Where your Amazon revenue actually leaks
Marketplace Diagnostics
Amazon publishes a report that shows you exactly where your revenue is leaking. It is free, it is first-party, and nearly every brand that opens it reaches the wrong conclusion — because four of its rules are counterintuitive and none of them are labeled.
When Amazon revenue softens, the diagnostic conversation almost always happens at the wrong altitude. Sales are down. Ad spend goes up. ACOS climbs, margin compresses, and the underlying problem — which may have had nothing to do with visibility — is still there in ninety days.
The reason is not laziness. It is that the standard reporting stack answers the wrong question. Business Reports tell you what happened to a product. The advertising console tells you what happened to a campaign. Neither tells you what happened to a shopper's intent — the specific moment, on a specific query, where someone who was looking for exactly what you sell chose someone else instead.
One report does. It is called Search Query Performance, it sits four menus deep in Brand Analytics, and it is the closest thing Amazon has ever given sellers to a competitive audit of their own demand.
Most brands treat SQP as a keyword list. It is not a keyword list. It is a loss report.
The instrumentWhat the report actually measures
Search Query Performance lives in Seller Central under Brands → Brand Analytics → Search Analytics. Access requires an active Brand Registry enrollment, a Professional selling account, and a Brand Representative role on the brand. There is no fee. Data can be pulled weekly, monthly, or quarterly, and it reaches back several years.
For each of your top-ranked search queries, it reports four counts at the market level and the same four counts for you — impressions, clicks, cart adds, purchases — and expresses your slice of each as a share.
That structure is the whole point. A raw sales number tells you that you sold less. A share number tells you that someone else sold more, on this exact query, at this exact stage of the funnel. Those are entirely different problems with entirely different remedies, and only one of them is visible in your own P&L.
The Share-Gap Diagnostic
Where the funnel breaks tells you what to fix
An illustrative brand on a single high-intent unbranded query. Share holds through impressions and clicks, then collapses at the cart. The collapse point is the diagnosis.
Read left to right, the story is not "we need more traffic." Visibility is intact. Creative is working. The shopper wanted the product enough to put it in a cart and then abandoned it — which points at price, delivery promise, offer structure, or a Buy Box problem, not at advertising.
The frameworkFour gaps, four prescriptions
The diagnostic value of SQP comes from comparing each share to the one above it. A drop between two adjacent stages localizes the failure. Each localization has a distinct set of causes — and treating one with another's remedy is how brands spend six figures on advertising to fix a shipping problem.
| Where share drops | What it means | Where to intervene |
|---|---|---|
| Impression share is low outright |
You are not being retrieved for the query. Amazon does not consider your listing a relevant candidate. | Indexation and relevance: title and attribute coverage, backend terms, category and browse node placement, structured attribute completeness. |
| Impressions → Clicks |
Shoppers see you and pick someone else from the results grid. The decision is made on the tile, not the page. | Main image, first 80 characters of the title, price shown in results, review count and star rating, Prime badge, coupon flag. |
| Clicks → Cart adds |
They arrived and were not persuaded. The detail page failed to close the intent the query created. | A+ content, image stack sequencing, bullet hierarchy, variation clarity, comparison and specification gaps against the competing listing. |
| Cart adds → Purchases |
They wanted it and did not buy it. Almost never a content problem. | Price and competitive undercut, delivery promise, Buy Box ownership, bundle or multipack economics, stock status at checkout. |
This is the difference between an optimization list and a diagnosis. Anyone can produce a list of things that could be improved on a listing. The report tells you which of them is actually costing you the sale.
The hard partFive rules that produce confident wrong answers
Here is why this report has not become standard practice despite being free and available for years: its counting rules are unusual, they are not surfaced in the interface, and each one quietly invalidates an obvious-looking conclusion.
It uses a 24-hour attribution window
For a purchase to be credited to a query, the entire journey — impression, click, checkout — must complete within one day. A shopper who searches Monday and buys Tuesday shows up as a click and a cart add with no purchase. The higher your price and the longer your consideration cycle, the more purchases fall outside the window. Read a considered-purchase category literally and you will conclude your conversion is broken when it is only delayed.
It only sees traffic that originated in search
Direct-to-detail-page visits, external traffic, and Sponsored Brands placements are not in the dataset. Organic results and Sponsored Products are. If a meaningful share of your demand arrives from off-Amazon or from repeat customers going straight to your page, the report is describing a subset of your business and never claims otherwise.
Purchase counts include cancellations and returns
Amazon does not net these out. In a high-return category this inflates purchase share and can mask a conversion-quality problem as a conversion-rate success. Any serious read pairs SQP purchase share with your actual return rate on the same ASINs.
Cart-add attribution breaks on variation families
A cart add is credited only when the shopper adds the exact product they clicked. On a listing where someone clicks the black one and adds the navy one, that cart add disappears from the query. The wider your variation family, the less reliable this stage becomes — which means the cart-add column requires a completely different level of trust on a two-SKU listing than on a twenty-SKU one.
Branded and unbranded queries must be separated before anything else
Your own brand name will show extraordinary share at every stage, because you are effectively the only relevant answer. Leave those queries in the dataset and they will pull every average upward and hide the unbranded queries — the ones that represent new demand — underneath a comfortable-looking blended number. This single filtering step changes the conclusion more often than any other.
Taken together, these explain the most common objection we hear when a brand first looks at the report: the purchase numbers don't match my sales reports, so the data must be wrong. The data is not wrong. It is measuring something narrower and more specific than total sales, and the gap between the two is structural, permanent, and expected. Once you stop trying to reconcile the absolute numbers and start reading the share ratios, the report becomes useful immediately.
Absolute counts in this report will mislead you. Share ratios will not.
The methodHow we run it
A framework that is not operationalized is a blog post. This is the sequence we use on client catalogs, and the order matters — each step removes noise that would corrupt the next one.
- Split branded from unbranded.Two datasets, always. Branded queries measure loyalty and defense. Unbranded queries measure whether the brand is winning new demand. Growth decisions come from the second set.
- Pull quarterly for strategy, weekly for defense.Quarterly data absorbs Amazon's back-fills and corrections and is the right basis for roadmap decisions. Weekly is a monitoring instrument — it exists to catch a competitor climbing the funnel on a query that matters, while there is still time to respond.
- Rank by revenue at risk, not by search volume.The largest share gap on the highest-volume head term is usually the most expensive fight on the board. A smaller gap on a high-intent long-tail query attached to a high-margin ASIN is frequently the better trade. Volume tells you the size of the room, not the size of the opportunity.
- Localize the break for each priority query.Impression, click, cart, or checkout. One stage per query. If the answer is "all of them," the diagnosis is not finished.
- Match the intervention to the stage — and then leave the others alone.A checkout-stage collapse does not get an advertising budget. A retrieval failure does not get new A+ modules. Discipline about what not to touch is what makes the result measurable.
- Re-pull the same queries and confirm the share moved.The share ratio is the scoreboard. If the intervention was correct, the specific gap you targeted closes, and it closes before total sales move. That lag is the proof the diagnosis was right.
The stakesWhy this compounds
The reason share ratios matter more than sales figures is that share is zero-sum. Every point of purchase share you are not holding on a high-intent query is being held by a competitor — and on Amazon, holding it feeds them. Sales velocity on a query improves ranking on that query. Better ranking produces more impressions. More impressions produce more velocity.
Which means a share gap is not a static loss. It is a gap that widens on its own while you are looking at a flat sales chart and concluding nothing much is happening.
That is the argument we have made consistently: Amazon is an active channel, not a passive one. Having a listing is compliance. Having content is baseline. Knowing precisely which query is losing you which stage of which funnel — and what that specific loss costs in margin, not in impressions — is the work.
See where your funnel actually breaks
Seed Ventures runs share-gap diagnostics as the first step of every marketplace engagement — before any recommendation, and before any spend. We work on a zero-management-fee model: we purchase at wholesale and sell at retail, so our return comes from goods sold. If the diagnosis doesn't produce growth, we don't get paid for it.
Request a marketplace diagnosticReport behavior described here reflects Amazon Brand Analytics as of July 2026. Amazon revises Brand Analytics reporting periodically; verify current attribution rules and report availability in Seller Central before acting on historical exports. Funnel figures shown are illustrative and do not represent a specific client account.