTP 404.
+- **Amazon Ads India help, new-to-brand** (G5EWB37XK4RS4J4T on advertising.amazon.in): returned an empty page.
+- **SP-API Product Fees use-case guide** (`developer-docs.amazon.com/sp-api/docs/product-fees-api-v0-use-case-guide`): HTTP 404. The `.md` docs and the model were used.
+- **sell.amazon.com FBA fees page** (`https://sell.amazon.com/fulfillment-by-amazon/fees`): "Page not found". The FBA page was used.
+- **Dorfman and Steiner (1954), the AER article itself**: not tried (JSTOR); the Wikipedia statement was used.
+- **Seller Central help pages named by the 2026 and 2027 fee letters and the Data Kiosk schema (gap fill)**: FBA fee details (G201411300), storage fees and storage utilization surcharge (G3EDYEF6KUCFQTNM), the returns processing fee (G64LS955WNFT6EDP), the low-inventory-level fee (GV43F6S76Y9DHYRH), and the 2027 detail pages. A plain GET of each returns the same 336-word JavaScript shell; the content call is the forbidden POST. Peak storage rates, the 2027 rate cards and the returns-fee category thresholds are therefore not stated from Amazon here.
+- **sell.amazon.com pages on MCF, merchant-fulfilled selling and FBA low-price rates** (`/fulfillment-by-amazon/mcf`, `/fulfilled-by-merchant`, `/programs/fba-low-price`): "Page not found" (gap fill). The supplychain.amazon.com MCF page and the optional-costs page were used.
+- **SP-API economics dataset guide** (a guessed `developer-docs.amazon/sp-api/docs/...economics...md` path): a soft-404 page (gap fill). The GraphQL schema on GitHub was used.
+- **INFORMS page for Gordon et al. (2019)** (pubsonline): HTTP 403 (gap fill). The EconPapers record was used.
+- **Pacvue, Canopy Management and Helium 10 on a price cut against a bid raise**: no page found by search (gap fill); nothing is attributed to them on this point.
+
+## Open questions
+
+1. **Does Amazon's "cost of goods sold" in Ads Math include Amazon's fees?** The page does not say. The answer decides whether Amazon's own break-even ROAS formula matches the practitioners' margin before ads.
+2. **Is there an SP-API read of FBA capacity limits and the IPI score?** None was found in the documents read. If there is none, cx reads them from the seller or from Seller Central pages.
+3. **What is the status of an SP ROAS-optimising bid strategy with no bid and no target?** It is reported by one practitioner (Nova Analytics, 2026-02-18) and not found on an Amazon page or in the v1 `BidStrategy` enum read here.
+4. **Will long-term sales, NTB ROAS and S&S metrics come to Sponsored Products?** The v1 pages list them as not SP-eligible today. v3 SP report types carry no NTB columns at all, while v1 SP does.
+5. **What happens to auction pricing after the FTC case?** Amazon describes the soft-reserve rule, the FTC alleges an 80% own-bid rate, and the case is untried. Whether pricing or help content changes again is open; a skill reads the current help content and the seller's own CPC-to-bid ratios.
+6. **What is the low-inventory-level fee's current rule?** Not readable here by permitted routes. The planning report's flags show whether it applied. Partly answered in the gap fill: Amazon's 2027 letter says the fee "is assessed using trailing 30- and 90-day sales" today and, from February 1, 2027, "will apply only when your inventory falls below 28 days of supply, based on actual shipments in both the four weeks before and after the assessment date". The fee amounts and exemptions stay on the unreachable help page.
+7. **Does Amazon fund 5% on all auto-enrolled S&S products, or only on orders of five items or more?** The S&S seller page states both.
+8. **What elasticity of total sales to SP spend holds on Amazon?** No published study was found that measures it; the meta-analyses are for brand advertising across channels.
+9. **Are returned sponsored-ads sales ever removed from attributed sales?** Amazon's attribution page lists cancellations within 72 hours and payment failures, not returns (section O). Owner of the question: `beyond/metrics-and-attribution.md` (its open question 5).
+10. **Do Sponsored Products ads and their attributed sales on Amazon.co.jp show or subtract points?** No Amazon page read says. The Product Pricing model nets points out of landed price; the attribution page speaks only of cart discounts.
+11. **What are the US peak monthly storage rates and the 2027 age-based storage rate card?** Amazon's staff post presupposes a peak storage rate, and the 2027 letter announces an age-based rate from February 1, 2027. The rate pages are help-hub shells. PPC Land's figures ($0.78 and $2.40 per cubic foot for standard size in 2026) and other vendors' ($0.87 off-peak) disagree.
+12. **Which categories carry the returns processing fee, at what return-rate threshold and rate?** On help G64LS955WNFT6EDP, unreachable. The 2027 dates (January 15 for apparel and shoes, February 1 for other high-return categories) are PPC Land's report.
+13. **Does Amazon keep part of the referral fee on a refund, and how much?** Nova Analytics says "Amazon keeps 20% of referral on returns"; no Amazon page read here states a refund administration fee.
+14. **Is the JP Double Points weekend program still running?** The staff post read is about three years old.
+
+## Positions
+
+**P-goals-and-economics-1.** Amazon defines ACOS as ad spend divided by ad-attributed revenue, times 100, and ROAS as ad revenue divided by ad spend, the inverse of ACOS. Sources: *What is advertising cost of sales (ACOS)?* (https://advertising.amazon.com/library/guides/acos-advertising-cost-of-sales); *Advertising budget* (https://advertising.amazon.com/library/guides/advertising-budget). [official]
+
+**P-goals-and-economics-2.** Amazon states that to maintain a profit, ACOS must be lower than the profit margin, and calls this break-even ACOS. Source: *What is advertising cost of sales (ACOS)?* (https://advertising.amazon.com/library/guides/acos-advertising-cost-of-sales). [official]
+
+**P-goals-and-economics-3.** Amazon's Ads Math gives gross profit margin as (revenue minus cost of goods sold) divided by revenue, break-even ROAS as 1 divided by that margin, and treats a campaign as not losing money when its ROAS is at or above break-even ROAS. Source: *Ads Math* (https://advertising.amazon.com/library/guides/ads-math). [official]
+
+**P-goals-and-economics-4.** Amazon states there is no general threshold for a good ACOS or ROAS, that a good ROAS is one that is profitable for the advertiser, and that ROAS is a preliminary metric to be read with margin, conversions, lifetime value, profit or incremental lift; one Amazon FBA guide nonetheless prints a benchmark of $2–4 per $1 of ad spend. Sources: *What is advertising cost of sales (ACOS)?*; *Ads Math*; *How to Help Increase Sales with FBA and Advertising* (https://advertising.amazon.com/library/guides/fba). [official]
+
+**P-goals-and-economics-5.** Amazon's click-chain formulas are CPA = CPC ÷ conversion rate, profit per sale = average order value × gross profit margin, and net profit per sale = profit per sale − CPA; Amazon says this calculator is for one-off purchases and ignores repeat customers. Source: *Ads Math* (https://advertising.amazon.com/library/guides/ads-math). [official]
+
+**P-goals-and-economics-6.** Amazon defines customer LTV as average order value × average orders and affordable CPA as LTV × profit margin, and calls affordable CPA the upper limit for ad spend per new customer to break even over the customer's lifetime. Source: *Ads Math* (https://advertising.amazon.com/library/guides/ads-math). [official]
+
+**P-goals-and-economics-7.** Amazon states that new campaigns tend to have a high ACOS because of their newness, and that this does not make them failures but calls for other KPIs. Source: *What is advertising cost of sales (ACOS)?* (https://advertising.amazon.com/library/guides/acos-advertising-cost-of-sales). [official]
+
+**P-goals-and-economics-8.** Amazon's referral fee for each item is a category-specific percentage of the total price (list price plus shipping and gift wrap) or a minimum amount, whichever is greater; a product's fee category may differ from the category shoppers see, and media items carry an additional closing fee. Source: *Selling on Amazon fees* (https://sell.amazon.com/pricing). [official]
+
+**P-goals-and-economics-9.** Amazon lists FBA costs as fulfillment (based on the product's price, weight and dimensions), monthly storage by daily average cubic feet, aged inventory for items stored more than 181 days, returns processing, removal, disposal and liquidation per item, and inbound placement, and tells sellers to calculate margins with storage and fulfillment costs included. Source: *Fulfillment by Amazon (FBA)* (https://sell.amazon.com/fulfillment-by-amazon). [official]
+
+**P-goals-and-economics-10.** The SP-API Product Fees API returns, per SKU or ASIN at a caller-supplied listing price and fulfillment channel, a `TotalFeesEstimate` and a `FeeDetailList` of fee types and amounts; Amazon states these estimates are not guaranteed, and ASIN-based estimates use catalog size, which may differ from the item sent to Amazon. Sources: *Product Fees API* (https://developer-docs.amazon/sp-api/docs/product-fees-api.md); *productFeesV0.json* (https://github.com/amzn/selling-partner-api-models/blob/main/models/product-fees-api-model/productFeesV0.json). [official]
+
+**P-goals-and-economics-11.** The FBA Fee Preview report (`GET_FBA_ESTIMATED_FBA_FEES_TXT_DATA`) holds per-SKU estimated referral, closing and fulfillment fees for active FBA offers, is updated at least once every 72 hours and can be requested once per day per seller; actual charges are returned by the Finances API's `listTransactions`, which may omit the last 48 hours. Sources: *Fulfillment by Amazon (FBA) Reports* (https://developer-docs.amazon/sp-api/docs/report-type-values-fba.md); *finances_2024-06-19.json* (https://github.com/amzn/selling-partner-api-models/blob/main/models/finances-api-model/finances_2024-06-19.json). [official]
+
+**P-goals-and-economics-12.** Amazon defines a bid as the maximum an advertiser is willing to pay for a click, and states that under its reserve-price system the winning advertiser pays the soft reserve when the bid exceeds it and pays the bid when the bid exceeds the hard reserve but not the soft reserve, never more than the bid. Sources: *FAQs* (https://advertising.amazon.com/resources/faq); *Amazon's response to the FTC's lawsuit regarding Sponsored Ads* (https://www.aboutamazon.com/news/company-news/amazon-ftc-sponsored-ads-lawsuit-response). [official]
+
+**P-goals-and-economics-13.** The FTC and states' complaint filed 2026-08-31 in the Western District of Washington alleges that by 2024 Sponsored Products advertisers paid their own winning bid about 80% of the time after Amazon introduced a soft reserve price; the allegations are untried. Source: *FTC, States Sue Amazon over Secret Ad Surcharge Scheme* (https://www.ftc.gov/news-events/news/press-releases/2026/08/ftc-states-sue-amazon-over-secret-ad-surcharge-scheme). [government; not Amazon's words]
+
+**P-goals-and-economics-14.** Ad Badger holds that an advertiser can only infer its auction pricing from its own account, by dividing average CPC by the current bid per target, with ratios near 0.90–1.00 meaning effectively first-price. Source: Ad Badger, *Amazon Says Its Ad Auction Was Second-Price...* (https://www.adbadger.com/blog/amazon-says-its-ad-auction-was-second-price-the-ftc-says-you-paid-your-own-bid-anyway/). [opinion, Ad Badger]
+
+**P-goals-and-economics-15.** Amazon's Sponsored Products guidance names objectives to set before launching a campaign, including brand exposure, new product promotion, inventory clearance and peak season promotion, and Amazon's measurement guide pairs awareness, consideration, purchase and loyalty goals with different metrics. Sources: *Best practices for your Sponsored Products ads* (https://advertising.amazon.com/library/guides/sponsored-products-best-practices); *How to measure and improve your campaigns* (https://advertising.amazon.com/library/guides/measure-improve-campaigns). [official]
+
+**P-goals-and-economics-16.** In Ads API v1 the campaign's `optimizations.goalSettings` is not available for Sponsored Products; for Sponsored Brands the `goal` is read-only and the `kpi` is required. Source: *The campaign entity in the Amazon Ads API* (https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/campaign-management/entities/campaign.md). [official]
+
+**P-goals-and-economics-17.** The v1 Ads API model's `Goal` values are AWARENESS, CONSIDERATION, CONVERSIONS and ENGAGEMENT, its `KPI` values include ROAS, PURCHASES, NEW_TO_BRAND_PURCHASES and TOP_OF_SEARCH_IMPRESSION_SHARE, and its `BidStrategy` values include NEW_TO_BRAND, PRIORITIZE_KPI_TARGET, RULE_BASED and SPEND_BUDGET_IN_FULL. Source: *Amazon Ads API v1 merged OpenAPI* (https://d1y2lf8k3vrkfu.cloudfront.net/openapi/en-us/dest/AmazonAdsAPIALLMerged_prod_3p.json). [official]
+
+**P-goals-and-economics-18.** Sponsored Brands offers two campaign goals: "Drive page visits", measured by clicks and paid by CPC, and "Grow brand impression share", measured by top-of-search impression share and paid by vCPM; earlier SB campaigns are tagged "Drive page visits". Source: *Sponsored Brands goal-based campaigns* (https://advertising.amazon.com/resources/whats-new/sponsored-brands-goal-based-campaigns). [official]
+
+**P-goals-and-economics-19.** Sponsored Products rule-based bidding lets an eligible campaign carry a ROAS guardrail within which Amazon adjusts base bids; Amazon does not guarantee the guardrail, bases it on the previous 21 days, disables the rule (restoring the previous bidding strategy and target bids) when the campaign is not meeting the guardrail and its last-20-day ROAS falls under 30% of its ROAS over a 20-to-40-day period, and warns that aggressive guardrails can reduce conversions. Source: *Sponsored Products rule-based bidding overview* (https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/rules/bidding-rules/overview.md). [official]
+
+**P-goals-and-economics-20.** Amazon's 2026 goal-led products (Full-Funnel Campaigns and DVA+ through Amazon Ads Agent) take a budget, products, creative and a business objective, and report new-to-brand and long-term sales metrics. Sources: *Introducing Amazon Ads Agent, your AI-powered ad platform* (https://advertising.amazon.com/library/news/amazon-ads-agent); *unBoxed 2026 news: Explore the latest advertising announcements* (https://advertising.amazon.com/library/news/unboxed-2026-news-announcements). [official]
+
+**P-goals-and-economics-21.** Amazon defines a new-to-brand purchase as one by a shopper who had not purchased from the brand in the previous 12 months, and computes cost per new-to-brand purchase as total cost divided by new-to-brand purchases. Source: *Purchases (new to brand)* (https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/ads-v1/metrics/amazon-retail-conversions/purchases-new-to-brand.md). [official]
+
+**P-goals-and-economics-22.** In v1 reporting, new-to-brand purchases and sales are available for Sponsored Products but new-to-brand ROAS is not; in Reporting v3, new-to-brand columns appear in Sponsored Brands, Display, TV and DSP report types and in no Sponsored Products report type. Sources: *Purchases (new to brand)*; *Reporting columns (version 3)* (https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/v3/columns.md). [official]
+
+**P-goals-and-economics-23.** Amazon's long-term sales metric is attributed sales plus the estimated 365-day value of new-to-brand shopper engagements, modelled from the brand's historical shopping patterns; it and long-term ROAS are available for Sponsored Brands, Display, TV, DSP and Full-Funnel, not Sponsored Products. Sources: *Long-term sales* (https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/ads-v1/metrics/amazon-retail-conversions/long-term-sales.md); *Campaign reporting: Analyze advertising performance* (https://advertising.amazon.com/measurement-analytics/campaign-reporting). [official]
+
+**P-goals-and-economics-24.** Subscribe & Save subscriptions attributed to ads are reported for Sponsored Brands, Display, TV, DSP and Full-Funnel, not Sponsored Products. Source: *Subscribe & Save subscriptions* (https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/ads-v1/metrics/amazon-retail-conversions/subscribe-save-subscriptions.md). [official]
+
+**P-goals-and-economics-25.** Amazon's Brand Analytics Repeat Purchase report gives per-ASIN orders, unique customers, repeat-customer share and repeat-purchase revenue by week, month or quarter, for brand-registered selling partners with the Brand Analytics role. Source: *Analytics Reports* (https://developer-docs.amazon/sp-api/docs/report-type-values-analytics.md). [official]
+
+**P-goals-and-economics-26.** The SP-API Replenishment API reports Subscribe & Save metrics including subscriber lifetime value by customer segment over 24 months, subscriber against non-subscriber revenue and reorders, 30- and 90-day retention, and revenue lost to out-of-stock, for up to the trailing two years. Sources: *Replenishment API* (https://developer-docs.amazon/sp-api/docs/replenishment-api.md); *replenishment-2022-11-07.json* (https://github.com/amzn/selling-partner-api-models/blob/main/models/replenishment-api-model/replenishment-2022-11-07.json). [official]
+
+**P-goals-and-economics-27.** Amazon states that Subscribe & Save carries no additional selling costs, that enrolled products start at a 0% seller-funded discount the seller can raise to 5% or 10%, that eligibility needs Brand Registry, and that keeping products in stock is a key consideration. Source: *Subscribe & Save, Sell on Amazon* (https://sell.amazon.com/programs/subscribe-and-save). [official]
+
+**P-goals-and-economics-28.** The AMC dataset Amazon Retail Purchases covers up to 60 months of a brand's purchase activity in the Amazon store regardless of advertising, and is free to query from 2026-06-01 to 2026-12-31 where available. Sources: *Amazon Retail Purchases in AMC* (https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/amazon-marketing-cloud/datasources/amazon_retail_purchase_paid.md); *AMC paid features overview* (https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/amazon-marketing-cloud/datasources/paid_features_overview.md). [official]
+
+**P-goals-and-economics-29.** Amazon names TACOS (total advertising cost of sales) on two of its library pages. One gives it as ad spend as a percentage of total sales; the other calls total sales "organic and paid". Neither page names the sales field, the ad products counted or the window, and no Ads API metric of that name was found in the v1 metric pages or v3 columns read. Sources: *The author's guide to Sponsored Products* (https://advertising.amazon.com/library/guides/authors-guide-to-sponsored-products); *Holiday advertising myths: What experts want you to know* (https://advertising.amazon.com/library/guides/common-holiday-advertising-myths). [official]
+
+**P-goals-and-economics-30.** Amazon states that an ad does not display when the product is out of stock or is not the Featured Offer. Source: *FAQs* (https://advertising.amazon.com/resources/faq). [official]
+
+**P-goals-and-economics-31.** Amazon staff state that FBA capacity limits for professional accounts active in FBA over 39 weeks are based on IPI score and sales performance, that usage counts inventory on hand plus open inbound shipments in cubic feet, and that Capacity Manager accepts limit-increase requests at a seller-specified reservation fee. Source: Amazon staff post, *Refresher: FBA Capacity Limits & How They Work* (https://sellercentral.amazon.com/seller-forums/discussions/t/7dd1b421-5863-46c8-a88a-ba2cdcabad94). [official]
+
+**P-goals-and-economics-32.** Amazon staff state that the IPI score ranges from 0 to 1,000, directly influences FBA capacity limits, and reflects excess inventory (over 90 days of supply on forecast demand), sell-through, stranded inventory and in-stock rate of replenishable best sellers, with new ASINs not affecting IPI for their first 90 days. Source: Amazon staff post, *IPI Score Dropping?* (https://sellercentral.amazon.com/seller-forums/discussions/t/a57e02a2-b759-4d73-aa84-ba5f795156bf). [official]
+
+**P-goals-and-economics-33.** Amazon's FBA inventory planning report gives per-SKU days of supply, weeks of cover, estimated excess quantity, inventory age buckets, aged-inventory surcharge exposure, low-inventory-level fee flags, seasonality flags and recommended actions with a recommended sales price; the restock report gives total days of supply including open shipments and a recommended replenishment quantity and ship date. Source: *Fulfillment by Amazon (FBA) Reports* (https://developer-docs.amazon/sp-api/docs/report-type-values-fba.md). [official]
+
+**P-goals-and-economics-34.** When a Subscribe & Save offer is in `REPLENISHMENT_ONLY_ORDERING`, Amazon blocks one-time purchases to preserve inventory for existing subscriptions. Source: *replenishment-2022-11-07.json* (https://github.com/amzn/selling-partner-api-models/blob/main/models/replenishment-api-model/replenishment-2022-11-07.json). [official]
+
+**P-goals-and-economics-35.** My Amazon Guy defines break-even ACoS as the margin before ad costs (price minus landed cost, referral fee, FBA fee and an allowance for returns and storage) and derives a campaign ACoS target by subtracting a target net margin and dividing by the share of total sales that comes from ads. Source: My Amazon Guy (Ken Zhou), *How to Find the Ideal True ACoS on Amazon PPC and Set Campaign Targets* (https://myamazonguy.com/advertising/how-to-find-the-ideal-true-acos-on-amazon-ppc/). [opinion, My Amazon Guy]
+
+**P-goals-and-economics-36.** EightX holds that break-even TACoS equals the pre-ad gross margin after cost of goods, referral fee and FBA fee, and that a sustainable TACoS target is that margin minus the net-margin goal. Source: EightX, *Average Amazon TACOS Benchmarks by Category (2026)* (https://eightx.co/blog/average-tacos-benchmarks). [opinion, EightX]
+
+**P-goals-and-economics-37.** Perpetua sets initial target ACoS from the last 30 days' historical ACoS and budget, with lower targets for branded segments and higher ones for competitor segments. Source: Perpetua, *Sponsored Products Goal Setup* (https://help.perpetua.io/en/articles/4859772-sponsored-products-goal-setup-targeting-types-target-acos-and-budget). [opinion, Perpetua]
+
+**P-goals-and-economics-38.** Canopy Management, EightX and My Amazon Guy each publish launch-stage ACoS or TACoS bands above their mature-stage bands, and the bands do not agree with one another; each firm calls its bands starting points or guidelines. Sources: Canopy Management (https://canopymanagement.com/ultimate-guide-to-acos-and-tacos/); EightX (https://eightx.co/blog/average-tacos-benchmarks); My Amazon Guy (https://myamazonguy.com/advertising/how-to-find-the-ideal-true-acos-on-amazon-ppc/). [opinion, Canopy Management; EightX; My Amazon Guy]
+
+**P-goals-and-economics-39.** Ad Badger and AdLabs set a bid or target CPC as revenue per click (order value × conversion rate) times target ACoS. Sources: Ad Badger, *Calculating a Bid Strategy for Amazon Ad Campaigns* (https://www.adbadger.com/blog/calculating-amazon-bid-strategy/); AdLabs, *What is Revenue Per Click (RPC) Bidding?* (https://help.adlabs.app/en/articles/1-what-is-revenue-per-click-rpc-bidding). [opinion, Ad Badger; AdLabs]
+
+**P-goals-and-economics-40.** Helium 10 holds that ad spend should follow inventory state, redirecting demand before a stock-out and raising demand for overstock, and that stock-outs and forced discounting are costs. Source: Helium 10, *Sync Ad Spend with Stock Levels Across Marketplaces* (https://www.helium10.com/blog/how-to-sync-ad-spend-with-stock-levels-across-marketplaces/). [opinion, Helium 10]
+
+**P-goals-and-economics-41.** The Dorfman–Steiner condition states that a firm's profit-maximising ratio of advertising to sales equals its price-cost margin times the advertising elasticity of demand. Source: Dorfman and Steiner (1954), *American Economic Review* 44, as stated in *Dorfman–Steiner theorem* (https://en.wikipedia.org/wiki/Dorfman%E2%80%93Steiner_theorem). [research]
+
+**P-goals-and-economics-42.** A meta-analysis of 751 short-term and 402 long-term brand advertising elasticities from 56 studies (1960–2008) found mean elasticities of 0.12 short term and 0.24 long term, declining over time and higher early in the product life cycle than in maturity. Source: Sethuraman, Tellis and Briesch (2011), *Journal of Marketing Research* 48(3), read through the IDEAS record of its French-language version in *Recherche et Applications en Marketing* (https://ideas.repec.org/p/hal/journl/hal-02055301.html). [research]
+
+**P-goals-and-economics-43.** A study of TV advertising for 288 brands found negative ROI at the margin for more than 80% of brands and positive overall ROI for only one third. Source: Shapiro, Hitsch and Tuchman (2021), *Econometrica* 89(4) (https://econpapers.repec.org/RePEc:wly:emetrp:v:89:y:2021:i:4:p:1855-1879). [research]
+
+**P-goals-and-economics-44.** Large paid-search field experiments at eBay found returns a fraction of non-experimental estimates, no measurable short-term benefit from brand-keyword ads, and positive effects on new and infrequent users only. Source: Blake, Nosko and Tadelis (2015), *Econometrica* 83, NBER w20171 (https://www.nber.org/papers/w20171). [research]
+
+**P-goals-and-economics-45.** Across 25 online advertising field experiments, the median confidence interval on ROI was over 100 percentage points wide. Source: Lewis and Rao (2015), *Quarterly Journal of Economics* 130(4) (https://econpapers.repec.org/RePEc:oup:qjecon:v:130:y:2015:i:4:p:1941-1973). [research]
+
+**P-goals-and-economics-46.** Customers acquired through Google search advertising at one firm had a higher transaction rate than those from other channels, and their value after future purchases was much higher than conventional first-purchase measures. Source: Chan, Wu and Xie (2011), *Marketing Science* 30(5) (https://econpapers.repec.org/RePEc:inm:ormksc:v:30:y:2011:i:5:p:837-850). [research]
+
+**P-goals-and-economics-47.** In a mobile-app ad shutoff, organic installs fell 20–30%, with the spillover running through app store category rankings. Source: Ju, Zhao and Aral (2025), arXiv 2504.16151 (https://arxiv.org/abs/2504.16151). [research]
+
+**P-goals-and-economics-48.** Click-prediction curves against bid are concave beyond a turning point, so each further increase in bid buys fewer additional clicks. Source: Kong, Shmakov and Yang, arXiv 2212.13923 (https://arxiv.org/abs/2212.13923). [research]
+
+**P-goals-and-economics-49.** Amazon's 2026 US fee letter raised FBA fees by an average of $0.08 per unit sold from January 15, 2026, with no new FBA fee types; its 2027 letter (November 5, 2026) raises FBA fulfillment fees by an average of $0.19 per unit, about 0.8% of the average selling price, with most changes effective January 15, 2027. Sources: *Update to U.S. Referral and Fulfillment by Amazon fees for 2026* (https://sellingpartners.aboutamazon.com/update-to-u-s-referral-and-fulfillment-by-amazon-fees-for-2026); *Update to US Referral and Fulfillment by Amazon fees for 2027* (https://sellingpartners.aboutamazon.com/update-to-us-referral-and-fulfillment-by-amazon-fees-for-2027). [official]
+
+**P-goals-and-economics-50.** Amazon's 2027 letter eliminates the storage utilization surcharge from February 1, 2027, and combines the aged inventory surcharge with monthly storage fees into a single age-based rate. Amazon says this gives flat or lower fees for inventory that turns in under six months and higher fees for aged inventory. From the same date the low-inventory-level fee applies only below 28 days of supply, measured on shipments in the four weeks before and after the assessment date. Source: *Update to US Referral and Fulfillment by Amazon fees for 2027* (as above). [official]
+
+**P-goals-and-economics-51.** From April 17, 2026 Amazon applies a 3.5% fuel and logistics-related surcharge to FBA fulfillment fees in the US and Canada, calculated on the fulfillment fee, not the sale price, averaging $0.17 per US FBA unit; it applies to Buy with Prime and MCF from May 2, 2026, and on top of holiday peak fulfillment fees. Sources: `News_Amazon`, *Fuel and logistics-related surcharge: FBA, MCF, and BWP in US and CA* (https://sellercentral.amazon.com/seller-forums/discussions/t/7cbc0233-ee5b-4359-978a-dee7cad5c6f4); *Holiday 2026: Same fees, same eligibility, earlier deadlines* (https://sellercentral.amazon.com/seller-forums/discussions/t/3e31fbb7-04e0-4ed4-873e-f74b1052e2ff). [official]
+
+**P-goals-and-economics-52.** Holiday peak fulfillment fees apply from October 15, 2026 to January 14, 2027 for FBA, Remote Fulfillment, MCF and Buy with Prime, averaging $0.32 per unit over non-peak rates. Source: `News_Amazon`, *Holiday 2026: Same fees, same eligibility, earlier deadlines* (as above). [official]
+
+**P-goals-and-economics-53.** Amazon itemises, among the fee types in its Data Kiosk economics dataset, a storage utilization surcharge, an aged inventory surcharge, a low-inventory-level fee, an inbound placement service fee, a returns processing fee for apparel and shoes, a high return rate fee for other categories, and Sponsored Products charges. The dataset shows a fee separately on each side of an Amazon fee-change date and previews fees up to 120 days ahead. Source: *analytics_economics_2024_03_15.graphql* (https://github.com/amzn/selling-partner-api-models/blob/main/schemas/data-kiosk/analytics_economics_2024_03_15.graphql). [official]
+
+**P-goals-and-economics-54.** Multi-Channel Fulfillment fulfils orders placed on the seller's own website or other sales channels, not Amazon-store orders. Seller Fulfilled Prime commits the seller to fast free delivery and free returns for domestic Prime customers. Sources: *Costs for optional selling services | Sell on Amazon* (https://sell.amazon.com/pricing/optional-costs); *Seller Fulfilled Prime | Sell on Amazon* (https://sell.amazon.com/programs/seller-fulfilled-prime). [official]
+
+**P-goals-and-economics-55.** Sponsored-ads attributed sales do not reflect discounts applied in the cart. They deduct payment failures and purchases cancelled within 72 hours of order creation, with deductions taking up to 7 days. Returns are not among the listed deductions. Source: *Ad campaign attribution* (https://advertising.amazon.com/help/GX7KDKHMWQYMJ385). [official]
+
+**P-goals-and-economics-56.** Amazon's Data Kiosk economics dataset defines net product sales as ordered product sales minus refunded product sales, counted by refund request time. It defines net proceeds as net product sales minus Amazon fees, ad charges and seller-entered unit costs times net units sold, by child ASIN, parent ASIN or SKU. Source: *analytics_economics_2024_03_15.graphql* (as above). [official]
+
+**P-goals-and-economics-57.** Per-ASIN returned units, refund amount and returned rate (returned over shipped units) are in the Data Kiosk Seller Analytics `returns` group. They are mapped to the day each unit shipped and computed about 45 days late, so recent rates read low. Sources: *analytics_sellerAnalytics_2025_03_31.graphql* (https://github.com/amzn/selling-partner-api-models/blob/main/schemas/data-kiosk/analytics_sellerAnalytics_2025_03_31.graphql); *Seller Analytics Dataset Use Case Guide* (https://developer-docs.amazon/sp-api/docs/seller-analytics-dataset-guide.md). [official]
+
+**P-goals-and-economics-58.** FBA customer returns, with ASIN, SKU, quantity, reason and disposition, are in `GET_FBA_FULFILLMENT_CUSTOMER_RETURNS_DATA`. Seller-fulfilled returns are in the returns-by-return-date reports, which exclude FBA returns and returnless refunds. In the Sales and Traffic report, refunded units and refund rate exist only in the by-date section, not by ASIN. Sources: *Fulfillment by Amazon (FBA) Reports* (https://developer-docs.amazon/sp-api/docs/report-type-values-fba.md); *Returns Reports* (https://developer-docs.amazon/sp-api/docs/report-type-values-returns.md); *sellerSalesAndTrafficReport.json* (https://github.com/amzn/selling-partner-api-models/blob/main/schemas/reports/sellerSalesAndTrafficReport.json). [official]
+
+**P-goals-and-economics-59.** Nova Analytics (Antoine), FiveX and My Amazon Guy put a returns allowance into margin before ads when computing break-even ACoS. Nova states that "Amazon keeps 20% of referral on returns". None of the practitioner pages read computes ACoS or TACoS on revenue net of returns. Sources: Nova Analytics (https://novadata.io/resources/blog/amazon-break-even-acos-guide); FiveX (https://fivex.com/blog/amazon-ads-break-even-acos/); My Amazon Guy (§What practitioners add). [opinion]
+
+**P-goals-and-economics-60.** On Amazon.co.jp, points are granted at a rate set by Amazon or by the seller. One point is used as one yen, and the seller bears one yen per point granted. Rates are set in Points Central, up to 50% per listing, with a default 1% auto-applied from November 1, 2021 to listings the seller had not set. Sources: *Amazonポイント規約* (https://www.amazon.co.jp/gp/help/customer/display.html?nodeId=201998170); `News_Amazon`, Points Central announcement (https://sellercentral.amazon.co.jp/seller-forums/discussions/t/e5d3ea81f020e093d688303ec86c80f5). [official]
+
+**P-goals-and-economics-61.** In Amazon's Product Pricing model, landed price is listing price plus shipping minus points. An offer may be ineligible for the Buy Box when price plus shipping minus points exceeds the competitive price threshold. The Product Fees estimate accepts points as an input, and the points program exists only in Japan. Sources: *productPricingV0.json* (https://github.com/amzn/selling-partner-api-models/blob/main/models/product-pricing-api-model/productPricingV0.json); *productFeesV0.json* (https://github.com/amzn/selling-partner-api-models/blob/main/models/product-fees-api-model/productFeesV0.json). [official]
+
+**P-goals-and-economics-62.** In the Dorfman–Steiner framework a firm can raise sales by cutting price or by advertising. At the optimum the ratio of advertising to sales equals the price-cost margin times the advertising elasticity. Source: *Dorfman–Steiner theorem* (https://en.wikipedia.org/wiki/Dorfman%E2%80%93Steiner_theorem). [research]
+
+**P-goals-and-economics-63.** A meta-analysis of 1,851 price elasticities from 81 studies found an average of −2.62, with sales elasticities growing in magnitude over four decades and larger magnitudes where price endogeneity is accommodated. Source: Bijmolt, Van Heerde and Pieters (2005), *Journal of Marketing Research* 42(2) (https://api.crossref.org/works/10.1509/jmkr.42.2.141.62296). [research]
+
+**P-goals-and-economics-64.** From sales ranks of about 20,000 books in 2001, Chevalier and Goolsbee estimated Amazon's own-price elasticity around −0.6, against close to −4 at Barnes and Noble's site. Bach et al. find strong product-level heterogeneity in price elasticity among toy cars on Amazon.com. Sources: NBER w9085 (https://www.nber.org/papers/w9085); arXiv 2501.00382 (https://arxiv.org/abs/2501.00382). [research]
+
+**P-goals-and-economics-65.** Shapiro, Hitsch and Tuchman state that advertising is not randomly assigned, so estimated effects have no causal interpretation without an identification strategy. Theirs rests on TV ads being bought largely in advance (up to 80% upfront), checked against a border strategy. With it they find mean and median long-run advertising elasticities of 0.023 and 0.014, two thirds not statistically different from zero. Source: *Econometrica* 89(4), public PDF copy (https://www.gwern.net/doc/economics/advertising/2021-shapiro.pdf). [research]
+
+**P-goals-and-economics-66.** Against Facebook field experiments, observational methods often failed to reproduce experimental ad effects even with extensive demographic and behavioural controls (15 experiments). With over 5,000 user-level features, double/debiased machine learning and stratified propensity score matching still overstated median lifts (for example 29% experimental against 83% DML and 173% SPSM for upper-funnel outcomes; 663 experiments). Sources: Gordon, Zettelmeyer, Bhargava and Chapsky (2019), *Marketing Science* 38(2) (https://econpapers.repec.org/RePEc:inm:ormksc:v:38:y:2019:i:2:p:193-225); Gordon, Moakler and Zettelmeyer, arXiv 2201.07055 (https://arxiv.org/abs/2201.07055). [research]
+
+**P-goals-and-economics-67.** Pauwels et al.'s Sponsored Brands video estimates come from brands' own adoption choices, adjusted by matching and causal Gaussian processes (one month) and propensity score stratification (one year) on 50+ observed features. Nothing was randomised. The authors did not analyse ROAS. Source: the paper's PDF (https://cdn.amazon.science/d1/66/f222c8524c4d87fc1b0b342345b7/sponsored-brands-video-rings-up-clicks-and-sales-in-the-short-and-long-run.pdf). [research]
+
+**P-goals-and-economics-68.** Amazon's scientists run price experiments as switchbacks over time, because all shoppers see one price at a time. They report that random-days switchbacks cut standard error by 60%, and that carryover arises when treatment-raised sales lead recommendations to show a product more. Amazon-authored work on advertiser experiments states that auction competition between treated and control advertisers biases simple advertiser randomisation. Sources: Cooprider and Nassiri (https://www.amazon.science/blog/the-science-of-price-experiments-in-the-amazon-store); Jain, Hut, Islam and Pan (2023) (https://www.amazon.science/publications/cross-unit-spillovers-in-a-b-testing-empirical-evidence-from-ads). [research]
+
+**P-goals-and-economics-69.** Jonathan Hershaff, a seller, reports a daily on/off switchback of his Amazon ads over about a month. It gave an incremental ROAS of 1.25 on total sales against a platform ROAS of about 2.35, and he treats 1.25 as a floor because a daily switchback ignores carryover. Source: *Estimating Amazon Ad Incrementality for Sellers* (https://whatstheimpact.substack.com/p/estimating-amazon-ad-incrementality). [opinion]
+
+**P-goals-and-economics-70.** Repricer.com (Colin Palin) and AMALYZE hold that price changes move a listing's Featured Offer status and conversion rate, and through them its ACoS. AMALYZE says repricing makes an ACoS target "a moving goalpost". Ad Badger says coupons lift conversion by about 5–10% whatever the discount size. Quartile reports a 15–20% TACoS reduction for coordinated ads and promotions. None of these pages publishes data or a method. Sources: https://www.repricer.com/blog/amazon-ppc-vs-repricing-buy-box-impact/; https://amalyze.com/blog/price-cart-repricer-ppc-impact; https://www.adbadger.com/blog/amazon-coupons/; https://www.quartile.com/blog/why-the-smartest-brands-on-amazon-are-rethinking-promotions-in-2026. [opinion]
+
+### Seat's positions (enter no skill)
+
+- [seat] ACOS = CPC ÷ (conversion rate × average order value), so break-even CPC = average order value × margin × conversion rate, and Amazon's "net profit per sale > 0" test is the same as ACOS < margin.
+- [seat] TACoS = ACOS × (attributed sales ÷ total sales). With one margin across units, profit after ads as a share of total sales = margin − TACoS.
+- [seat] Ads Math's margin names only cost of goods sold. A margin that leaves out referral and FBA fees overstates break-even ACOS by those fees' share of price.
+- [seat] The Dorfman–Steiner condition is equivalent to a marginal total-sales ROAS of 1 ÷ margin. Under diminishing returns, a campaign's average ACOS sits below the ACOS of its last dollar. Example: going from $1,000 at ROAS 3.2 to $1,600 at ROAS 2.8 adds $1,280 of sales on $600, an incremental ROAS of about 2.13.
+- [seat] Bid-from-ACoS formulas hold on paid CPC. Paid CPC equals the bid whenever the bid is under the soft reserve, so the bid–CPC ratio a seller observes decides how far a formula's target is met.
+- [seat] `costPerNewToBrandPurchase` charges all spend to NTB purchases. It is an upper bound on the cost of a new customer, not that cost.
+- [seat] Halo sales (`salesOtherSku7d`) carry the margins of the SKUs that sold, not of the advertised SKU.
+- [seat] For a SKU being cleared, the comparison is ad cost per unit against the per-unit cost of removal, disposal, liquidation or continued storage and aged-inventory surcharge, not against margin.
+- [seat] Ad-driven units shorten days of supply one for one. Stock cover and the ad budget are therefore one decision when cover is short, and capacity limits cap the replenishment that could relieve it.
+- [seat] A Sponsored Products campaign has no goal field in Ads API v1. Its purpose exists only as the seller's label, outside the campaign object (the seller's words, a portfolio name, notes).
+- [seat] Attributed sales are gross of returns, so spend over kept revenue is ACOS ÷ (1 − return rate on revenue). A per-unit fee change of Δ moves break-even ACOS by Δ ÷ price. A JP points rate p lowers break-even ACOS on the full price by p.
+- [seat] With price at its own optimum, the Dorfman–Steiner ratio of ad spend to sales equals advertising elasticity ÷ |price elasticity|. The cheaper lever for the same extra sales depends on the ratio of the two elasticities, neither of which a seller's own regression of sales on spend or on price recovers without randomised variation.
+
+## Verification
+
+Verified 2026-11-09 by an adversarial verifier seat.
+
+**Method.** Every page in *Pages read* was re-opened on 2026-11-09 (76 rows, plus one page the verifier added). Amazon Ads library, objective, resources and news pages, sell.amazon.com pages, the two Seller Forums staff posts, the practitioner pages and the research abstract pages were fetched by plain GET with a browser user agent, converted to text, and each quoted string was checked by exact match after normalising curly quotes, dashes and whitespace; misses were read in context. Ads API docs were read from the cloudfront markdown host (`d3a0d0y2hgofx6.cloudfront.net/en-us/...md`) and parsed per field (program eligibility, description, v3 name), and the v1 enums and paths were read from the merged OpenAPI (`AmazonAdsAPIALLMerged_prod_3p.json`). SP-API docs were read as `developer-docs.amazon/....md`, and the models (Product Fees v0, Finances 2024-06-19, Replenishment 2022-11-07, the Sales and Traffic schema) from the `amzn/selling-partner-api-models` GitHub repository (raw copies of the cited blob URLs). The FTC press release was re-read with WebFetch (a plain GET returns 404). Page titles were taken from each page's `<title>`, H1 or front matter; dates and authors from the page's structured data. Searches were run for 2025–2026 changes that could overturn a claim: an SP ROAS auto-bidding launch (still only practitioner reports, the earliest dated 2026-02-03 and 2026-02-18; no Amazon page and no v1 `BidStrategy` value), a native TACoS metric (none found in the v1 metric library, the v3 columns or the merged OpenAPI), 2026 FBA fee and low-inventory-level-fee changes (practitioner reports only; the file states no figure from them), 2026 capacity-limit changes (practitioner reports only), and Amazon's reserve-price help content (not found by search). Every positions line was re-read for behavioural wording.
+
+No Ads API host (advertising-api*.amazon.com), no Login with Amazon host (api.amazon.com/auth), no Amazon Ads MCP server endpoint, no SP-API host and no Seller Central `mons-api/GetHelpTopic` call was made. Hosts called: advertising.amazon.com, d3a0d0y2hgofx6.cloudfront.net, d1y2lf8k3vrkfu.cloudfront.net, developer-docs.amazon, raw.githubusercontent.com, sell.amazon.com, sellercentral.amazon.com (forum posts and help-page shells by plain GET only), www.aboutamazon.com, www.ftc.gov (via WebFetch), the practitioner sites, en.wikipedia.org, ideas.repec.org, econpapers.repec.org, www.nber.org, arxiv.org, www.amazon.science, www.gwern.net, and web search.
+
+**What held.** All Amazon quotations in sections A–L matched their pages on 2026-11-09, including Ads Math's formulas and table, the ACOS and budget guides, the FAQ, Pricing transparency, the SB goal page (datePublished 2023-10-25), Ads Agent and unBoxed 2026 (both datePublished 2026-09-29), the fee, FBA and Subscribe & Save pages, Amazon's FTC response (all quoted sentences and figures), both staff forum posts (dates 2024-03-14 and 2026-08-13), the campaign entity table, the v1 metric eligibilities (NTB purchases, sales, cost per NTB purchase and percent NTB are SP-eligible; NTB ROAS, long-term sales, long-term ROAS and S&S subscriptions are not), the v3 columns' report types (no NTB column in any `sp*` report type, which the v3 campaign report page also bears out), the rule-based bidding figures, the AMC dataset notes, the Product Fees model and tutorials, the FBA and Analytics report-type columns, and the Replenishment and Finances models. Practitioner quotations matched, with the attribution fixes below. Research abstracts matched.
+
+**Corrections (was → now).**
+
+1. Tags line: no tag for a government source → added "[government]" for the FTC's words, which are neither Amazon's nor research.
+2. Section E and P-goals-and-economics-13: FTC press release tagged "[official, government]" / "[official]" → "[government; not Amazon's words]"; *Pages read* type "official (government)" → "government (FTC; not Amazon)". [official] is kept for Amazon's own words only.
+3. Section E: Amazon's FTC response cited at `aboutamazon.com/news/company-news/...` with the md copy secondary → the HTML URL returned 404 at verification; the `/md/` copy is named as the text read (section E, *Pages read*, *Unreachable*).
+4. Section F goal table: Awareness metrics given as CTR only ("the guide lists CTR under each goal") → "campaign reach, cost per mille (CPM), click-through rate (CTR)", as the guide lists; Purchase and Loyalty rows gained the guide's "Sales per visitor, sales per order, units per order".
+5. Section G `KPI`: three deprecated values → four (`COST_PER_ACTION`, "Please use COST_PER_CONVERSION_OFF_AMAZON", was missing), each with its replacement.
+6. Section G `AdProduct`: "includes `FULL_FUNNEL`" → seven values, including `FULL_FUNNEL` and `RMS` ("Sponsored ads for external retailers").
+7. Section G and P-goals-and-economics-19, rule-based bidding auto-disable: the quote dropped the page's condition → the rule is disabled when "your campaign is not meeting the guardrail and" the 20-day ROAS falls under 30% of the 20-to-40-day ROAS, and Amazon then "will enable your previous bidding strategy and targeting level bids".
+8. Section A and *What cx must read at run time*: no mention that Reporting v3 is being retired → the v3 columns page's banner, "Reporting API v3 will be deprecated on June 30, 2027.", is recorded where v3 columns are cited and where the v3 report call is named.
+9. *What cx must read at run time*: v1 reporting and the v1 campaign query had no operation named → `POST /adsApi/v1/create/reports` / `POST /adsApi/v1/retrieve/reports` and `POST /adsApi/v1/query/campaigns` (`QueryCampaign`), read from the merged OpenAPI; and a note that `acosClicks7d`/`roasClicks7d` are not listed for `spCampaigns` or `spPurchasedProduct`.
+10. Page names corrected to the pages' own titles: *Pricing transparency* guide → *Your guide to Amazon Ads pricing transparency*; *Amazon Ads Agent* news → *Introducing Amazon Ads Agent, your AI-powered ad platform*; *unBoxed 2026 news and announcements* → *unBoxed 2026 news: Explore the latest advertising announcements*; *Campaign reporting* → *Campaign reporting: Analyze advertising performance*; *Costs* → *Costs (Costs and fees)*; *Selling on Amazon fees* → *How much does it cost to sell on Amazon?* (short name kept); *Fulfillment by Amazon (FBA)* → *Amazon FBA (Fulfillment by Amazon)* (short name kept); "Increase sales and conversion" → *How to increase sales and conversions for your business*; Jungle Scout's title gained "Total Advertising Cost of Sale in 2023"; My Amazon Guy's title gained "and Set Campaign Targets"; Optmyzr's title gained "(And How to Diagnose It)".
+11. Attributions: My Amazon Guy "author block names Ken, COO" → Ken Zhou, Chief Operating Officer (modified 2026-10-31); AdLabs "author not named on the page" → Stephen Noch (page metadata; the text is addressed to "Stephen"); Helium 10 "Lauren" → Lauren Stair; Canopy → Chuck Kessler (page modified 2026-08-20); Perpetua → Ashley Chung; Optmyzr → Ronia, Content Strategist; Ad Badger's FTC article → byline "The Badger"; Nectar page dated July 22, 2025, Jason Landro "Co-CEO and Co-founder".
+12. Canopy quotation: two sentences quoted as one run → split with an ellipsis; the page puts "Break-even ACoS = (18 / 50) x 100 = 36%" between them.
+13. AdLabs formula: "Target CPC = RPC × Target ACOS" → "Target CPC = RPC × Target ACOS%", as printed.
+14. Li, Yuan, Yang, Chen and Song (2026) quotation: "was followed by an approximately 15-percentage-point reduction" (not the abstract's words) → "supported budget and traffic strategy adjustments that were followed by an approximately 15-percentage-point reduction in the measured cannibalization rate", "across multiple global TikTok markets".
+15. Sethuraman, Tellis and Briesch (2011), *Published research*, *Pages read* and P-goals-and-economics-42: cited as the JMR article through an IDEAS record → the IDEAS record is the French-language version in *Recherche et Applications en Marketing* (2011, DOI 10.1177/076737011102600204), whose abstract carries the same figures; the JMR page itself was not reached.
+16. Section K: Capacity Manager described from a 2024 forum post only → added Amazon's own announcement (*Amazon announces new streamlined FBA capacity management system*, dated March 1, 2023: a single monthly limit announced in the third full week of each month; reservation fees offset by performance credits "designed to offset up to 100% of the reservation fee"), and a note that practitioner reports of 2026 capacity cuts are not confirmed by any Amazon page read.
+17. Seat's positions: "Its purpose is the seller's label, which cx keeps outside the campaign" (words about what cx does) → a fact: the purpose exists only as the seller's label outside the campaign object.
+
+**Removals.** None. No position was worded as behaviour (no list of when to ask, fixed sequence, threshold as a rule, ceiling, or way of talking), and every numbered position carries a source; the one seat sentence that described what cx does was reworded in place (correction 17).
+
+**Pages that stayed unreachable.** Seller Central help GAFNWEYTJUV2GBFC (capacity limits), G202174810 (IPI) and GV43F6S76Y9DHYRH (low-inventory-level fee): re-tried by plain GET, still 336-word JavaScript shells. sell.amazon.com/fulfillment-by-amazon/fees: still "Page not found" (404). The aboutamazon.com HTML page of the FTC response: 404 (the `/md/` copy was read). The Sethuraman, Tellis and Briesch JMR article: no English record reached, and the public PDF could not be read (poppler fails; custom font encoding). The rest of the *Unreachable* list (SSRN 1866002, EconBiz, ScienceDirect, LandingCube, the Teikametrics 2022 blog, Amazon Ads India NTB help, the SP-API Product Fees use-case guide, the AER original, the Revenue Calculator) was not re-tried beyond what the list records; nothing in the file rests on those pages.
+
+### Gap fill 2026-11-09
+
+Six gaps from three completeness critics and one contradiction, worked on 2026-11-09 by the owning research seat. Every page added was read that day by plain GET (or a public copy), under the brief's limits: no Ads API, SP-API or LWA host, no MCP endpoint, no `mons-api` POST, no signed-in page, no seller data.
+
+**Added.**
+
+1. **Gap 1, FBA cost list.** Section C, "Fee changes by date": the 2026 letter (+$0.08, January 15, 2026), the 3.5% fuel and logistics surcharge (April 17 / May 2, 2026), holiday peak fulfillment fees 2026–27 (October 15 to January 14, about $0.32) with deal fees and the AWD off-peak storage note, the 2027 letter (+$0.19; storage utilization surcharge ends and the aged surcharge merges into an age-based rate on February 1, 2027; new low-inventory-level test), Data Kiosk's itemised fee types (storage utilization surcharge, aged, LILF, returns processing and high return rate fees), PPC Land's report of the 2027 price bands and returns-fee dates (as a report, not as Amazon's words), and SFP's seller-side costs. Seat bullets on what dated fees do to break-even. Section M rows; open question 6 partly answered; open questions 11–12; stable item 20; positions 49–54.
+2. **Gap 2, returns.** New section O: what attributed sales deduct (cart discounts not reflected; 72-hour cancellations and payment failures; returns not listed), Data Kiosk net product sales and net proceeds, Seller Analytics `returns` (per-ASIN returned rate, 45-day lag, ship-date mapping), the FBA customer returns report, seller-fulfilled returns reports, Sales and Traffic refunds by date only, returns-fee types, Customer Feedback return insights. Practitioners: Nova Analytics, FiveX, EightX ("Returns in the margin"), disagreement 5. Run-time reads, stable items 21–23, open questions 9 and 13, positions 55–59, a seat line.
+3. **Gap 3, Amazon Points (JP).** New section P: Amazon.co.jp points terms, Points Central (default 1%, maximum 50%, seller pays one yen per point), Double Points weekends, points in Product Pricing's landed price and Buy Box threshold, points in the fee estimate, a seller thread (opinion). Run-time reads, stable item 24, open questions 10 and 14, positions 60–61.
+4. **Gap 4, price against advertising.** New section Q (Amazon on price and stock gating the ad, deals with ads, cart discounts in attribution, Amazon's price switchbacks, the Dorfman–Steiner ratio of the two elasticities). *Published research*: a new subsection on price elasticity (Bijmolt et al. 2005; Chevalier and Goolsbee; Bach et al.). Practitioners: Repricer.com (Colin Palin), AMALYZE, Ad Badger on coupons, Quartile (Sandie Shin). Stable item 26, positions 62–64 and 70, a seat line.
+5. **Gap 5, Pauwels et al. 2023.** The paper's PDF was read. The entry now gives its identification (matched sample, causal multi-task Gaussian processes, propensity score stratification on 50+ observed features), says nothing was randomised and adoption was self-selected, quotes the authors' "we did not analyze ... ROAS", and cites Gordon et al. (2019; 2023) on how far such methods miss experiments. Position 67. No position or stable item had carried the figures as causal; the transfer note had called the study evidence on sales growth, which it now qualifies.
+6. **Gap 6, estimating ε.** Under the Dorfman–Steiner subsection: why a seller's own regression is biased (spend set with demand; Amazon's dynamic bids move with conversion likelihood), Shapiro, Hitsch and Tuchman's ad-buying-process identification and border strategy with their 0.023/0.014 elasticities, Lewis and Rao on selection, Gordon et al. 2019 and 2023, and the experimental route (Hershaff's daily switchback; Amazon's balanced switchbacks, Masoero et al. 2025; auction spillovers, Jain et al. 2023; Lewis and Wong). Stable item 25, positions 65, 66, 68, 69.
+
+**Settled.**
+
+- **Contradiction 1, whether Amazon defines TACoS.** Both Amazon pages were re-opened on 2026-11-09. *The author's guide to Sponsored Products* (dateModified 2026-10-09) says "TACOS shows your ad spend as a percentage of total sales". *Holiday advertising myths: What experts want you to know* (dateModified 2026-10-09) says "TACOS reveals whether your total sales (organic and paid) are growing in relation to your ad spend". So Amazon names TACOS and gives its ratio on two library pages. Neither names fields, and no API metric exists in what was read. This file was wrong to call the author's guide "the only Amazon page read here that defines TACoS". Section J, section N item 6, P-goals-and-economics-29 and the author's guide row in *Pages read* now say so, and a row for the myths page was added. The owner, `areas/ads-organic-and-total.md`, already reads this way in its P-ads-organic-and-total-6 ("Amazon names TACOS on two of its guidance pages ..."), so this file now agrees with it. Restating P6 is the owner's, and it needs no change on this file's evidence.
+
+**Corrected in the task's own framing.**
+
+- **Sales and Traffic refund fields.** The task named "Sales and Traffic refund fields" as a per-ASIN return source. The schema has `unitsRefunded` and `refundRate` only in `salesAndTrafficByDate`, across the catalog, and `refundRate`'s description inverts its own ratio. Per-ASIN returns come from Data Kiosk Seller Analytics or the returns reports instead (section O).
+- **"Attributed sales are not net of returns (only of cancellations within 72 hours)."** Amazon's list also deducts payment failures, and it says attributed sales do not reflect cart discounts. Its silence on returns is stated as silence (open question 9), not as a rule that returns are never deducted.
+
+**Declined, in part.**
+
+- **Gap 1, "MCF costs for sellers not on FBA".** MCF fulfils orders from the seller's own site and other channels, not Amazon-store orders, so its fees are not a cost of an Amazon ad sale. Amazon's MCF page describes it as serving "all your sales channels—Amazon, DTC, social platforms, and more—by using a single pool of inventory", where an Amazon-store order from that pool is an FBA order with FBA fees. The file states this (section C, P-goals-and-economics-54) and gives MCF's 2026–27 fee dates, but no MCF rate card. For sellers not on FBA, the cost is the seller's own fulfilment cost, which Amazon's economics schema asks the seller to enter. SFP's free delivery and free returns are stated.
+- **Gap 1, peak storage rates, low-price FBA rates and returns-processing thresholds as Amazon figures.** The pages carrying them are Seller Central help-hub shells (*Unreachable*). The file gives practitioner and press figures, attributed and marked as disagreeing, and lists the Amazon figures as open questions 11 and 12. No Amazon figure was inferred.
+- **Gap 4, Canopy, Helium 10 and Pacvue on a price cut against a bid raise.** No such page of theirs was found. Other named vendors were used, and the absence is recorded.
+
+**Checks.** Every quote added was matched against the saved page text or schema on 2026-11-09. Positions 49–70 are facts with sources (Amazon rules, documented mechanisms, measured effects, or a named practitioner's claim as theirs). None is a list of when to ask, a sequence, a threshold rule, a ceiling on what cx may change, or a way of talking. Figures are for human readers only. The seat's arithmetic is tagged [seat] and sits outside the numbered positions.