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

package/skills/ads-seasons/references/deals.md

@@ -0,0 +1,42 @@
+# Deals, coupons and promotions
+
+## Deal types
+
+A Lightning Deal runs 4–12 hours, as Amazon determines, and a Best Deal several days. (E9)
+Both need a Professional plan and can carry a Deals badge. (E9)
+For deals, "Seller promotions" names a four-star seller rating and the Sell on Amazon guide an overall rating of at least four stars. (E9)
+*Amazon deals* requires a sales history on Amazon, a product rating of at least four stars and a seller rating of at least 3.5, and makes deals invite-only. (E9)
+Coupons need a 3.5-star rating, hold up to 200 ASINs, expire after 90 days and show a coupon badge with the discount in shopping results. (E9)
+Price discounts run up to 30 days on limited units. (E9)
+Amazon's Prime Day coupon criteria are a discount above 5%, dates coinciding with Prime Day and the audience set to Prime members. (E9)
+
+## Deadlines and fees
+
+Deal submission windows differ by marketplace and year, and deal fees by marketplace; for holiday 2026 Amazon stated them in News_Amazon's Seller Forums posts "Holiday 2026: Same fees, same eligibility, earlier deadlines" for the US (https://sellercentral.amazon.com/seller-forums/discussions/t/3e31fbb7-04e0-4ed4-873e-f74b1052e2ff) and "Peak season 2026: Inventory dates, peak fees, deal cutoffs and promotions" for the UK and EU (https://sellercentral.amazon.co.uk/seller-forums/discussions/t/d8c34b0b-86ed-4aec-9b0a-8be6eabb8fb0). (E10)
+In those posts Best Deals, Lightning Deals and Prime Exclusive Price Discounts cost an upfront fee per promotion plus a capped share of promotional sales, and the UK and EU post discounts the upfront fee for early submission. (E10)
+A promotion's own fee rates are in the Promotions API's fee snapshot. (E10, E14)
+Prime Exclusive Discounts in the UK and EU need a minimum star rating, which News_Amazon's "Peak season 2026: Inventory dates, peak fees, deal cutoffs and promotions" (https://sellercentral.amazon.co.uk/seller-forums/discussions/t/d8c34b0b-86ed-4aec-9b0a-8be6eabb8fb0) states. (E10)
+PBDD 2026 promotional prices are excluded from the 30- and 60-day lookbacks that set the maximum BFCM deal price. (E11)
+Amazon's Prime Day 2026 staff post for sellers, "Prime Day 2026: Your Deals Deadline Extended!" (https://sellercentral.amazon.com/seller-forums/discussions/t/9412e643-a006-464d-824a-0121d07c1b63), extended the Prime-Exclusive Best Deal and Lightning Deal deadline. (E26)
+The same post required inventory in the network or in transit, and stated that deals are suppressed if ASINs are not buyable at event time. (E26)
+
+## Reading promotions
+
+The SP-API Promotions API (2025-12-01), for sellers, lists coupons, deals, price discounts and basket-building promotions with status, schedule (`startDate`, `endDate` and the event's `eventId`) and validation issues. (E14)
+It also carries Prime-exclusive and brand-tailored customer segments, badge display, a stacking flag and a fee snapshot. (E14)
+It is read-only, has no Brand Tailored Promotions type, and no operation read creates or reports Brand Tailored Promotions or Customer Engagement. (E14)
+
+## Brand Tailored Promotions
+
+Brand Tailored Promotions are for Brand Representatives in Brand Registry, in the stores sell.amazon.com's "What are Amazon Brand Tailored Promotions?" lists (https://sell.amazon.com/blog/brand-tailored-promotions). (E21)
+They give 10–50% off to an audience of at least 1,000 shoppers chosen by goal (acquisition, re-engagement, retention, cross-sell), each eligible customer redeeming once. (E21)
+One starts at least 48 hours after creation and within 60 days, and ends within 90 days of its start; a brand holds up to 20 at once. (E21)
+Once one is live, only its budget can change, and only upward. (E21)
+It shows a badge only to the audience and stacks with coupons, and its fees are stated only on a Seller Central help page. (E21)
+Practitioners disagree on the best audience: Desert Cactus reports recent customers, Intentwise says many agencies favour cart abandoners, and none publishes a holdout test. (E21)
+
+## Customer Engagement
+
+The Customer Engagement tool sends free marketing emails, delivered by Amazon, to a Brand Registry brand's followers and, since 2022, its repeat, recent and high-spend customers. (E22)
+It reports open rate, click-through, delivered emails, opt-outs, sales and conversion. (E22)
+Amazon says brands with 1,000 or more followers "will have the best experience". (E22)

package/skills/ads-seasons/references/measurement.md

@@ -0,0 +1,26 @@
+# Reading an event
+
+## Attribution
+
+Sponsored Products attributes seller conversions over 7 days and vendor conversions over 14. (E8)
+Sponsored Brands and Sponsored Display attribute over 14 days, click or view. (E8)
+Conversions are dated to the ad interaction and restated 1, 7 and 28 days after the conversion. (E8)
+Amazon advises checking 7–14-day attribution before changing campaigns after a peak, because peak-event sales often occur days after the click. (E8)
+Marketplace Ad Pros found the clicks of Prime Day 2026's last day fully mature about two weeks later. (E8)
+Attribution and metrics at large: ../ads-review/references/metrics.md.
+Report types and columns: ../ads-review/references/reports.md.
+
+## Benchmarks across a moving event
+
+Data Kiosk's week-level `yearOverYear` benchmark compares the same ISO week number a year earlier, and its month-level benchmark the same calendar month. (E15)
+Data Kiosk has no benchmark at day level, so a moving event such as Prime Day lands in different benchmark weeks. (E15)
+Search Query Performance runs on Sunday–Saturday weeks, calendar months or quarters, and a request cannot span periods. (E15)
+
+## Promotions over time
+
+Store-level scanner research by Van Heerde, Leeflang and Wittink (2000) puts the post-promotion dip at 4–25% of the promotion's current sales effect, usually missed by simple models and visual inspection. (E18)
+Panel research by Mela, Gupta and Lehmann (1997) found consumers growing more price- and promotion-sensitive as promotions increase and advertising falls. (E18)
+Amazon's own guide says regular promotions "can condition customers to wait for deals". (E18)
+A review by Kaul and Wittink (1995) found price advertising raising price sensitivity and non-price advertising lowering it. (E18)
+Across 70 French brands, Ataman, van Heerde and Mela (2010) found the total sales elasticity of discounting at .04 and of advertising at .13, and discounting's long-term effect one third of its short-term effect. (E18)
+In Erdem, Keane and Sun (2007), advertising raised marginal consumers' willingness to pay for 17 of 18 brands in four categories. (E18)

package/skills/ads-seasons/references/spend.md

@@ -0,0 +1,44 @@
+# Spend and performance around events
+
+## Amazon's holiday guidance
+
+For the 2025–2026 holiday Amazon advises launching Sponsored Products campaigns by early October. (E5)
+It advises raising daily budgets 20–30% around Black Friday and Cyber Monday, starting three to five days ahead of each peak, as guidance figures, not limits. (E5)
+Amazon describes the season in three phases: build (October), peak (November through Cyber Monday) and extend (December and beyond). (E5)
+It defines the extend phase by last-minute gift shoppers, post-holiday deal seekers, visitors who browsed but did not buy during peak events, and seasonal buyers to turn into repeat customers; its New Year guide adds January resolution shoppers. (E5)
+No Amazon page read on 2026-11-09 uses the term "Q5" or describes CPC after December 25. (E5)
+In Amazon's post-event teaching scenario, "Extend holiday deals into January and maintain Sponsored Brands investment" is the answer marked incorrect. (E5)
+The answer marked correct moves budget to high-performing branded terms and repositions seasonal stock with non-holiday messaging. (E5)
+
+## Event CPC and budgets
+
+Marketplace Ad Pros measured Prime Day 2026 median Sponsored Products CPC up 10% ($1.19 to $1.31) and Sponsored Brands CPC up 2% against matched pre-event days, across several hundred US advertisers. (E6)
+In Marketplace Ad Pros' data, 51% of advertisers cut SP budgets and 13.7% raised them by 20% or more, and raisers and holders showed the same 1.30x ad-sales lift, which the authors call correlational. (E6)
+Skai measured Prime Day 2025 CPC at $1.80, down 10.4% from 2024, in a same-store cohort. (E6)
+Ad Badger recommends 1.5–2.5× budgets in the two days before a Prime event and 2–3× on event days. (E6)
+Against Ad Badger's figures, Amazon's holiday guide and Florian Nottorf (Adference) give +20–30% starting 3–5 days ahead. (E6)
+
+## Conversion during an event
+
+Amazon's holiday "myths" page says conversion rates often drop during major sales events while shoppers browse. (E7)
+The same page says a temporary ACOS spike is read against TACOS, which "reveals whether your total sales (organic and paid) are growing in relation to your ad spend". (E7)
+Ad Badger reports impressions rising while clicks and conversions stay flat in the 48 hours before a Prime event. (E7)
+ACOS, TACOS and the other metrics, defined: ../ads-review/references/metrics.md.
+
+## After the event
+
+Practitioners disagree on post-event spend. (E16)
+Pare holds ad velocity for 48–72 hours and retargets detail-page viewers at 20–30 days. (E16)
+Ad Badger returns bids to normal the next day and runs Sponsored Display retargeting for days 1–7. (E16)
+Acadia finds "consistently thin post-event lift" in its going-dark tests and trims lead-out media. (E16)
+Acadia reports brands running no deals lifting 46% over their 30-day baseline at Prime Day 2025 and 116% at Prime Day 2024, and calls non-participation "a penalty" in 2026. (E16)
+Acadia treats year-over-year growth rather than period-over-period lift as comparable across events, because 2025's baseline week was unusually low. (E16)
+Amazon's 137% (Prime Day 2025) and 145% (holiday 2025) "increase in sales" figures compare advertisers with average category growth over tentpole windows against the adjacent preceding weeks. (E16)
+
+## December 26 to mid-January
+
+Practitioners call December 26 to mid-January "Q5" and disagree on it. (E20)
+Evolve AMZ says CPMs "commonly fall 28% or more against the Q4 peak" as seasonal advertisers exit and gift-card redemption concentrates there, and Pattern says CPCs go down. (E20)
+Ad Badger's app data puts January 2026 at the year's highest ACOS (32.5%, against 28% in October 2025). (E20)
+Intentwise's data has conversion falling from 13.4% in the week ending December 23 to 11.03% the week after. (E20)
+Teikametrics reported, across more than 700 sellers in 2019, ad conversion rising and cost per conversion falling from Cyber Monday to the holiday shipping cutoff before Christmas, which it attributed to brands having spent their seasonal budgets. (E20)

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

@@ -0,0 +1,26 @@
+# Vendor promotions
+
+## Reports
+
+Amazon's promotion performance report covers vendor promotions, with types including Best Deal, Deal of the Day, Lightning Deal, Price Discount, Sales Discount, Coupon and Promo Code. (E24)
+Its statuses include `NEEDS_YOUR_ATTENTION`, and its creation channels are self-service, full-service (account managers), automated (for example Deal-O-Matic) and manual (Vendisto). (E24)
+It carries the vendor-only funding agreement id and the funding amount spent, times in UTC, and two years of start dates. (E24)
+The vendor coupon report gives clips, gross redemptions, total discount and the vendor's budget spent, including clip and redemption fees, per coupon and campaign, in UTC. (E24)
+The SP-API Promotions API (2025-12-01) is available to sellers; vendor promotions are reported through the promotion performance and coupon reports. (E24)
+
+## Amazon's 2017 vendor guides
+
+Amazon's March 2017 vendor self-service promotions guide listed Price Discount up to 12 months, submitted at least 24 hours ahead. (E25)
+It listed Best Deal up to 2 weeks, at least 10% off, submitted 4 or more weeks ahead. (E25)
+It listed Lightning Deal up to 6 hours, at least 15% off, with a maximum unit quantity, submitted 4 or more weeks ahead. (E25)
+It listed Promo Code up to 30 days, at least 15% off, submitted at least 24 hours ahead. (E25)
+The guide recommended 4 weeks for all, "or Amazon may not have time to buy inventory", and said minimum discounts may rise for Prime Day, Black Friday and Cyber Monday. (E25)
+It stated that Amazon often buys more product to prepare for an approved promotion, sets prices independently, generally aims to pass promotional funding on to customers, and may not pass it on if the product does not meet financial thresholds. (E25)
+Amazon's August 2017 AMS toolkit stated that "Product Display Ads are the only ad unit that supports VPCs" (vendor-powered coupons) and that Subscribe & Save VPCs were not eligible. (E25)
+The 2017 toolkit stated that a Product Display ad whose VPC is no longer valid reverts to the normal format. (E25)
+
+## Funding
+
+Amazon's purchase-order model carries a `dealCode`, a promotional or deal number used to obtain a price discount on items on the order. (E27)
+AMALYTIX (Trutz Fries, October 2026) states that vendors fund Lightning Deal discounts per unit, in local currency and excluding tax, and names no separate vendor deal fee. (E27)
+Wake Commerce (James Wakefield, 2024) describes co-op, also called ContraCoGS, as deducted through PO invoices. (E27)

package/skills/ads-structure-and-targets/SKILL.md

@@ -0,0 +1,38 @@
+---
+name: ads-structure-and-targets
+description: "How a seller's sponsored ads are built and what they target: Sponsored Products, Sponsored Brands and Sponsored Display campaigns, ad groups, portfolios, fields fixed at creation, archiving, limits, global campaigns, match types, auto groups, product, category, theme and audience targets, and brand and competitor targeting with what research and practitioners report on it. For reading, auditing or changing a seller's campaign structure and targets on Amazon."
+---
+
+## Structure
+
+- Sponsored Products nests campaign, ad group, then product ads and targets; an ad group's products share its bids, and no ad group holds both keyword and product targets. (T1)
+- Budget and bidding strategy live on the SP campaign; an ad group has no budget. (T3)
+- Fixed at creation: SP auto or manual (`targetingType`), a keyword's match type, a Sponsored Brands campaign's goal and other fields listed per product; changing one means archiving and recreating, though a keyword in another match type can be added beside the old one; while both run, the higher-bid one is used on a query both match, and Amazon's pages do not say whether across ad groups or campaigns. (T4, T22)
+- `ARCHIVED` is terminal, reached through a delete. (T6)
+
+## Sponsored Brands
+
+- A Sponsored Brands ad group holds one ad type and targets keywords or products; v4 ad groups carry no bid. (T2, T20)
+
+## Matching
+
+- SP auto targets through close match, loose match, substitutes and complements. (T5)
+- No match type is a literal string match. (T23)
+- An SP product target holds one ASIN, exact or expanded. (T27)
+- Which keyword or target matched a search term, its match type, ad group and campaign: ../ads-search-terms/references/search-term-report.md.
+
+## Brand and competitors
+
+- Any advertiser may target any brand's name or ASIN (Amazon's Protect and Conquest). (T35)
+- None of Amazon's share metrics names a competitor; Brand Analytics Top Search Terms names a term's three most-clicked products and needs Brand Registry. (T32)
+- No Amazon field flags a search term as branded. (T42)
+- Field experiments on search engines, not Amazon, found a brand's own ad on its query added little when no competitor advertised and more when competitors did. (T43)
+
+## References
+
+- references/structure.md: hierarchy, fixed fields, archiving, portfolios, limits, quota errors, global campaigns, vendors.
+- references/sponsored-brands-and-display.md: SB goals, formats, v4, retirement; SD targets.
+- references/targets.md: match types, auto, product and category targets, themes, recommendations.
+- references/brand-and-competitors.md: brand and conquest targeting, share, Top Search Terms, experiments, law.
+- The amazon-ads skill's references/door.md: Ads API paths and reports. The amazon-sp-api skill's references/door.md: SP-API reports, Brand Analytics among them.
+- ../ads-search-terms/SKILL.md: search terms. ../ads-bids-and-budgets/SKILL.md: bids, budgets. ../ads-review/references/metrics.md: metrics.

package/skills/ads-structure-and-targets/references/brand-and-competitors.md

@@ -0,0 +1,60 @@
+# Brand and competitor targeting
+
+Metric definitions are in ../ads-review/references/metrics.md.
+
+## What Amazon allows
+
+- Any advertiser may target any brand's name as a keyword and any ASIN as a product target, which Amazon calls Protect and Conquest. (T35)
+- Amazon suggests bidding on one's own branded keywords and own product pages. (T35)
+- Ad creative may not name or disparage a competitor, and may say "a leading brand". (T35)
+
+## Reserve share of voice
+
+- Reserve share of voice sells Sponsored Brands top-of-search placement on the advertiser's own exact branded keywords at a fixed upfront price, through v1 operations. (T36)
+- Its keywords, flight and budget are fixed once it starts. (T36)
+- Amazon's own pages state different flight lengths, minimums and marketplaces. (T36)
+- Amazon reports its beta raised top-of-search share from 62.7% to 99.3%, a before/after comparison. (T36)
+
+## Share and branded-search data
+
+- Search term impression share is the account's share of one ad product's impressions on a term, account-wide, not splittable by campaign, and in windows that differ by surface. (T31)
+- Top-of-search impression share divides a campaign's top-of-search impressions by those it was eligible for. (T31)
+- v3's columns page lists top-of-search impression share for `spCampaigns`, `sbCampaigns` and `sbTargeting` and the report type pages add `spTargeting`, while v1 offers it with campaign-level dimensions only. (T31)
+- Sponsored Brands v2's search-term share fields stop on 2026-11-10. (T31)
+- No API was found for the console's brand impression share dashboard. (T31)
+- None of Amazon's share metrics names a competitor. (T32)
+- Brand Analytics Top Search Terms names the three most-clicked products per term and needs Brand Registry. (T32)
+- Branded-search metrics (searches for the brand after an ad interaction) exist for Sponsored Brands, display, TV and DSP but not Sponsored Products. (T42)
+- No Amazon field flags a search term as branded; keyword recommendations carry a `keywordTheme` label. (T42)
+
+## Field experiments on brand queries
+
+- The experiments below ran on search engines, not Amazon's store. (T43)
+- On a brand's own query, the brand's own ad added little when no competitor advertised: on Bing about 1–4% of clicks, and on eBay no measurable short-term benefit, with organic clicks capturing 99.5% of the forgone paid clicks. (T43)
+- It added more when competitors advertised: brands not advertising lost 18–42% of clicks. (T43)
+- Competitors' ads took 1–2% of a brand query's traffic on Bing. (T43)
+- At Edmunds.com, shutting off paid links lost more than half the paid traffic. (T43)
+- A firm pausing Google search ads in random US markets lost customers, but only 63% as many as a non-experimental estimate implied. (T43)
+
+## Models and law
+
+- Game-theory models find mutual bidding on each other's brand names can be a prisoner's dilemma that profits the platform, and competing bids can cycle (Edgeworth cycles). (T44)
+- When Google opened trademarks to rivals' keyword bidding in continental Europe in 2010, navigational searches became less likely to reach the trademark owner, more so for lesser-known marks. (T45)
+- The Ninth Circuit held in 2015 that Amazon showing other brands' watches for a search of "mtm special ops" was not likely to confuse shoppers, because Amazon labels each result with its name, manufacturer and photo. (T45)
+- The Second Circuit in 2021 vacated the FTC's order against 1-800 Contacts' settlement agreements requiring rivals to refrain from search advertising on its trademarks and to use them as negative keywords. (T45)
+
+## Practitioners and Amazon figures
+
+- Perpetua's rule of thumb is that branded spend above 40% of ad spend likely buys traffic the brand would get anyway. (T46)
+- Haus reports, from Google branded-search tests, significant lift for 82% of brands facing high competition on their brand terms against 35% facing low competition. (T46)
+- Amazon's 2023 goal-based campaigns page states that, from Amazon data of 6,000 advertisers (June 2023), on average 60% of shoppers who searched for a brand bought from a competitor brand within 30 days. (T48)
+- Practitioner figures on Sponsored Brands video against competitors (Skale Strategy, ATTN Agency, BTR Media) are published without a stated method. (T48)
+- No Amazon page read states whether Store pages carry other brands' ads, which Perpetua's 2021 guide says they cannot. (T48)
+
+## Competitor stock-outs
+
+- Perpetua's Competitor Signals raises bids when a competitor goes out of stock and normalises them on restock. (T47)
+- Jarvio and Netpeak advise product-targeting a rival's page during its stock-outs. (T47)
+- Eva calls a rival's stock-out a window of "typically 2 to 6 weeks" and warns that conquest-driven sales can stock out the conquering ASIN. (T47)
+- None of these four publishes a measured effect. (T47)
+- Stock-out studies outside advertising find that a model accounting for availability predicts significantly larger stock-out effects on profitability than one ignoring it. (T47)

package/skills/ads-structure-and-targets/references/sponsored-brands-and-display.md

@@ -0,0 +1,44 @@
+# Sponsored Brands and Sponsored Display
+
+## Sponsored Brands campaigns
+
+- A Sponsored Brands ad group holds one type of ad and targets either keywords or products. (T2)
+- Automatic collections take keyword targets only. (T2)
+- A Sponsored Brands v4 campaign holds several ad groups with several ads each, and its ad groups carry no bid. (T20)
+- Moderation rejects an ad, not the campaign. (T20)
+- Legacy v3 campaigns hold read-only ad groups, and a migrated campaign gets a new id, with the old one kept as `legacyCampaignId`. (T20)
+
+## Goals
+
+- Sponsored Brands goal-based campaigns have two goals: drive page visits, on CPC and the default when neither goal nor cost type is sent, and grow brand impression share, on vCPM and measured by top-of-search impression share. (T13)
+- Earlier Sponsored Brands campaigns are tagged "Drive page visits". (T13)
+- The marketplaces where goal-based campaigns are not supported are named in Amazon's *Sponsored Brands v4* model, https://d3a0d0y2hgofx6.cloudfront.net/openapi/en-us/sponsored-brands/4-0/openapi.json. (T13)
+
+## Formats
+
+- Sponsored Brands formats are manual and automatic collections (3–10 products, the type fixed once the ad group exists), store spotlight and video. (T38)
+- Under drive page visits every format can serve on search, detail pages and the home page. (T38)
+- Sponsored Brands video has its own placement and auction, and video creative follows chapters 1 to 5 of Amazon's moderation guide. (T38)
+- The ASIN list of a Store page returns only ASINs placed directly on that page. (T38)
+
+## Product Collection retirement
+
+- Sponsored Brands Product Collection ads are being retired: since August 2026 a create makes a Manual Collection ad. (T19)
+- Amazon's pages give January 2027 for creates to error and February 2027 for the end of support. (T19)
+- The Sponsored Brands v4 migration list returns recommended replacements. (T19)
+
+## Sponsored Brands model files
+
+- Amazon publishes Sponsored Brands targeting across three model files. (T39)
+- The SB 3.0 YAML holds the keyword, negative keyword, target, negative target and theme operations. (T39)
+- The JSON file titled "Sponsored Brands 3.0" and the SB 4.0 JSON hold the category target, product target count and negative brand target recommendation operations, and neither of those two carries the negative keyword operations. (T39)
+
+## Sponsored Display targets
+
+- Sponsored Display contextual targets with refinements, and single products, serve on Amazon only. (T37)
+- Unrefined categories and "similar to advertised products" also serve off Amazon; the console adds "similar to advertised products" on its own, and the API does not. (T37)
+- Audience campaigns exclude shoppers who bought the advertised or similar products in the past 30 days. (T37)
+- View audiences reach back up to 90 days and purchase audiences up to 365, including viewers or buyers of a given brand. (T37)
+- Audience ids come from the audience list. (T37)
+- On the unified console, contextual targeting is part of multi-signal targeting. (T37)
+- Randomised experiments outside Amazon found retargeting incremental for the advertisers studied, and showing the exact product a shopper viewed was not always the better ad. (T41)

package/skills/ads-structure-and-targets/references/structure.md

@@ -0,0 +1,75 @@
+# Campaign structure
+
+## The Sponsored Products hierarchy
+
+- In Sponsored Products the hierarchy is campaign, ad group, then product ads and targets, with negatives at campaign or ad-group level. (T1)
+- Products in one ad group share its bids and targeting, ad group names are unique within a campaign, and one ad group cannot hold both keyword and product targets. (T1)
+- The budget, the bidding strategy with its placement adjustments, the dates, the portfolio, the site restriction, the off-Amazon setting and the tags live on the SP campaign, and an ad group has no budget. (T3)
+- Amazon reports that ad groups of similar products sold 6% more units on average. (T14)
+- Amazon marks a campaign it manages with `autoManageCampaign`. (T16)
+- Ads API v1 models campaigns, ad groups, ads, targets and ad associations across SP, SB, SD, Sponsored TV and DSP in one schema. (T10)
+
+## Fixed at creation
+
+- Fixed at creation: whether an SP campaign is auto or manual (`targetingType`), a keyword's match type, a Sponsored Brands campaign's goal, and several other fields Amazon lists per ad product. (T4)
+- Changing one of them means archiving and recreating; a keyword in another match type can be added beside the old one. (T4)
+- While both run, the higher-bid match type of the two is used on a query both match; Amazon's pages do not say whether this reaches across ad groups or campaigns. (T4, T22)
+
+## Archiving
+
+- `ARCHIVED` is terminal: Amazon's model calls it "permanently stopped and cannot be reactivated". (T6)
+- Every API reaches `ARCHIVED` through a delete. (T6)
+- An archived parent refuses updates and creations of its children, and a bulksheet archive cascades. (T6)
+
+## Portfolios
+
+- A campaign belongs to at most one portfolio, and portfolios do not compete with each other. (T7)
+- Portfolios are managed per profile, in that marketplace's currency. (T7)
+- Amazon suggests organising portfolios by brand, category or season. (T7)
+
+## Limits
+
+- Entity limits are published on Amazon's *Sponsored ads campaign limits* page, which says Amazon raises limits as an account nears them. (T8)
+- A write past a quota is refused with `QUOTA_EXCEEDED` or `NON_ARCHIVED_QUOTA_EXCEEDED`. (T8)
+- Third-party limit tables disagree with each other and with Amazon. (T8)
+- Campaign names are limited to 128 characters for sellers and 116 for vendors, ad group names to 255, and an entity takes up to 50 tags. (T9)
+
+## Global campaigns and other marketplaces
+
+- A global SP campaign (`marketplaceScope` `GLOBAL`) holds one budget per marketplace in local currency. (T11)
+- It carries a top-level delivery status with a per-marketplace one, and its targets carry a `marketplaces` list, where a child's marketplaces are a subset of its parent's. (T11)
+- Only the `spCampaigns` report surfaces global campaigns. (T11)
+- The Localization API maps keywords, products, targeting expressions and currencies between marketplaces. (T12)
+- Amazon's keyword translation combines machine translation, linguists and closest local terms. (T12)
+- Valid scripts and symbols differ by marketplace, and the API accepts hiragana, katakana, kanji and both widths in keywords. (T12)
+- Amazon publishes a checklist for new international ad campaigns. (T12)
+
+## Bulksheets and exports
+
+- A bulksheet row carries a product, an entity and an operation, over the same entities as the API. (T15)
+- An export from the Exports API is a snapshot of current metadata with no history, with five jobs per endpoint and a download link valid for one hour. (T15)
+
+## Vendors and sellers
+
+- Vendors identify a Sponsored Products product ad by ASIN and sellers by SKU; in Sponsored Display vendors use the `asin` field where sellers use `sku`. (T49)
+- The v3 advertised-SKU column is not available for vendors, and Review Requests take an ASIN from both. (T49)
+- Only vendors may give a Sponsored Brands campaign a custom landing page URL, which must include the ASINs of at least three products advertised in the campaign and cannot be combined with an `asins` list; the merged model's `CUSTOM_URL` landing-page type is "Available for vendors only". (T49)
+- Amazon's *Sponsored Products 3.0* model, https://d1y2lf8k3vrkfu.cloudfront.net/openapi/en-us/dest/SponsoredProducts_prod_3p.json, defines custom text ads only for KDP authors and book vendors, in the marketplaces it names, and its brand negative-target recommendations are "Only available for Sellers and Vendors". (T49)
+
+## Which products are advertised
+
+- Amazon labels products "suggested" on the create page and sorts ASINs by life cycle and sales in its ASIN matrix guide. (T21)
+- Amazon advises distinct complementary products, not variations, in a collection. (T21)
+- m19 advises advertising "only the top-performing children" of a family, and Canopy's Chuck Kessler gives "the 20% of products driving 80% of revenue" a defensive budget. (T21)
+
+## Practitioner structures
+
+- Perpetua isolates search terms per ASIN across auto, manual and exact campaigns. (T17)
+- Helium 10 names four SP campaigns. (T17)
+- Seller Labs lists signs of over- and under-structure. (T17)
+- Single-keyword campaigns have advocates (Scale Insights, Quartile, JumpFly) and critics. (T17)
+- Mina Elias keeps one ad group per campaign against "keyword bullying". (T17)
+- Ad Badger segments by targeting type. (T17)
+- Ad Badger calls a branded/non-branded campaign split "more of a reporting strategy than a performance strategy". (T18)
+- Practitioners' naming conventions put ad product, product, targeting and goal in the name. (T17)
+- AMZ Pathfinder names scattered ASINs the most visible audit problem. (T17)

package/skills/ads-structure-and-targets/references/targets.md

@@ -0,0 +1,68 @@
+# Keywords and targets
+
+Search terms and negatives are ../ads-search-terms/SKILL.md's; the search term report is in ../ads-review/references/reports.md.
+
+## Match types
+
+- Broad, phrase and exact all account for plurals, misspellings and translations. (T23)
+- Broad matches the terms in any order, with synonyms and related terms. (T23)
+- Amazon's own exact example matches "outside lights" to "outdoor lights", its guides disagree on whether exact tolerates reordering, and no match type is a literal string match. (T23)
+- Where no search term exists, Amazon infers one from context and serves keywords there, and the report shows such inferred terms unmarked beside typed ones. (T24)
+- Sponsored Brands keywords add semantic matching on the same meaning or intent, even in phrase match, and accept a `+` modifier in broad. (T25)
+- Sponsored Brands negatives take exact and phrase only, with limits of 10 words per keyword and 4 per negative phrase. (T25)
+
+## Compounds, scripts and spacing
+
+- Amazon's Japanese guides show phrase match finding a keyword inside unspaced text. (T26)
+- Amazon's German guide shows phrase „Laufschuhe" serving „Laufschuhe für Damen" but not „Schuhe zum Laufen", which Sponsored Brands semantic matching can reach. (T26)
+- No Amazon page read states whether SP broad or exact bridges a compound and its split or hyphenated form, which AMALYZE lists as a close variant. (T26)
+- Amazon researchers report that Amazon's Japanese query reformulation normalises script variants and tokenises unspaced queries, in a paper that does not describe advertiser match types or negatives. (T26)
+
+## One keyword, several entries
+
+- When one keyword is set in several match types with different bids, the match type with the highest bid is used, and Amazon's page says the advertiser is not bidding against itself. (T22)
+- Amazon's pages read do not say whether this reaches across ad groups or campaigns, to same-match-type copies, or to different keywords and targets of one advertiser matching one query. (T22)
+- Amazon's targeting guide says bids set by product priority give the top-priority product "the best opportunity to get top ad placements". (T22)
+- Practitioners disagree on how a seller's own campaigns meet on one query, and none cites an Amazon page or measured data. (T22)
+
+## Auto campaigns
+
+- An SP auto campaign targets through four groups: close match and loose match on search queries, substitutes and complements on detail pages. (T5)
+- The four groups can be paused or bid but not archived, and v1 names them `THEME` targets. (T5)
+- Amazon says auto reads the product's titles and descriptions; ClearAds says bullets and backend terms too. (T5)
+- Ad Badger says close match takes most of an unsegmented auto campaign's traffic. (T5)
+
+## Product and category targets
+
+- An SP product target holds one ASIN and is exact or expanded ("similar"); expanded is the console default and adds "a limited number of additional products with the highest performance potential" that Amazon chooses. (T27)
+- Amazon says a product target appears on the targeted product's detail page and on search result pages where that product ranks among the top results. (T27)
+- In one seller's 2020 placement data published by Perpetua, 85–92% of category, ASIN and brand targeting fell on detail pages. (T27)
+- Category targets refine by brand, price range and star rating in SP, SB and SD, and by Prime shipping, age range and genre in SP and SD. (T28)
+- Sponsored Brands has no expanded-ASIN target. (T28)
+
+## Amazon-built targets
+
+- Sponsored Brands theme targets are keywords related to the brand or to the landing pages, which Amazon generates "based on historical performance and brand interactions", and a goal-based campaign creates them by default. (T29)
+- SP keyword groups are Amazon-maintained keyword sets built from the advertised ASINs and refreshed over time; the marketplaces where they are available, and whether they are still a beta, are stated in Amazon's Ads API guide *Sponsored Product keyword groups targeting*, https://advertising.amazon.com/API/docs/en-us/guides/sponsored-products/keyword-group-targeting. (T29)
+- The v1 theme for peer brands' category searches picks the peer brands automatically, and the pages read do not say which ad product accepts it. (T29)
+
+## Recommendations and keyword sources
+
+- SP keyword recommendations rank by `CLICKS`, `CONVERSIONS` or `DEFAULT` and carry no search volume. (T30)
+- Amazon says keyword recommendations draw on past performance and similar products. (T30)
+- Product recommendations come in themes. (T30)
+- The v2 suggested-keyword endpoints were shut off in June 2026. (T30)
+- Third-party keyword volumes are estimates, and Helium 10 and Jungle Scout each published a test favouring itself. (T33)
+- Reverse-ASIN tools list the queries an ASIN ranks for. (T33)
+- Product Opportunity Explorer groups search terms into niches, and practitioners disagree on its access terms. (T33)
+- The Customer Feedback API returns an ASIN's top review topics. (T33)
+
+## Relevance
+
+- Keywords and product targets must be relevant to the advertised product, and Amazon lists restricted targets. (T34)
+- Amazon's auctions weigh the ad's expected relevance to the shopper. (T34)
+
+## Research on matching
+
+- Amazon's published research describes organic product search matching as behaviour-trained semantic matching, motivated by lexical matching's failures on synonyms, morphological variants and misspellings, and does not describe the ads matcher. (T40)
+- Amaldoss, Jerath and Sayedi model broad match as the platform bidding on the advertiser's behalf, helping advertisers when its accuracy is high and a prisoner's dilemma at moderate accuracy, in a theory of web search ads, not an Amazon measurement. (T40)