amazon-listing-optimizer
Skill8 starsChanged 45 days agoMITView on GitHub
Optimize an Amazon listing from your own live data, tuned for 2026 Amazon (the 75-character title cap, the COSMO intent layer and the Rufus / Alexa for Shopping assistant). It benchmarks your search funnel against the market to find whether you leak at impressions, clicks, cart-adds or purchases, separates an exposure problem from a copy problem, finds the converting themes your title forgot to say, audits intent coverage and listing health, and returns a rewritten, policy-safe, upload-ready title, bullets, backend terms and attribute fills. Live from DataDoe. Use for "optimize my listing", "listing optimization", "fix my title", "improve my bullets", "keyword gaps", "why isn't my listing ranking", "why isn't it converting", "backend keywords", "listing audit", or "is my listing Rufus ready".
1 other project uses a version of this.
Adopted by
Most recently taken up by
- SkillMedev/ecommerce-dtc-opstheir own version, first seen Oct 5, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- SkillMedev/ecommerce-dtc-ops★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository26 other files it gives its agents.
- amazon-asin-search-auditorSkill
- buy-box-loss-root-causeSkill
- create-amazon-reconciliation-dashboardSkill
- create-orders-managerSkill
- daily-account-health-checkSkill
- daily-performance-health-checkSkill
- fba-inventory-ledger-reportSkill
- fba-stock-restock-plannerSkill
- keyword-rank-sqp-trackerSkill
- net-profit-pl-analyzerSkill
- ppc-asin-acos-alertSkill
- ppc-bid-optimizer-applySkill
- ppc-campaign-acos-alertSkill
- ppc-negative-keyword-applierSkill
- ppc-profit-bid-applySkill
- ppc-wasted-spend-watchdogSkill
- promotions-managerSkill
- restock-priority-alertSkill
- return-refund-analyzerSkill
- sales-movers-scannerSkill
- suppressed-inactive-listings-checkSkill
- vendor-inventory-reportSkill
- vendor-performance-dashboardSkill
- vendor-pnl-dashboardSkill
- weekly-business-reviewSkill
- weekly-sales-briefingSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
13 rules. About 0 are common, 9 are distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈9
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 39
Before a push/relaunch, or after a title/bullet edit, to check coverage + funnel.
- Distinctive: line 70
impressions -> clicks -> cart-adds -> purchases: Read each stage against the market average for that same query, not against an absolute …
- Distinctive: line 105
Before rewriting anything, run this once - the most common, highest-ROI miss:
- Distinctive: line 110
Check whether that theme's words are in the title: If your best-converting theme is missing, that is the #1 fix - and since the title now …
- Very specific: line 200
Use child_asin_search_query_score.
- Distinctive: line 223
Read the marketplace's caps now (Copy rules below).
- Distinctive: line 234
Never present a one-week sample as an 8-week trend.
- Very specific: line 261
Check price vs featuredoffer_price and available if pulled.
- Distinctive: line 289
If the current title is over 75 chars, treat compliance as an urgent fix (Amazon will otherwise auto-rewrite it).
- Very specific: line 294
Backend search terms: read them from amazon_listings_raw.attributes (Amazon attribute generic_keyword).
- Distinctive: line 316
Compliance: - never invent claims (medical, "cures", etc.).
- Distinctive: line 398
Do the 5 bullets each answer a question / cover a distinct dimension?
- Very specific: line 440
Read-only (analysis + copy).
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Amazon Listing Optimizer
Optimizes an Amazon listing from your own live data, tuned for how Amazon actually ranks and recommends in 2026 (the 75-character title cap, the COSMO intent layer, and the Rufus / "Alexa for Shopping" AI assistant). It diagnoses where the listing leaks against the market (are you seen, clicked, added to cart, bought?), separates an exposure problem from a copy problem, finds the converting themes your title forgot to say, audits intent coverage the way COSMO reads a listing, and returns a rewritten, policy-compliant, upload-ready title, bullets, backend terms and attribute fills with a prioritized, measurable fix list. All from DataDoe - no manual exports, no screenshots, no third-party tools.
When to use this
- A listing under-converts or isn't ranking for the terms it should.
- Before a push/relaunch, or after a title/bullet edit, to check coverage + funnel.
- New variation or ASIN that needs its copy dialed in.
- To bring an old, long, keyword-stuffed title into 2026 policy before Amazon auto-rewrites it for you.
- Trigger phrases: "optimize my listing", "listing optimization", "fix my title", "improve my bullets", "keyword gaps", "why isn't my listing ranking", "why isn't it converting", "backend keywords", "listing audit", "is my listing Rufus ready".
What changed in 2026 (design this skill around it)
- Title cap is 75 characters (including spaces) for all categories except media, enforced from 27 July 2026. Over-length or non-compliant titles are auto-rewritten by Amazon's own AI without seller approval - so a tight, compliant title is not optional. The title can no longer hold every keyword; it holds the brand + the one best-converting theme + the audience/use-case. Keywords now live in bullets, backend and structured attributes.
- COSMO is Amazon's intent/semantic layer on top of A9. A9 still uses keyword match, sales velocity, CTR and CVR; COSMO adds "does this listing actually match what the shopper means". It rewards breadth of intent coverage (who it's for, what it does, where/when it's used, what it is, what it pairs with) and penalizes keyword repetition, generic descriptors and empty attribute fields. Breadth of intent beats depth of keyword repetition.
- Rufus / "Alexa for Shopping" now mediates a large share of shopping queries and can surface an "AI research" summary above listings. It favours listings that plainly answer who the product is for, what problem it solves, what it is made of, how it compares, and that pre-empt objections - in natural language, backed by complete data and ratings.
The framework. Diagnose the funnel vs the market, then fix the right thing
For every important query the ASIN touches, the search funnel is impressions -> clicks -> cart-adds -> purchases. Read each stage against the market average for that same query, not against an absolute number - a 2% click rate can be excellent on one query and terrible on another.
Per query, compute:
- Impression share =
child_asin_impression_count/search_query_total_impression_count. - Your CTR =
child_asin_click_count/child_asin_impression_count, and the market CTR =search_query_total_click_count/search_query_total_impression_count. - Cart rate =
child_asin_add_to_cart_count/child_asin_click_count. - Your CVR =
child_asin_purchase_count/child_asin_click_count, and the market CVR =search_query_total_purchase_count/search_query_total_click_count. - Search Query Score =
child_asin_search_query_score, kept per ASIN, marketplace and period as context. Amazon ranks queries relative to other queries for the same ASIN based on overall query performance; 1 is highest. This is not the ASIN's organic search position. Do not take a minimum or aggregate scores across periods or ASINs.
Then classify each high-value query:
- Exposure problem - you win on CTR/CVR (at or above market) but impression share is low. Investigate: ads, offer, stock and relevant keyword coverage before rewriting copy; stronger exposure may be the highest-upside opportunity.
- Keyword coverage gap - a relevant, high-volume query has low impression share and its terms are absent from the listing. Fix: consider a bullet / backend / attribute (not necessarily the 75-char title). Low share or a Search Query Score does not prove that the term is not indexed; verify indexing separately.
- Click-rate (below-market CTR) - impressions but CTR below the market CTR for that query -> main image / title / price / badge. (If CTR is low in absolute terms but at market, there is nothing to fix - don't chase it.)
- Conversion (below-market CVR, or cart-adds that don't convert) -> bullets / A+ / images / price / Prime / reviews.
- Relevance guard - low impression share AND below-market CTR AND below-market CVR is usually not your product. Do not chase it into the title.
The single highest-leverage move: does your title say what converts?
Before rewriting anything, run this once - the most common, highest-ROI miss:
- Cluster the queries into themes.
- Per theme, sum volume and take the CVR and CTR-vs-market.
- Find the theme with the best conversion and real volume - what buyers actually want from this product.
- Check whether that theme's words are in the title. If your best-converting theme is missing, that is the #1 fix - and since the title now holds only 75 characters, it must earn its place with the single best-converting theme, not a keyword pile.
Intent-coverage audit (how COSMO reads the listing)
Across the title + bullets + description + attributes, check that the listing explicitly answers each dimension at least once (most listings cover only 2-3; aim for the lot):
- WHO it's for (audience / use case).
- WHAT it does (function + what its materials/features enable, not just specs).
- WHERE / WHEN it's used (location, occasion, frequency).
- WHAT it is (product type / category).
- WHAT it pairs with / compares to. Each real keyword appears once, in the field where it fits best - repetition is wasted space and a COSMO/title-policy penalty.
Rufus-readiness check
Would the listing let an AI assistant answer: who is this for, what problem does it solve, what is it made of, how does it compare, and what do buyers complain about (objections)? If any answer is missing, add it in a bullet or A+ in plain language. Mine 3-4 star reviews for the real audience language and objections.
Configuration
- MCP base:
https://mcp.datadoe.com/mcp/v1 - Data sources (resolve each by table name with
exports_sources_get):amazon_child_product_organic_search_ranks_per_week(SQP) - the funnel engine and the one required source. Per query:search_query_volume,child_asin_impression_count+search_query_total_impression_count,child_asin_click_count+search_query_total_click_count,child_asin_add_to_cart_count,child_asin_purchase_count+search_query_total_purchase_count,child_asin_search_query_score,child_asin_median_click_price_value/_currency. Coverage is per configured ASIN and often sparse (weeks can be missing and the latest week stale), and each week carries at most ~100 queries (the report cap, not the full query set) - probe coverage first (workflow step 4a). Fallback when weekly coverage is thin:amazon_child_product_organic_search_ranks_per_month(same funnel columns, one row per month).amazon_products_by_child_asin(fallback / supplement for content) -product_name(the displayed title),product_bullet_point_1..5,product_description,product_brand, category, BSR,product_image_url,marketplace_country_code. It returns a row for every marketplace the seller is in (e.g. CA / DE / GB for one UK seller) - always filtermarketplace_country_codeto the target marketplace (rows useGB;sellers_and_vendors_listreportsUKfor the same seller). Preferamazon_listings_rawfor submitted title/bullets/description/brand; use this for category / BSR or whenattributesis sparse.amazon_listings_raw- the real backend-keyword + attribute source, and the place to see which attribute fields are empty (the "death of null" gap COSMO penalises). One row per SKU/marketplace, a current snapshot (uselast_seen_atfor freshness, not a time series). JSON columns:attributes,summaries,issues,offers. Parseattributes:- Backend search terms =
attributes.generic_keyword[].value- an array of{language_tag, marketplace_id, value};valueis Amazon's real "Search Terms" blob (space-separated). Pick the entry matching the targetmarketplace_id. Nogeneric_keywordkey = the SKU has no backend terms at all (a concrete miss). - Listing content (richer than the catalog source):
item_name,bullet_point[],product_description(HTML),brand,color,size, image locators (main_product_image_locator,other_product_image_locator_1..5),list_price. - Two titles, not one.
attributes.item_nameis the title you submitted;summaries.itemNameis the title Amazon displays (it equalsamazon_products_by_child_asin.product_name). They can differ wildly - 66 vs 149 chars observed on the same SKU. Measure the 75-char rule onsummaries.itemNameand flag the divergence when the two differ: Amazon is not showing your contribution, so a compliant submitted title does not make the listing compliant. - Empty-attribute gaps: compare the
attributeskeys against what's expected for the SKU'ssummaries.productType(types differ -SHARP_PINcarries far fewer fields thanSHOE_INSERT); an expected field absent or an empty array is a gap. - Health:
issues[](code,severity,categories,enforcements.actions) - flag suppressions so you don't "just add keywords" to a blocked listing. - Multiple SKUs per ASIN: a
child_asin+marketplace_idcan have more than one row - different seller SKUs for the same ASIN (new vs used, FBA vs FBM, a superseded listing). Don't treat the first or richest row as canonical: checksummaries.statuson each and audit theBUYABLEone(s) - that is what's actually live. If a non-buyable sibling SKU carries better content (populatedgeneric_keyword, fullerbullet_point/product_description) than the buyable SKU(s), flag it explicitly: the good copy exists but isn't on the listing that sells. Usesummaries.lastUpdatedDateto tell which is the newer edit. Rows are large JSON - pull only the columns you need and download to a file; parse programmatically, don't dump raw JSON into context.
- Backend search terms =
amazon_ads_search_terms_by_campaign_by_date- converting ad search terms (keyword harvest the SQP set may miss).amazon_fba_inventory_health(optional, premium) -your_price,featuredoffer_price,lowest_price_new_plus_shipping,availablefor a price/stock read (a CVR leak is often price or an out-of-stock, not copy). Premium note: a premium export costs 5 AI Tokens instead of 2 - nothing else differs, and the table is part of the always-on default dataset, so it is never disabled. A 0-row export means no data in the window or an initial load still in progress - say which, and useamazon_listings_with_cogs.listing_price_value/fba_quantity_availableinstead; never render zeros.amazon_brand_analytics_search_terms_weekly(Brand Registry only, optional) - top terms with #1-3 click/conversion share. It is org/marketplace-scoped, not account-scoped (noseller_or_vendor_idcolumn) - filter by marketplace and treat the clicked ASINs as market-wide (often competitors). If absent, skip and say so - never block.
- Use
child_asin_search_query_score.child_asin_organic_search_rankis a deprecated compatibility name for the same data and should not be used in new exports. - The SQP table name stays unchanged for compatibility; the source does not provide organic search position or separate organic and sponsored visibility.
- Inputs: the target ASIN/SKU (+ marketplace if multi).
Step-by-step workflow (MCP-native)
sellers_and_vendors_list-> pick the seller; get the target ASIN.exports_sources_get-> confirm sources; SQP is required, the rest degrade gracefully.- Current listing:
exports_createonamazon_listings_raw(columnssku,child_asin,marketplace_id,attributes,summaries,issues,last_seen_at) filtered to the ASIN; download to a file and parse. Fromattributesreaditem_name,bullet_point[],product_description,brand; the backend terms fromattributes.generic_keyword[].valuefor the targetmarketplace_id;productTypeanditemNamefromsummaries; andissues[]. Record both titles with their lengths: displayed =summaries.itemName(this is the one the 75-char compliance verdict uses) and submitted =attributes.item_name. If they differ, say so in the header - Amazon is displaying a different title than the one on file, and the submitted title's length proves nothing. Fall back toamazon_products_by_child_asinfor title / bullets / category / BSR ifattributesis sparse - filtered to the targetmarketplace_country_code, since it returns one row per marketplace of the seller. Read the marketplace's caps now (Copy rules below). - SQP coverage probe, then the funnel per query (the core):
- 4a. Coverage probe first.
exports_createon the SQP weekly source for the ASIN over the intended window (>= 8 weeks),groupBy: ["date"]withcountDistinct(search_query)andsum(child_asin_impression_count). Report which weeks exist and the age of the latest one (observed: 4 non-contiguous weeks in a year, latest 8 weeks old, exactly 100 queries per week - the per-ASIN report cap, not the full query set). If fewer than ~4 weeks are present in the window, either fall back toamazon_child_product_organic_search_ranks_per_month(same funnel columns, monthly rows) or run in "snapshot" mode on the weeks that exist - and say so in the header: which periods the analysis rests on and how stale they are. Never present a one-week sample as an 8-week trend. - 4b. Funnel pull.
exports_createon the SQP source for the ASIN over the window (or the periods found in 4a), grouped bysearch_query, pulling BOTH thechild_asin_*and thesearch_query_total_*columns so you can benchmark against the market. Compute impression share, your-vs-market CTR, cart rate, your-vs-market CVR. Sum funnel counts over the window before dividing; if showing Search Query Score, pulldateseparately and show per-period values without aggregation.
- 4a. Coverage probe first.
- Theme check (first): cluster into themes, find the best-converting theme with real volume, check whether its words are in the title.
- Classify each high-value query as exposure / keyword coverage / CTR / CVR / relevance using the market benchmarks. Split "you win but aren't seen" (investigate ads/offer/coverage) from "you're seen but lose" (copy).
- Intent-coverage + Rufus audit: score the listing on the COSMO dimensions and the Rufus questions above; note which are missing.
- Harvest + competitive gap: pull converting ad terms; if Brand Analytics is available, find terms where a competitor takes the #1-3 share you should own; else skip and note it.
- Coverage + backend-keyword + attribute audit:
- Tokenise title + bullets +
attributes.generic_keyword[].value(backend). Flag the high-value SQP money-keywords (from step 4) that are absent from the backend string- a concrete "add to your Search Terms" fix; if the SKU has no
generic_keywordat all, flag that as a clear miss.
- a concrete "add to your Search Terms" fix; if the SKU has no
- Empty-attribute gaps: for the SKU's
summaries.productType, list expected fields that are missing or empty (at minimumbullet_point,product_description,generic_keyword,main_product_image_locator) - "death of null". - Match everything by marketplace (
marketplace_id) so a multi-marketplace ASIN's keywords/attributes don't bleed across markets. - Check price vs
featuredoffer_priceandavailableif pulled; flag listingissues[]/ suppressions (don't just add keywords to a blocked listing).
- Tokenise title + bullets +
- Rewrite (2026-compliant, upload-ready):
- Title <= 75 chars: brand first, then the single best-converting theme, then audience/use-case. One keyword each, no repetition, no promo/subjective words, no banned symbols. This is a precision line, not a keyword dump.
- 5 bullets: each answers a customer question / covers a distinct COSMO dimension; benefit first, then the feature that proves it and the context it's used in; keep any legally required text; distinct keywords across the five.
- Backend search terms: fill the byte cap (commonly ~250 bytes; verify per marketplace) with the relevant terms that did NOT fit the title/bullets - this is now where most keyword coverage lives; no title repeats, no brand/competitor names, include synonyms, misspellings, other-language terms.
- Attributes: list the empty structured fields to fill (audience, intended use, occasion, material, etc.) - "death of null".
- Set a measurement plan: ship the single highest-impact change first (usually the title), then wait - semantic/indexing changes take days to a few weeks to reflect. Re-run this skill and compare the same queries. Don't change everything at once.
Copy rules (Amazon 2026 policy + best practice, marketplace-safe)
- Title (<= 75 chars, non-media): brand first; primary converting keyword early;
readable human phrase, not a list; each word appears at most twice (prepositions,
articles, conjunctions exempt) and real keywords ideally once; no promotional words
("best", "sale", "#1", "free shipping"), no subjective claims, no ALL-CAPS blocks;
banned symbols
! $ ? _ { } ^ ¬ ¦unless part of the brand, and~ # < > *only for a style identifier or a measurement; capitalise the first letter of each major word. If the current title is over 75 chars, treat compliance as an urgent fix (Amazon will otherwise auto-rewrite it). - Bullets: 5, benefit first then the proof feature and the use context; one COSMO dimension / customer question each; distinct keywords across the five; keep required regulatory/safety text; no keyword-stuffing.
- Backend search terms: read them from
amazon_listings_raw.attributes(Amazon attributegeneric_keyword). They are not a column onamazon_products_by_child_asin. Fill the marketplace byte cap; NO words already in the title/bullets; no brand or competitor names; synonyms, misspellings, other-language terms; space-separated, no commas needed. - Attributes: fill every relevant structured field (audience, use, occasion, material, scent, etc.) - empty fields are COSMO knowledge gaps.
- Localize: marketplace language + currency; never reuse another marketplace's copy.
What the data can't see - flag these to check manually
- Organic search position and indexing - SQP cannot establish either. Search Query Score is relative query performance for one ASIN, and SQP shares do not isolate organic visibility. Verify positions or indexing separately if needed.
- Main image + image count - a below-market CTR is very often the main image; aim for a clean main image plus 6-7+ images with use-case/infographic content (COSMO now reads use-case imagery, not decorative icons).
- A+ Content / Brand Story / video - major CVR + intent-coverage levers; use A+ comparison charts to answer "how does it compare".
- Star rating + review count/velocity - a below-market CVR with good copy is often a reviews problem.
- Price / coupon / deal / badges - pull
featuredoffer_priceif available, else verify price and whether a coupon/Best-Seller badge would lift CTR/CVR. - Compliance - never invent claims (medical, "cures", etc.); keep required regulatory text; restricted-claim categories (supplements, biocides, medical devices) need extra care.
Output format
Listing Optimizer - {ASIN} - {marketplace} (displayed title {chars}/75 chars {FLAG if >75})
{if submitted != displayed: "Submitted item_name is {chars}/75 but Amazon displays a different {chars}-char title - contribution not shown."}
SQP coverage: {n} weeks present ({first}..{last}), latest {age} weeks old{; "snapshot mode" / "monthly fallback" if < ~4 weeks}
THEME CHECK (what actually converts)
theme vol/mo your CVR CTR vs mkt in title? -> action
{best theme} {v} {cv}% {x}x NO add to title (headline)
...
FUNNEL DIAGNOSIS (top queries, vs market)
query vol impr-share yourCTR/mkt cart% yourCVR/mkt query score (context) verdict
...
EXPOSURE UPSIDE (you win but aren't seen)
{kw}: impr-share {s}%, CTR {x}x market, CVR at/above market -> investigate ads / offer / stock / keyword coverage
INTENT COVERAGE (COSMO): WHO {y/n} · WHAT-does {y/n} · WHERE/WHEN {y/n} · WHAT-is {y/n} · PAIRS/COMPARE {y/n}
RUFUS-READY: for-whom {y/n} · problem {y/n} · made-of {y/n} · compare {y/n} · objections {y/n}
COMPETITIVE GAP (only if Brand Analytics available)
{term}: your share {x}% (top competitor {y}%) [or: "Brand Analytics not enabled - skipped."]
KEYWORD COVERAGE: {covered}/{topN} high-value terms in the listing
Missing (relevant): {kw (vol, impression share)} -> {title/bullet/backend}
EMPTY ATTRIBUTES to fill: {field, field, ...}
CHECK MANUALLY (data can't see): main image / images / A+ / rating / price-coupon
--- UPLOAD-READY (paste into listing / flat file) ---
item_name (Amazon listing attribute for catalog `product_name`, {chars}/75):
{new title, <=75 chars, compliant}
bullet_point1..5:
1) {benefit + proof + context, one COSMO dimension}
...5
generic_keyword (inside `amazon_listings_raw.attributes`, {bytes}/{cap}):
{relevant terms not in title/bullets}
attributes_to_fill:
{audience=..., intended_use=..., occasion=..., material=...}
FIX FIRST (by impact, measurable)
1) {title: add best-converting theme / bring under 75 chars (headline)}
2) {below-market CTR on "Y" -> new main image + title hook}
3) {exposure: you win "Z" but see {s}% -> investigate ads / offer / stock / keyword coverage}
Measure: ship #1 now, allow days-to-weeks for re-index, re-run and compare.Worked example (illustrative)
An odor spray out-clicks the market ~3x on the "against odor" theme and converts well, but "odor" is absent from the title -> headline fix: since the title now holds only 75 chars, spend them on brand + "against odor" + the product type, and push the rest of the keywords into bullets and backend. On its best terms it shows for only ~4% of impressions while beating market CTR and CVR -> exposure upside: investigate ads, offer, stock and keyword coverage before rewriting copy. A "disinfectant" theme clicks at market rate but converts below market -> a page/trust fix (and a defensible claim only if the product qualifies). The intent audit shows the listing never says WHERE it's used or WHAT it pairs with -> add those to bullets for COSMO. Output: a compliant title, question-led bullets, a full backend line, a list of empty attributes to fill, and a "do this first, measure in days-to-weeks" plan - a diagnosis and the rewrite, not a keyword list.
Quality self-check
- Is the rewritten title <= 75 characters, brand-first, one keyword each, no
promo/subjective words, no banned symbols? Did I flag the current title if it's over -
measured on the displayed
summaries.itemName, not only the submittedattributes.item_name, and did I flag it when the two differ? - Did I probe SQP coverage (
groupBy date) before the funnel and state which weeks the analysis rests on, how stale they are, and whether I fell back to monthly / snapshot mode? - Did I filter
amazon_products_by_child_asinto the target marketplace? - Did I benchmark each funnel stage against the market (not absolute numbers) and separate exposure (ads/offer/coverage) from copy (image/page) problems?
- Did I run the theme check first and put the best-converting theme in the title?
- Did I audit COSMO intent coverage (who/what-does/where-when/what-is/pairs) and the Rufus questions, and route missing keywords to bullets/backend/attributes?
- Did I use the cart-add stage to catch last-mile (price/Prime/stock) leaks?
- Did I apply the relevance guard before putting any term in the title?
- Do the 5 bullets each answer a question / cover a distinct dimension? Backend filled, no title repeats, no brand/competitor, within the byte cap? Empty attributes listed?
- Did I read the actual backend terms from
attributes.generic_keyword[].value(for the right marketplace) and flag high-value SQP terms missing from them - or flag a SKU with nogeneric_keywordat all? - Did I judge empty attributes against the SKU's
productType, not a universal schema? - Did I flag the blind spots (image, A+, rating, price) and keep compliance text?
- Did I give a measurement plan (one change first, days-to-weeks re-index, re-run)?
Common mistakes
- Writing a long, keyword-stuffed title - it now breaks the 75-char policy and COSMO flags it; Amazon may auto-rewrite it against you.
- Measuring title compliance on
attributes.item_namealone - a 66-char submitted title can sit behind a 149-char displayedsummaries.itemName; the displayed one is what Amazon enforces. - Running the funnel on whatever SQP rows come back without checking coverage - one stale week dressed up as an 8-week trend, or 100 capped queries read as the whole market.
- Reading
amazon_products_by_child_asinunfiltered and mixing another marketplace's title / bullets into the audit. - Reading CTR/CVR as absolute numbers instead of against the market for that query.
- Treating Search Query Score as organic position or proof that a term is not indexed.
- Aggregating query scores into a "best organic rank" across periods or ASINs.
- Treating an exposure problem (you win but aren't seen) as a copy problem.
- Repeating a keyword across title/bullets/backend instead of covering a new intent dimension (breadth beats repetition).
- Leaving structured attribute fields empty ("death of null").
- Blaming copy for a below-market CVR that is really reviews or price.
- Ignoring listing issues/suppressions that block the listing regardless of copy.
- Assuming a fixed attribute schema across product types - expected fields differ by
productType; judge gaps per type, and read backend terms fromattributes.generic_keyword[].value(per marketplace), not a guessed field. - Auditing whichever
amazon_listings_rawrow comes back first when an ASIN has multiple seller SKUs - always checksummaries.statusand audit theBUYABLEone(s); a rich but non-buyable sibling SKU is a real, easy-to-miss gap (good backend keywords sitting on a listing nobody can buy). - Fluent AI-written copy with no structured intent signals - reads well, ranks poorly.
- Changing everything at once, then judging before the days-to-weeks re-index.
- Inventing medical/other restricted claims to win a keyword.
Notes
- Read-only (analysis + copy). Applying the changes is a separate write skill via
AMAZON_LISTINGS_UPDATE(dryRun-gated). - Tuned for 2026 Amazon: 75-char title cap, COSMO intent ranking, Rufus / Alexa for
Shopping. Built on DataDoe SQP (with market benchmarks), listing attributes + backend
keywords (
amazon_listings_raw.attributes.generic_keyword), catalog, search-term, inventory and (when available) Brand Analytics data.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

