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

package/skills/ads-economics/SKILL.md

@@ -0,0 +1,40 @@
+---
+name: ads-economics
+description: "What a seller's sponsored ads earn: Amazon's ACOS, ROAS and break-even, Ads Math, the fees and returns behind the margin, attributed against ordered and total sales, TACOS, new-to-brand and long-term sales, what organic rank and published research show, and the vendor and hybrid side. For judging whether Sponsored Products, Sponsored Brands or Sponsored Display ads pay, and setting their targets."
+---
+
+## Break-even
+
+- Amazon defines ACOS as ad spend over ad-attributed sales and ROAS as its inverse, and calls the profit margin break-even ACOS; break-even ROAS is 1 ÷ margin. (G1)
+- Amazon states no general threshold makes an ACOS or ROAS good, and that new campaigns tend to run a high ACOS without failing. (G2)
+- Ads Math's affordable CPA is LTV × profit margin; its per-sale calculator ignores repeat customers. (G3)
+
+## Margin
+
+- The margin before ads rests on Amazon's fees; Product Fees estimates are not guaranteed, and `listTransactions` returns actual charges. (G4)
+- Attributed sales deduct purchases cancelled within 72 hours; returns are not among the listed deductions. (G7)
+
+## Attributed against total
+
+- Amazon dates an attributed purchase to the ad interaction in the campaign's time zone; Sales and Traffic dates ordered sales to the period. (G20)
+- A seller's Sponsored Products attributed sales include other products the same merchant sells. (G20)
+- A new-to-brand purchase is by a shopper with no purchase from the brand in the previous 12 months, judged on all shopping within the brand on Amazon, not only attributed purchases. (G10)
+- Amazon's help says new-to-brand data "isn't available for Sponsored Products"; v1's *Purchases (new to brand)*, https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/ads-v1/metrics/amazon-retail-conversions/purchases-new-to-brand.md, lists it for new-to-brand purchases and sales. (G10)
+- Long-term sales and ad-attributed Subscribe & Save cover Sponsored Brands and Display among others, not Sponsored Products. (G11, G12)
+- No Ads API metric, report column or help page read defines TACOS. (G14)
+- A total-account series alone does not separate the ads' effect from price, deals, stock or seasonality. (G26)
+- In published field experiments, attributed or observational estimates of advertising returns exceeded the experimental ones. (G19)
+- No Amazon source read states whether ad-attributed sales feed organic rank; no Amazon API read returns organic position. (G21, G28)
+
+## Vendors
+
+- A vendor's attributed sales are retail sales; its receipts are shipped COGS. (G33)
+
+## References
+
+- references/margin.md: ratios, Ads Math, goals, targets.
+- references/fees.md: fee estimates, peak fees, returns, points, capacity.
+- references/total-sales.md: other-SKU sales, new-to-brand, long-term sales, TACOS, organic rank, search counts.
+- references/evidence.md: published experiments, Amazon's carry-over study.
+- references/vendors.md: vendors, hybrid.
+- ../ads-review/references/metrics.md: attribution windows by ad product, metric definitions.

package/skills/ads-economics/references/evidence.md

@@ -0,0 +1,39 @@
+# What published research measures
+
+## Advertising, price and margin
+
+- In economics the Dorfman–Steiner condition sets the profit-maximising advertising-to-sales ratio at the price-cost margin times the advertising elasticity of demand, for a firm raising sales by price or by advertising. (G17)
+- Meta-analyses put brand advertising elasticity at about 0.12 short term and 0.24 long term, and price elasticity at −2.62. (G17)
+- TV advertising showed negative marginal ROI for more than 80% of 288 brands. (G17)
+- Click curves against bid are concave beyond a point. (G17)
+- 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. (G18)
+- Bach et al. find strong product-level heterogeneity in price elasticity among toy cars on Amazon.com. (G18)
+
+## Attributed against experimental returns
+
+- In published field experiments, observational or attributed estimates of advertising returns exceeded the experimental ones. (G19)
+- At eBay, paid-search returns were a fraction of non-experimental estimates, with no measurable short-term benefit from brand-keyword ads and effects only on new and infrequent users (Blake, Nosko and Tadelis). (G19)
+- A geographic Google-ads shutoff lost only 63% as many customers as a non-experimental estimate implied. (G19)
+- On Meta, observational methods often failed to reproduce experimental lifts even with over 5,000 user features (29% experimental against 83% DML and 173% SPSM). (G19)
+- On Meta, 7-day last-click attribution predicted experimental incrementality with R² 0.19, against 0.88 for a model trained on experiments. (G19)
+- Advertising is not randomly assigned in observational data; with an identification strategy, Shapiro, Hitsch and Tuchman found mean and median long-run TV elasticities of 0.023 and 0.014. (G19)
+- Customers acquired through Google search ads at one firm were worth more than first-purchase measures showed. (G19)
+
+## Sponsored listings and organic outcomes
+
+- On marketplaces and app stores, field experiments found sponsored listings' effect on organic outcomes negative in some settings and positive in others, and none measures a seller's Sponsored Products against its Amazon organic sales. (G25)
+- On a brand search, the brand's own ad added little without competitors and more with them. (G25)
+- On Amazon, one position up raised sales quantity by about a third of a percent. (G25)
+- A mobile game's app-install ad shutoff cut organic installs by 20–30%. (G25)
+- On 2023 Amazon.com first pages the median ad appeared 17 ranks above the same product's organic listing, with 48.3% of ads having an organic duplicate there. (G30)
+- On a shopper panel, being sponsored moved a product up 7 positions on average. (G30)
+- Over 4,800 simulated searches in four stores, items beyond organic rank 100 appeared as sponsored above the top organic results, top sponsored results mostly costlier and of poorer quality than top organic ones. (G30)
+- On the Kindle Daily Deal page, Amazon-published ebooks ranked 23.6% better conditional on observables and sold 50% less conditional on rank, quantities inferred from sales ranks. (G30)
+
+## Amazon-authored studies
+
+- Amazon Ads' Qin (2023), on 15 US brands that asked for a carry-over study (weekly 2021–22 data, a seven-week VARMAX impulse response), found that "SP does not show significant carry-over effects beyond three weeks for any brand". (G31)
+- Qin found that lower-funnel products realise 60–90% of their effect within two weeks, against about 40% of DSP display and about 30% of Sponsored Display effect in week 5 and beyond. (G31)
+- The paper states the sample is "only … brands that asked for a long-term effects assessment" and that the estimate nets out company feedback effects and competitive response. (G31)
+- Pauwels and co-authors' Amazon-authored Sponsored Brands video estimates rest on brands' own adoption choices, adjusted by matching and causal Gaussian processes over 25,364 advertisers for one month and by propensity score stratification on 50+ observed features for one year. (G32)
+- In that study nothing was randomised, and the authors did not analyse ROAS. (G32)

package/skills/ads-economics/references/fees.md

@@ -0,0 +1,47 @@
+# Fees, returns, points and stock behind the margin
+
+## Amazon's fees
+
+- The margin before ads rests on Amazon's fees. (G4)
+- The referral fee is a category-specific percentage of the total price or a minimum, whichever is greater; the fee category may differ from the one shoppers see, and media adds a closing fee. (G4)
+- FBA charges fulfilment by price, weight and dimensions, monthly storage, aged inventory, returns processing, removal, disposal, liquidation and inbound placement; Amazon's US *Fulfillment by Amazon (FBA)* page (https://sell.amazon.com/fulfillment-by-amazon) lists them, for the US, with the storage age from which inventory is charged as aged. (G4)
+- The SP-API Product Fees API estimates fees per SKU or ASIN at a supplied price; Amazon states the estimates are not guaranteed. (G4)
+- The FBA Fee Preview report holds per-SKU estimates updated at least every 72 hours. (G4)
+- The Finances API's `listTransactions` returns the actual charges and may omit the last 48 hours. (G4)
+
+## Fees move
+
+- The 2026 US letter raised FBA fees by $0.08 a unit on average from 2026-01-15. (G5)
+- Amazon's 2027 US fee letter, *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), gives the average change in US FBA fulfilment fees per unit, most changes taking effect from 2027-01-15. (G5)
+- From 2027-02-01 the 2027 US letter drops the storage utilization surcharge and folds aged inventory into an age-based storage rate that Amazon says is flat or lower for inventory turning in under six months and higher for aged inventory. (G5)
+- From 2027-02-01 the 2027 US letter applies the low-inventory-level fee only below 28 days of supply. (G5)
+- A 3.5% fuel and logistics surcharge on the fulfilment fee applies from 2026-04-17 in the US and Canada, to Buy with Prime and MCF from 2026-05-02, and on top of peak fees; Amazon's news post *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) speaks for the US and Canada and states the rate, its dates and its average per US FBA unit. (G5)
+- Holiday peak fulfilment fees for FBA, Remote Fulfillment, MCF and Buy with Prime apply for a holiday season at dates and an average per-unit increase over non-peak rates that Amazon states for that season; for the US in 2026 they are stated in Amazon's news post *Holiday 2026: Same fees, same eligibility, earlier deadlines* (https://sellercentral.amazon.com/seller-forums/discussions/t/3e31fbb7-04e0-4ed4-873e-f74b1052e2ff). (G5)
+- The Data Kiosk economics dataset itemises the fee types, shows a fee separately on each side of a change date and previews fees up to 120 days ahead. (G5)
+
+## Fulfilment programs
+
+- Multi-Channel Fulfillment fulfils orders placed on the seller's own website or other sales channels, not Amazon-store orders. (G6)
+- Seller Fulfilled Prime commits the seller to fast free delivery and free returns for domestic Prime customers. (G6)
+
+## Returns and net sales
+
+- Attributed sales do not reflect cart discounts, and deduct payment failures and purchases cancelled within 72 hours, deductions taking up to 7 days; returns are not among the listed deductions. (G7)
+- Data Kiosk's net product sales subtract refunds by refund request time, and its net proceeds subtract fees, ad charges and seller-entered unit costs. (G7)
+- Per-ASIN returns sit in Data Kiosk's `returns` group, mapped to ship day and computed about 45 days late. (G7)
+- FBA returns are in `GET_FBA_FULFILLMENT_CUSTOMER_RETURNS_DATA` and seller-fulfilled returns in the returns reports. (G7)
+- Sales and Traffic carries refunds only by date, not by ASIN. (G7)
+- Nova Analytics, FiveX and My Amazon Guy put a returns allowance into the margin before ads, and none of their pages computes ACoS or TACoS on revenue net of returns. (G7)
+
+## Points on Amazon.co.jp
+
+- On Amazon.co.jp points are granted at a rate set by Amazon or the seller, in Points Central up to 50% per listing. (G8)
+- A default 1% has been auto-applied from November 1, 2021 to listings the seller had not set. (G8)
+- One point is used as one yen, and the seller bears one yen per point. (G8)
+- In Amazon's Product Pricing model, landed price is listing price plus shipping minus points, and an offer may be ineligible for the Buy Box when that exceeds the competitive price threshold. (G8)
+- The Product Fees estimate accepts points, a program that exists only in Japan. (G8)
+
+## Stock and capacity
+
+- Amazon staff state that FBA capacity for established accounts rests on IPI score and sales, and that IPI runs 0–1,000. (G15)
+- Amazon's inventory planning report gives per-SKU days of supply and excess. (G15)

package/skills/ads-economics/references/margin.md

@@ -0,0 +1,32 @@
+# Break-even, Ads Math and targets
+
+## Amazon's ratios
+
+- Amazon defines ACOS as ad spend over ad-attributed sales, times 100, and ROAS as its inverse. (G1)
+- Amazon states that to keep a profit ACOS must sit below the profit margin, and calls that margin break-even ACOS. (G1)
+- Amazon's Ads Math gives margin as (revenue − cost of goods) ÷ revenue and break-even ROAS as 1 ÷ that margin. (G1)
+- Amazon states there is no general threshold for a good ACOS or ROAS. (G2)
+- In Amazon's words a good ROAS is one profitable for the advertiser, read with margin, lifetime value and lift. (G2)
+- Amazon states that new campaigns tend to run a high ACOS without failing. (G2)
+- One Amazon FBA guide, *How to Help Increase Sales with FBA and Advertising*, still prints $2–4 per $1. (G2)
+
+## Ads Math's click chain
+
+- CPA = CPC ÷ conversion rate. (G3)
+- Profit per sale = average order value × gross profit margin; net profit per sale = profit per sale − CPA. (G3)
+- Amazon says this calculator is for one-off purchases and ignores repeat customers. (G3)
+- Customer LTV = average order value × average orders. (G3)
+- Affordable CPA = LTV × profit margin, which Amazon defines as the upper limit of ad spend per new customer to break even over the customer's lifetime. (G3)
+
+## Goals in Amazon's guidance and API
+
+- Amazon's Sponsored Products guidance names objectives to set before a campaign: brand exposure, new product, inventory clearance, peak season. (G9)
+- In Ads API v1, `goalSettings` is not available for Sponsored Products, while a Sponsored Brands goal is read-only with a required KPI. (G9)
+- The v1 model's `Goal`, `KPI` and `BidStrategy` enums name the options. (G9)
+
+## Practitioners' targets from margin
+
+- My Amazon Guy (Ken Zhou) computes a "true" break-even ACoS for setting campaign targets. (G16)
+- EightX takes break-even TACoS as the pre-ad margin. (G16)
+- Perpetua's Sponsored Products goal setup takes an initial target from 30 days' ACoS. (G16)
+- Helium 10 describes ad spend that follows stock levels across marketplaces. (G16)

package/skills/ads-economics/references/total-sales.md

@@ -0,0 +1,62 @@
+# Attributed sales, total sales and organic
+
+## Two sales figures
+
+- Amazon dates an ad-attributed purchase to the ad interaction in the campaign's time zone, without tax or cart discounts and net of cancellations within 72 hours. (G20)
+- Sales and Traffic counts ordered product sales as price × units for the period's date. (G20)
+- For a seller's Sponsored Products, attributed sales include the advertised product and any other product the same merchant sells. (G20)
+- Reporting v3 splits them by `SameSku`/`OtherSku` and `purchasedAsin`, and v1 by the Promoted and Halo qualifiers. (G20)
+- v1's base conversion metrics include halo; the promoted-only figure carries the `Promoted` qualifier. (G20)
+- Sponsored Products and Sponsored Brands CPC campaigns report 0 for conversions from views. (G20)
+- `getOrderMetrics` takes an IANA time zone for the day boundary, so seller-side daily ordered sales can be cut on the ads profile's time zone. (G23)
+- No Amazon page read states the time zone of a Sales and Traffic report day. (G23)
+- Data Kiosk Economics returns, per MSKU and day, ad charges by ad type beside ordered and net product sales; those charges are spend, not attributed sales. (G23)
+- ../ads-review/references/metrics.md: the ads metrics, their definitions and attribution.
+
+## New-to-brand
+
+- A new-to-brand purchase is one by a shopper with no purchase from the brand in the previous 12 months, judged on all shopping within the brand on Amazon, not only attributed purchases. (G10)
+- Amazon's new-to-brand help page says the metrics are for Sponsored Brands and Display and the data "isn't available for Sponsored Products". (G10)
+- Reporting v3 carries new-to-brand columns in no Sponsored Products report type. (G10)
+- v1's *Purchases (new to brand)* page, https://d3a0d0y2hgofx6.cloudfront.net/en-us/guides/reporting/ads-v1/metrics/amazon-retail-conversions/purchases-new-to-brand.md, names Sponsored Products in the eligibility of its new-to-brand purchase and sales fields, view-attributed fields included, although Amazon reports 0 view conversions for Sponsored Products. (G10)
+
+## Long-term sales
+
+- Long-term sales is attributed sales plus the estimated 365-day value of new-to-brand shopper engagements, modelled from the brand's historical shopping patterns. (G11)
+- It is offered with long-term ROAS for Sponsored Brands, Display, TV, DSP and Full-Funnel, not Sponsored Products. (G11)
+- Amazon calls it an ad-attributed measure that estimates incremental sales value, and no Amazon page read describes a holdout or unexposed comparison in its calculation. (G11)
+- Since 2026-09-25 its US estimates use machine-learning models in place of heuristics, with 30-, 60-, 90- and 180-day horizons beside 365 days. (G11)
+- Its cohort fields (`metric.cohortAccumulatedSales*`), which Amazon presents for checking estimates against actual cohort sales, require `campaign.id` and "aren't summable across cohorts". (G11)
+
+## Subscribe & Save and repeat purchase
+
+- Subscribe & Save subscriptions attributed to ads are reported for Sponsored Brands, display, TV, DSP and Full-Funnel, not Sponsored Products. (G12)
+- The Repeat Purchase report and the Replenishment API report repeat and subscriber value. (G12)
+- Subscribe & Save needs Brand Registry, starts at a 0% seller-funded discount and can block one-time purchases under `REPLENISHMENT_ONLY_ORDERING`. (G12)
+
+## TACOS
+
+- Amazon names TACOS on two Amazon Ads pages, once as ad spend as a percentage of total sales and once as total sales "(organic and paid)" set against ad spend. (G14)
+- Neither page names the sales field, the ad products counted or the window, and no Ads API metric, report column or help page read defines it. (G14)
+- The TACOS benchmark ranges read (Helium 10) cite no data source, a point PPC Land makes in general. (G14)
+
+## A total-account series
+
+- A total-account series (ad spend, attributed sales, ordered sales, their ratio) shows how the quantities moved together over the dates read. (G26)
+- Price, deals, stock, Featured Offer share, reviews, seasonality and competitors' actions move ordered sales over the same dates, and the series alone does not separate their effect from the ads'. (G26)
+
+## Organic rank and search counts
+
+- No Amazon help page, API document, policy or staff reply states whether ad-attributed sales feed organic rank; Amazon's words stop at Sponsored Products helping offers "rise through search results". (G21)
+- No Amazon API or report read returns a product's organic search position. (G28)
+- Catalog Items returns Best Sellers Rank, the Ads API an advertiser's paid impression share and rank per search term, and Brand Analytics search-funnel counts and shares. (G28)
+- The console's Search Term Impression Share report is listed as not available in the API. (G28)
+- Every organic-rank series is a third-party sample (Helium 10, SmartScout, DataDive, Jungle Scout), none read validated against Amazon data. (G28)
+- BSR is relative to the category, and its mapping to quantity is modelled as a power law (Chevalier & Goolsbee's slope 0.9 to 1.3 on 2001 book data). (G28)
+- Keepa's `monthlySold` ("bought in past month") is a variation-specific floor bucket most ASINs lack, moving only at bucket boundaries. (G28)
+- Search Query Performance gives, per query, an ASIN's and the brand's impressions, clicks, cart adds and purchases on the search results page beside the query's totals and shares, by Sunday–Saturday week, month or quarter. (G22)
+- It is for brand-registered sellers with the Brand Analytics role, through SP-API with an `asin` list of at most 200 characters. (G22)
+- Neither Search Query Performance nor Search Catalog Performance has a field separating paid from organic, and Search Catalog Performance's impressions "include sponsored search results". (G22)
+- Trellis reads Search Query Performance's counts as organic and paid combined. (G22)
+- ../ads-review/references/reports.md: the ads reports.
+- The amazon-ads skill's references/door.md: how a report is asked for and landed.

package/skills/ads-economics/references/vendors.md

@@ -0,0 +1,62 @@
+# Vendor and hybrid economics
+
+## A vendor's sales and receipts
+
+- A vendor's attributed sales are retail sales, which Amazon states "don't reflect any discounts applied in the cart" and "don't include tax". (G33)
+- Amazon defines the vendor's own receipts separately, as shipped COGS, "the price Amazon paid to a vendor for item procurement". (G33)
+- A vendor's total-sales side is Vendor Retail Analytics, by `MANUFACTURING` or `SOURCING` view, reported in Pacific Standard Time whatever the marketplace. (G27)
+
+## Vendor Sales report
+
+- The Vendor Sales report requires `reportPeriod`, `distributorView` (`MANUFACTURING` or `SOURCING`) and `sellingProgram` (`RETAIL`, `BUSINESS` or `FRESH`). (G35)
+- The manufacturing view covers ASINs the vendor manufactures, whoever sourced them to Amazon; the sourcing view covers ASINs sourced directly from the vendor group. (G35)
+- `orderedRevenue` and `orderedUnits` are populated only in the manufacturing view, while shipped revenue, shipped COGS and shipped units are in both. (G35)
+- Amazon's SLA is end of day Monday (48 hours) for a weekly period and 72 hours for a non-Saturday daily period. (G35)
+- A report requested before its period is available ends with `processingStatus` `FATAL`. (G35)
+- DAY data looks back 1,460 days with a 15-day maximum span. (G35)
+- In the Data Kiosk vendor dataset, the view a vendor sees follows the role assigned to it, and the sourcing view is the distributor view and has no glance views. (G35)
+- In that dataset a vendor group that manufactures nothing gets an error for the manufacturing view, and several manufacturing metrics cannot be grouped by `vendorCode`. (G35)
+
+## Glance views and real-time reports
+
+- Amazon counts vendor glance views only when Amazon retail is the featured merchant: Data Kiosk's `glanceViews` are "Glance views where Retail is the featured offer". (G36)
+- Data Kiosk's `lostFeaturedOffer` is the share of glance views where the vendor's offer does not win the Featured Offer. (G36)
+- The real-time traffic report's glance views differ from the Vendor Central dashboard's, are less accurate for concluded days and arrive about 65 to 115 minutes after the hour. (G36)
+- The real-time sales report gives hourly ordered units and revenue five minutes after the hour, with cancellations shown at their own time as negative units. (G36)
+- Amazon's tip computes conversion as ordered units over glance views. (G36)
+- Reason Automation (Andrew Hamada, 2025) reports that Vendor Central renamed glance views to "featured offer page views", which in their account count only when a 1P offer is featured, appear only in the manufacturing view and come with no sessions for 1P vendors. (G36)
+
+## Net PPM
+
+- Amazon gives two current definitions of Net PPM, its own margin on a vendor's product. (G34)
+- The Net PPM report schema's definition is "(Product Revenue – Product COGS – Vendor Funded Coop) / Product Revenue", a fraction from 0 to 1 measuring Amazon's profitability, in a vendor-only report that takes only `reportPeriod` and is in Pacific time. (G34)
+- Data Kiosk's definition is "(Shipped Revenue - Shipped CCOGS + CCOGS - Sales Discount) / Shipped Revenue", where contra-COGS is "the dollars realized by customer outbound shipments minus Display Ads Contra COGS". (G34)
+- MerchantSpring (Seiton) and SellerApp (Purohit) state that a vendor's advertising spend is not part of Net PPM. (G34)
+- Seiton, in a MerchantSpring article of 2026-04-06, updated 2026-09-11, read 2026-11-09 and naming no marketplace, gives approximate Net PPM targets of about 40–45% for hardlines, 30–37% for softlines and 27–35% for consumables, which she says vary by sub-category and over time. (G34)
+
+## Vendor TACoS
+
+- No Amazon page read on 2026-11-09 defines TACoS for a vendor or names its total-sales denominator. (G37)
+- m19 offers three vendor TACoS variants: spend over manufacturing ordered sales, over manufacturing shipped sales and over sourcing shipped sales, and says manufacturing metrics may include products sourced from distributors and wholesalers. (G37)
+- Pare.so (George Reid, 2026) states that a hybrid brand's two TACoS denominators come from Vendor Central and Seller Central and that blending them without correction produces a meaningless metric. (G37)
+
+## Brand Analytics for vendors
+
+- Search Query Performance and Search Catalog Performance are available through the SP-API to sellers only. (G38)
+- Repeat Purchase, Market Basket and Search Terms are available to sellers and vendors with the Brand Analytics role and Brand Registry, Search Terms also requiring a brand representative; Amazon's *Analytics Reports* page (https://developer-docs.amazon/sp-api/docs/report-type-values-analytics.md) lists the stores Repeat Purchase covers. (G38)
+- AMALYTIX (Trutz Fries, October 2026) reports that Vendor Central offers vendors Top Search Terms, Repeat Purchase Behavior and Market Basket Analysis, but not Search Query Performance or Search Catalog Performance, as checked in a vendor account in October 2026. (G38)
+
+## Hybrid brands
+
+- MerchantSpring (Rachel Seiton, "The Hybrid Playbook", 2026-08-13) advises two advertising strategies for a hybrid brand: 1P products on the upper funnel and broad category terms, and 3P on the lower funnel, retargeting, branded defence and conversion promotions. (G39)
+- Seiton warns that a budget too small for both underperforms on both. (G39)
+- Seiton states, citing ChannelEngine, that Amazon has terminated smaller vendor relationships in waves since late 2024, the latest with a termination date of 2 August 2026, which no Amazon page read states. (G39)
+
+## Research on reseller and marketplace
+
+- Hagiu and Wright (2015) model a platform's choice between marketplace and reseller as a choice of who holds control over a noncontractible marketing activity, spillovers across products shifting the choice toward reselling. (G40)
+- Abhishek, Jerath and Zhang (2016) find in a stylised model that agency selling is more efficient than reselling and gives the manufacturer control of retail prices, reselling being preferred when the e-tailer gains from sales of associated products. (G40)
+- Zhu and Liu (2018) find that Amazon is more likely to enter successful third-party product spaces and less likely where products need greater seller effort to grow, its entry discouraging affected sellers from pursuing growth on the platform. (G40)
+- Farronato, Fradkin and MacKay (2023) find that Amazon-branded products rank higher than observably similar products in search, with a prominence 30–60% of that given to sponsored products. (G40)
+- Bergen and John (1997) model co-op advertising and find that more generous manufacturer participation rates are called for with less targetable media, less differentiated retailers, more differentiated brands and more upscale products. (G41)
+- Besanko, Dubé and Gupta (2005) measure a retailer's pass-through of wholesale price changes and find it varies widely across products and categories, averages above 60% own-brand in 9 of 11 categories, and includes cross-brand effects. (G41)

package/skills/ads-launch/SKILL.md

@@ -0,0 +1,34 @@
+---
+name: ads-launch
+description: "Who may run sponsored ads on Amazon and what a new product's first campaign meets: seller, vendor and Brand Registry eligibility, product and program eligibility reads, the ad products and their placements, ad policy, moderation, billing and API terms, launch programs (new-seller credits, Vine, coupons, Review Requests), how a first campaign is built, its budget suggestions and time marks, practitioners' opening bids, and what is known of new-product ranking. For opening an account's sponsored ads or launching a product's ads."
+---
+
+## Door
+
+The amazon-ads skill (cx-amazon-ads) carries every call; its references/door.md: headers, paths, reports.
+
+## Who may advertise
+
+- Sponsored Products needs a Professional plan or vendor account and an eligible category; variation parents and suppressed listings cannot be advertised (L1).
+- Sponsored Brands, display ads, Streaming TV and Stores need Brand Registry or a vendor account; vendors need not enroll to advertise (L2, L22).
+- Product eligibility rates each product and ad type `ELIGIBLE`, `ELIGIBLE_WITH_WARNING` or `INELIGIBLE` with reasons (P1).
+- No eligibility reason reads reviews; Amazon's five or more reviews at 3.5 stars is guidance (L3).
+
+## New products
+
+- Product metadata returns `createdDate`, the item's first-available date on Amazon (L4).
+- Review Requests (marketplaces and stage: its *overview*, https://advertising.amazon.com/API/docs/en-us/guides/review-requests/overview): an SP campaign with goal `ENGAGEMENT`, billed per rating or review left on its ad card (L15).
+- No Amazon page claims a "honeymoon" ranking boost in Amazon's own voice; practitioners disagree (L13).
+
+## First campaign
+
+- Amazon recommends a new advertiser start with automatic Sponsored Products and use its search term report to find manual targets; its guides differ on when manual targeting joins (L10).
+- Amazon's guides build a campaign by separate creates, in order: campaign, ad group, product ad (SKU for sellers, ASIN for vendors), then targets; an auto campaign serves once those three are active, before any negative; both create APIs accept `PAUSED` as first state (L20).
+- Before a campaign exists, the API suggests keywords, new-ad-group bids and an initial budget (`-1` when information is short); Amazon's suggested starting budget exceeds the API minimum (L11).
+- Amazon's new-campaign time marks: metrics checked at least twice weekly in week one; match types tested at least one to two weeks before keywords pause; at least 20 clicks before a keyword becomes a negative; negatives left two weeks or more before a strategy change (L12).
+- Practitioners' opening bids: Canopy Management above Amazon's suggestion, Ad Badger conservative on manual bids, ecomcrew narrow on exact match, high bids, widening later (L19).
+
+## References
+
+- references/accounts.md: accounts, Brand Registry; references/products-and-policy.md: eligibility, ad products, policy, billing; references/new-products.md: credits, Vine, roadmap, coupons, rank manipulation; references/first-campaign.md: launch ACoS bands; references/new-product-ranking.md: cold start, launch multiples.
+- Existing campaigns' serving and budget: ../ads-review/references/serving.md, ../ads-bids-and-budgets/references/budgets.md.

package/skills/ads-launch/references/accounts.md

@@ -0,0 +1,27 @@
+# Who may advertise
+
+## Sellers and Sponsored Products
+
+- Sponsored Products needs a Professional plan or a vendor account and an eligible category. (L1)
+- Adult, used, refurbished and closed-category products are excluded from Sponsored Products. (L1)
+- Variation parents and suppressed listings cannot be advertised; only the child variation is eligible. (L1)
+
+## Brand Registry
+
+- Sponsored Brands, display ads, Streaming TV and Stores need Brand Registry or a vendor account. (L2)
+- Streaming TV is not open to Health and Personal Care advertisers. (L2)
+- Brand Registry needs the brand name on the product or packaging and, in one Amazon page's words, a pending or registered trademark, in another's an active registered trademark. (L2)
+
+## Vendors
+
+- Amazon defines a vendor as a business that sells its items directly to Amazon, which then sells them to customers. (L22)
+- Amazon's FAQ lists the vendors eligible for Sponsored Products, any one criterion sufficing: Hardlines, Softlines, Supplier Express (Vendor Express) vendors with a confirmed purchase order or direct fulfillment order, and Media, Books and Consumables vendors. (L22)
+- The same FAQ requires Sponsored Products listings to be eligible for the Featured Offer. (L22)
+- It lists Hardlines, Softlines, Media, Books and Consumables (non-Pantry/Fresh) vendors for Sponsored Brands. (L22)
+- Amazon states that only sellers must enroll in Brand Registry to advertise and vendors need not, while the SP-API Brand Analytics reports require Brand Registry for vendors and sellers alike. (L22)
+
+## Linking and permissions
+
+- An advertising account can be linked to only one Vendor Central account, through Access and Settings > Selling Accounts. (L23)
+- The vendor approves a link request by signing in with Vendor Central credentials; a user not eligible to approve forwards it to someone with admin privileges on the Vendor Central account. (L23)
+- Vendor permissions in the advertising console come in three default tiers, Viewer, Editor and Admin, while seller permissions are managed in Seller Central. (L23)

package/skills/ads-launch/references/first-campaign.md

@@ -0,0 +1,32 @@
+# The first campaign
+
+## Amazon's starting advice
+
+- Amazon recommends that a new advertiser start with automatic Sponsored Products and use its search term report to find manual targets. (L10)
+- Amazon's guides differ on when to add manual targeting: about two weeks in one, at the start in another. (L10)
+- Search term reports: ../ads-review/references/reports.md; their reading: ../ads-search-terms/SKILL.md.
+
+## How a campaign is built
+
+- Amazon's guides build a campaign by separate creates in order: campaign, ad group, product ad (by SKU for sellers, by ASIN for vendors), then targets. (L20)
+- An auto campaign is complete and serves once it has an active campaign, ad group and product ad, before any negative exists. (L20)
+- Both create APIs, SP v1 and SP 3.0, accept `PAUSED` as the first state. (L20)
+
+## Recommendations before a campaign exists
+
+- For a campaign that does not exist yet, the API returns keyword suggestions for ASINs, bids for a new ad group and an initial budget recommendation. (L11)
+- The initial budget recommendation is `-1` when there is not enough information, and comes with up to five special-event budgets. (L11)
+- Amazon's guide Best practices for your Sponsored Products ads (https://advertising.amazon.com/library/guides/sponsored-products-best-practices) suggests a starting daily budget above the API's minimum daily budget, which the Ads API *Limits, constraints, and quotas* page (https://advertising.amazon.com/API/docs/en-us/reference/concepts/limits) gives per marketplace and currency. (L11)
+
+## Amazon's time marks
+
+- Amazon's guides set time marks for a new campaign: in the first week, monitor metrics at least twice a week; test match types for at least one to two weeks before pausing keywords. (L12)
+- In the same guides a keyword is evaluated after at least 20 clicks before it is added as a negative, and negatives run for two weeks or more before a change of strategy. (L12)
+
+## Practitioners' launches
+
+- Canopy Management structures a launch as exploration, intent capture and brand presence and opens bids above Amazon's suggestion. (L19)
+- Ad Badger opens manual bids conservatively. (L19)
+- ecomcrew starts narrow with exact match and high bids and widens later. (L19)
+- Canopy Management, EightX and My Amazon Guy each publish launch-stage ACoS or TACoS bands above their mature bands; the bands do not agree with one another, and each firm calls its bands starting points or guidelines. (L19)
+- ACoS and TACoS as figures: ../ads-review/references/metrics.md.

package/skills/ads-launch/references/new-product-ranking.md

@@ -0,0 +1,24 @@
+# New-product ranking and launch evidence
+
+## The "honeymoon"
+
+- No Amazon page claims a "honeymoon" ranking boost for new products in Amazon's own voice. (L13)
+- Amazon's cold-start patents give example durations such as ninety days; these are examples, not operating figures. (L13)
+- Practitioners disagree: My Amazon Guy (Steven Pope, 2021) writes "Typically 15 – 30 days", m19 "a couple of weeks… not a hardcoded duration", Best at Amazon calls it a "myth", Code3 "debunks" it, and Canopy (Chuck Kessler) writes "Amazon has never confirmed one". (L13)
+
+## Cold-start research
+
+- Amazon researchers (Han et al., CIKM 2022) report that new items "typically cross the cold start barrier through very specific queries". (L14)
+- The same researchers report that a prior from similar items with early stopping raised new-product impressions and purchases in an online A/B test. (L14)
+- Amazon-affiliated researchers (Dong et al., arXiv 2602.14502) report that a ranking model lending new items behavioural signals borrowed from their substitutes has served customers since 2025, on a platform the paper names only as "a large E-commerce platform". (L14)
+
+## Amazon's launch multiples
+
+- Amazon's launch multiples (for example 36× and 231×) compare groups of new and not-new advertised products on ad-attributed outcomes, from internal data. (L17)
+- Amazon links launch advertising to organic visibility only in hedged terms. (L17)
+
+## Field research on advertising
+
+- In field research (Dai, Kim and Luca), advertising raised Yelp businesses' purchase-intention outcomes by 7–19% and reviews by 5%, effects that disappeared once the ads stopped. (L21)
+- On Zomato (Sahni and Nair), labelling a paid listing as an ad raised calls to the restaurant by 77% against the same listing unlabelled, and the ads' incremental conversions came mostly through the advertiser's organic listing. (L21)
+- On Amazon (He, Hollenbeck and Proserpio), products that bought fake reviews saw a short-lived rise in rating and sales, with keyword search position not reversing within two months, while their sellers raised sponsored listings and cut prices. (L21)

package/skills/ads-launch/references/new-products.md

@@ -0,0 +1,52 @@
+# New products and launch programs
+
+## Reviews and ad eligibility
+
+- Amazon advises that an advertised product have five or more reviews at 3.5 stars; that advice is guidance, and no eligibility reason reads reviews. (L3)
+
+## Finding a new ASIN
+
+- Product Selector's product metadata read returns `createdDate`, the date the item was first available on Amazon. (L4)
+- The recommendation API groups unadvertised, recently added ASINs as `NEW_ASIN`. (L4)
+
+## Amazon's 30-day roadmap
+
+- Amazon's 30-day roadmap puts active listings with stock, Sponsored Products campaigns and coupons on day one. (L9)
+
+## New-seller credits and promotions
+
+- A seller new to Amazon earns Sponsored Products credits by creating a first SP campaign within 90 days of listing the first buyable ASIN and spending within 30 days of that campaign's launch. (L5)
+- Under the 2024 terms the credit is issued within 14 days after that window, is valid 30 days and cannot pay for the spend that earned it. (L5)
+- The beta v1 promotion offers and promotion rewards queries list an account's promotion offers and rewards. (L6)
+- Promotional credits apply only to future costs, may be limited to one ad product and expire under their terms. (L6)
+
+## FBA New Selection
+
+- FBA New Selection, on Amazon's page dated Jan 14, 2026, enrols a new seller automatically when the first shipment is created within 90 days of listing the first buyable product. (L18)
+- PPC Land reports revised terms from July 30, 2026 that no Amazon page read confirms. (L18)
+
+## Vine
+
+- Vine needs a Professional plan, FBA offers and a Brand Representative or Reseller role on a Brand Registry brand or generic products. (L7)
+- Vine gives up to 30 reviews per parent and takes an existing product only with fewer than 30 reviews. (L7)
+- Amazon advises enrolling in Vine with stock at least three weeks before the planned launch. (L7)
+
+## Review Requests
+
+- Review Requests is a Sponsored Products campaign with goal `ENGAGEMENT`, a goal fixed once set; the marketplaces where it runs, and its stage, are on Amazon's *Review Requests overview* (https://advertising.amazon.com/API/docs/en-us/guides/review-requests/overview). (L15)
+- It is billed per rating or review left on the ad card, serves products with fewer than 1,000 reviews and pauses at 1,000. (L15)
+- Amazon frames Review Requests as an answer to new products' cold-start problem. (L15)
+
+## Coupons
+
+- Under Seller Central's *Coupon eligibility criteria* (GJHCPEFJ5JJQD52D) as captured in version 14 on 2026-10-14, a capture that names no marketplace, a product with no reviews is not barred by the rating rule, whose floors apply to products with 1–4 reviews (2.5 stars) and 5 or more reviews (3 stars). (L8)
+- Amazon's full new-product guide says "Coupon quality validation rules do not apply to new products". (L8)
+
+## New & Notable and newness messaging
+
+- New & Notable marks products new to market in the last 12 months, subject to eligibility. (L18)
+- Amazon reports at least 22% higher click-through for ads that message newness. (L18)
+
+## Rank manipulation
+
+- Amazon's Seller Code of Conduct, as clarified on the Seller Forums, prohibits search-rank manipulation by any name, including search-find-buy and two-step URLs. (L16)

package/skills/ads-launch/references/products-and-policy.md

@@ -0,0 +1,91 @@
+# Ad products and policy
+
+## Product eligibility
+
+- Amazon's product eligibility read takes up to 50 products, only products sold by the profile's merchant, and returns per product and ad type an overall status: `ELIGIBLE`, `ELIGIBLE_WITH_WARNING` ("may not receive impressions") or `INELIGIBLE` ("cannot be included in an advertisement"). (P1)
+- The overall status inherits the highest severity among the named reasons, such as `NOT_IN_BUYBOX`, `OUT_OF_STOCK`, `LISTING_SUPRESSED`, `MISSING_IMAGE`, `RESTRICTED_CATEGORY` and `VARIATION_PARENT`. (P1)
+- The product metadata read returns availability, price, best seller rank and eligibility for up to 300 products. (P1)
+
+## Program eligibility
+
+- Amazon's program eligibility read returns per marketplace and program (`SB`, `SD`, `SPOT`, `DTC`, `MAAS`) whether the account may use it, with reasons covering standing, terms, brand relation, business verification, tax, billing and payment. (P2)
+- Most advertisers are verified automatically; the others are verified through verification tasks. (P2)
+- Sponsored Brands does not support adult, used or refurbished products and is limited to a listed set of categories. (P3)
+- Streaming TV needs a professional seller with Brand Registry or a vendor account, is not available to Health and Personal Care advertisers and is offered through the API in the marketplaces, and at the stage, that Amazon's *Sponsored TV overview* gives (https://advertising.amazon.com/API/docs/en-us/guides/sponsored-tv/overview). (P3)
+
+## What an ad shows
+
+- A Sponsored Products ad includes the product's title, image and price, the list price when the offer is below it, and the Prime badge when the offer has Prime shipping. (P10)
+- Amazon's responsive display creatives add live deal, coupon and Subscribe & Save badges and star ratings from the listing, showing the highest discount when several are active. (P10)
+
+## Sponsored Products video
+
+- SP video runs only in the marketplaces Amazon's *Understand Sponsored Products video* names (https://advertising.amazon.com/help/G45ASMDJY8XXQQFK), is not for Books, is billed per click from the campaign budget and is approved in 48–72 hours. (P4)
+- It can be auto-assembled from existing detail-page or uploaded video, with no new video generated. (P4)
+- Amazon's help page calls auto-assembly "an opt-in feature", while the API guide says "All campaigns are opted in to auto-assembled videos by default". (P4)
+
+## Prompts
+
+- Prompts are an AI-generated extension of SP and SB campaigns in the marketplaces Amazon's *Understand prompts on Sponsored Products and Sponsored Brands* names (https://advertising.amazon.com/help/GLWNW26EWQZW46WZ), built from detail-page, Brand Store and campaign data, auto-enrolled, billed per click from the campaign budget and pausable per prompt. (P11)
+- A prompt can open a Rufus dialog, can show in shopping results and on product pages, and is not available to authors and publishers. (P11)
+- Prompts have been generally available since 2026-03-25, a date Ad Badger agrees on. (P11)
+- They are reported in the `spPromptAdExtension` and `sbPromptAdExtension` v3 report types, whose data delivery ends with v3 on 2027-06-30. (P11)
+
+## Alexa for Shopping, Rufus and Alexa+
+
+- In May 2026 (announced 2026-05-13) Amazon brought Rufus and Alexa+ together as Alexa for Shopping for US shoppers on the Amazon Shopping app, website and Echo Show, with Alexa+ continuing on Alexa-enabled devices; outside the US the assistant is Rufus. (P12)
+- Existing sponsored ads campaigns are automatically eligible to serve in Alexa for Shopping without additional setup, and advertisers with sponsored ads campaigns are automatically enrolled in Alexa+ Conversational Ads. (P12)
+- No Ads API release note, guide or contract names Alexa for Shopping, and no report field or placement value separates Alexa for Shopping or Alexa+ traffic, the v1 inventory type `ALEXA_DEVICES` aside. (P12)
+- Amazon said in 2024 that reports would not carry Rufus metrics, and in 2025 that Alexa+ metrics would be consolidated in existing reporting. (P12)
+- No page says which bid adjustments apply in Alexa for Shopping, and Branded Conversations has no API or date. (P12)
+
+## Sponsored Services
+
+- Sponsored Services, announced at unBoxed 2026, is a US campaign type for local and online service providers bought with Cost Per Outcome bidding, available only through the Amazon Ads Unified API from Q4 2026, with console access in 2027. (P14)
+
+## Creators
+
+- Creator Connections lets a brand offer creators a commission on net price for sales within 24 hours of a click. (P15)
+- It needs a Brand Registry role and an Amazon Ads account advertising in, and based in, the countries Amazon's *Understand Amazon Creator Connections* lists (https://advertising.amazon.com/help/GPARAL6JVGFEJKG2). (P15)
+- Its results are reported only in its own portal at campaign level, and its charges appear on the advertising invoice; in India the same name will cover SP creator content. (P15)
+- An invite-only Sponsored Content Requests program inside Creator Connections is described by trade press only. (P17)
+- SP creator placements are part of off-Amazon reach, with the CPC set from the maximum bid, expected performance and sales, and clicks drawing on the same campaign budget. (P16)
+- Reporting splits creator content as "Off Amazon" and creator storefronts as "Other – On Amazon"; "Limit reach to Amazon" keeps ads on Amazon properties, creator storefronts included. (P16)
+- A creator tag on the account deny list excludes that creator from all active SP and Display campaigns, and creators earn affiliate commission besides. (P16)
+- PPC Land and Ad Badger report the start of creator placements as 2026-08-10 with campaigns enrolled by default, and Ad Badger notes creator clicks can lower on-Amazon impression share while spend holds. (P16)
+
+## Ad content policy
+
+- On Amazon O&O supply, ads may not use all capitals (outside the marketplace exception that Amazon's *Prohibited content and products for Amazon O&O supply*, https://advertising.amazon.com/help/GTEEH3E3J2BDELBA, names), scare tactics, unsubstantiated warranty claims, calls to action in custom images, deal teasers, social links, contact details or disparagement of competitors. (P6)
+- On that supply, seasonal or holiday promotions start, in the policy's words, "no earlier than 8 weeks leading up to the event". (P6)
+- Claims must be accurate and substantiated. (P6)
+- Tobacco and nicotine, weapons (some knives excepted), THC or CBD, political ads, religious services, fundraising or proselytizing (religious books and merchandise excepted) and adult sexual products are prohibited across all supply. (P6)
+- Consumer healthcare products are tiered by sensitivity; tiers 2 and 3 are prohibited on Fire TV, Fire tablet, Alexa and the store homepage and barred from audience building. (P6)
+- Ad text must be in a language the marketplace's shoppers understand, and Amazon translates SB and display text automatically within 24 hours in the marketplaces Amazon's help page Sponsored ads translation lists (https://advertising.amazon.com/help/GJ67QHHYCKMEJGG9). (P7)
+- Creative may not reference Amazon products or services or use Amazon branded elements. (P8)
+
+## Moderation
+
+- Moderation combines automated and human review, usually within 24 hours and up to 3 business days. (P9)
+- Since February 2026 Amazon emails only rejections; an appeal goes through the rejection link, and repeat violations can lead to suspension. (P9)
+- Pre-moderation checks one entity's components before submission. (P9)
+- The moderation results read reports SP, SB and display moderation status (`APPROVED`, `IN_PROGRESS`, `REJECTED`, `FAILED`) with reasons and an expected time. (P9)
+- The legacy SB moderation endpoint shut off on 2025-12-15, and the asset library runs no eligibility check on images. (P9)
+
+## Billing
+
+- Sponsored ads are billed at most once per click. (P13)
+- Sellers pay by card, seller account funds or invoice, vendors by card, deduction from payments or invoice, and KDP authors by card or invoice; a payment failure pauses campaigns. (P13)
+- Invoices come each time accumulated cost reaches the credit limit and at least monthly, on the second day of the next month. (P13)
+- An invoice differs from campaign cost by invalid-traffic adjustments, over-delivery and promotional credits, taxes, regulatory advertising fees and charges carried across a month boundary. (P13)
+- Some countries add a regulatory advertising fee; Amazon stopped charging the Canadian one after 2025-07-31 and began refunding it on 2026-04-15. (P13)
+- Amazon's vendor payment methods are card (charged at each credit limit), deduction from retail proceeds by vendor code (which fails if that code's available funds are short) and monthly invoice due in 30 days. (P20)
+- Vendor invoicing is invite-only; deduction from payments is not available to third parties managing payments or to accounts without retail activity in the last 6 months; a failed payment pauses campaigns. (P20)
+
+## API terms
+
+- The Amazon Advertising API License Agreement limits confidential information to the advertising participant it relates to and to use on behalf of the participant who authorised it, and forbids generating repetitive or invalid interactions. (P19)
+- The Partner Network Policies forbid combining Amazon data with other third-party sources without Amazon's written approval and require deletion when it is no longer needed. (P19)
+- The Ads API is offered only to companies that do not operate a significant e-commerce business, a demand-side ad platform or an ad network. (P19)
+- For SP-API data Amazon requires incident notice within 24 hours, deletion within 30 days of the triggering event, and stored data kept separate or tagged by origin. (P19)
+- The SP-API Acceptable Use Policy forbids aggregating data across authorised users' businesses to provide to others, and requires being explicit about calculations and the use of AI models, their accuracy and data freshness, with data integrity and validation checks on AI analysis that materially affects a business. (P19)

package/skills/ads-review/SKILL.md

@@ -0,0 +1,38 @@
+---
+name: ads-review
+description: "How a seller's Sponsored Products, Sponsored Brands and Sponsored Display ads serve and perform, and why: the Featured Offer and stock behind an ad, serving statuses and money stops, what each metric counts and when it settles, which report or stream carries each figure, and a vendor's supply. For reviewing, watching or diagnosing a seller's sponsored ads."
+---
+
+## Serving
+
+A Sponsored Products or display ad serves only while its product holds the Featured Offer and is in stock, judged when the shopper views it; an out-of-stock product is paused within its campaign, and one without the Featured Offer gets zero impressions even when the campaign runs (A8).
+SP v3 with extended data returns every entity's serving status, so an ad can read `AD_NOT_IN_BUYBOX` or `AD_OUT_OF_STOCK` while its campaign runs; statuses and v1 `deliveryReasons` also name each money stop (A8, A15, A16, A17).
+`getItemOffers` names the current Buy Box holder (`IsBuyBoxWinner`) (A12).
+When another seller holds the Featured Offer on a brand's own ASIN, its Sponsored Brands ads keep serving, sending shoppers to a page whose default offer is the other seller's; Amazon routes the case to Seller Support (A9).
+
+## Diagnosis
+
+On a cost-per-click row spend equals impressions × CTR × CPC (A7).
+
+## Metrics
+
+Sponsored ads report a conversion on the date of the ad interaction, not necessarily the conversion date (M1).
+Sponsored Products credits 7 days of clicks for sellers, to the advertised product and anything else the merchant sells, and 14 for vendors; v3 carries SP 1-, 7-, 14- and 30-day purchases and sales with same-SKU and other-SKU variants (M3, M8).
+Impressions and clicks change for up to three days; conversions are first reported within 24 hours and restated up to 42 days for a 14-day window, and no report marks a day final (M14).
+
+## Reports
+
+Reporting v3, Sponsored Ads v2 reporting and Marketing Stream v1 stop delivering data on 2027-06-30; Reporting v1 and Stream v2 have been generally available since 2026-09-30 (R1).
+Reports hold only entities with activity in the range (R14).
+Hourly data beyond Reporting v1's span comes only by Marketing Stream, to the subscriber's own AWS destination, which ComChatX does not run (R12, R15).
+Asking for and landing a report: the amazon-ads skill's references/door.md.
+
+## References
+
+[serving.md](references/serving.md): Featured Offer, 2026 eligibility change, statuses, money stops.
+[reads.md](references/reads.md): hosts, profiles, accounts, offer price, Buy Box history.
+[metrics.md](references/metrics.md): attribution, definitions, invalid traffic, settling.
+[reports.md](references/reports.md): v3, v1, hourly data, limits, Stream, console.
+[decomposition.md](references/decomposition.md): metric chain, cause orders, benchmarks, shift or noise.
+[vendors.md](references/vendors.md): vendor profiles, purchase orders, supply.
+[history.md](../ads-change/references/history.md): change history.