Amazon Category Audit: What Stops Revenue Before the Ads?

How to find what holds revenue back before you blame the ads. A white paper for Amazon agency owners and the brands they serve.

Before you blame the ads, check the layers above them. Start with the category's structure and what shoppers pay for. Then read why products come back and what the page promises. The audit below reads those layers first, in a fixed order. The ad data comes last. By then it reads as a symptom with a known cause.

Executive summary

"Our ACOS on generic terms is out of control. Please fix the bids."

If you run an Amazon agency, you have read that email more than once. The complaint arrives in ad language. The answer goes out in ad language. The team moves bids and budgets. The report looks busy for a month. Then the same numbers return, because the cause never sat in the campaigns.

This method is a structured read of an Amazon listing's whole market before anyone touches a bid. It sizes the category and maps what shoppers pay for. It splits search demand into clean pools. It reads returns and reviews, then defines the real competitive set. Last, it judges the account's own page and ads. Its job is to find where the revenue stops and which fix comes first.

The sections below cover the problem and four common agency fixes. Then comes the method I use. After that, the paper walks through one anonymised account: a premium Home and Kitchen product, 3 months old on Amazon US, already ranking well in its premium lane. Execution was fine. The audit found the constraint through five findings:

Brand searchers converted at 7.40%. Category searchers converted at 1.22%.
45% of ad spend sat on generic category terms at 87.5% ACOS.
About 545,000 page views from external traffic converted at 0.068%, one order per 1,466 visits.
43.25% of category returns came from expectation failures. Photography and copy fix those at zero unit cost.
Every title in the competitive lane ran 156 to 198 characters against Amazon's 75-character cap.

The paper closes with a checklist. Use it to run the method yourself, or to judge an audit someone else sends you.

Why do agencies misread Amazon accounts as ad problems?

Agencies misread accounts as ad problems because the ad console is the most visible report in the business. It records every other weakness as ad cost. A page losing cold shoppers shows up there as a worse ACOS. A return at unboxing wastes the ad spend behind the sale. No bid fixes either of them.

Picture where a weak page shows up. A shopper lands from a generic search on a phone. The title stops after the first line. The main image says nothing about size. The shopper leaves. Amazon charged for the click. The campaign records a click with no sale. After a few thousand of those, the campaign shows a high ACOS. The campaign report now carries a loss the page caused.

Returns hide in the same place. A shopper sends the product back because it arrived larger than the shopper expected. The sale reverses. Amazon keeps the ad spend behind that sale. The campaign paid for a sale that no longer exists.

The client sends a screenshot of ACOS. You open the bulk file. Three weeks later, both sides argue about a number neither of them caused. The mistake is where the diagnosis starts: inside the campaign manager.

In the case below, about $20,000 of ad spend returned nothing or less than nothing. A bid-first review would have kept most of it running, at new bids. The client hired you to fix revenue. If your first 60 days go into the one lever that never caused the problem, you lose an account you were able to fix.

Why do agencies' first fixes miss the cause?

Agencies reach for four fixes before an audit like this: bid and budget tuning, keyword expansion, a listing refresh and a cold tool or AI audit. Each one misses for the same reason. It starts with the account and stops there. The cause sits further out, upstream of the campaigns.

Bid and budget tuning

When you lower bids on high-ACOS terms, the ratio comes down. Traffic comes down with it. The page should win those terms after someone fixes it. When a page converts category shoppers at a sixth of its brand-shopper rate, a bid cut makes the campaign smaller. The conversion gap stays where it was. The ACOS line on the report looks better, with revenue flat.

Keyword expansion

More keywords feel like more demand. On a page already ranking across its core terms, new keywords add reach to a page with a conversion problem. The new terms bring less relevant shoppers to the same weak page.

A listing refresh without category data

A copywriter rewrites the bullets. Nobody checks which buyer pays this price. Nobody reads why products in the category come back. The new bullets speak to the wrong shopper. Conversion stays where it was.

A cold tool or AI audit

A tool reads the account and returns a confident verdict: ACOS too high, cut bids. The tool reads only the campaign data. It treats a symptom as the cause.

I use AI every day. In my own work it cut keyword segmentation from 50 minutes to 5. It sorts and groups thousands of search terms in minutes. Deciding which demand earns the spend stays with the operator. So does matching the price to its buyer.

What is a category audit?

The category audit is my method for finding held-back revenue on an Amazon listing before any ad change. It reads the account in a fixed order from the outside in. Market first, then customer, competition and current position. Traffic and ads come last. Each layer explains the next one. By then the audit has mapped what causes the ad results.

Market → Customer → Competition → Current position → Traffic and ads

The order is the method. After the first four layers, the ad data shows a symptom with a cause you have already found.

The audit has thirteen parts across four layers. Below is what each part asks and where its data comes from. Each part also names the held-back revenue signal it catches. This paper shares what each part does and why it matters. The working templates stay inside client work.

Layer 1: the market

Category size and structure. How big is the category? Who holds it? How does it split? The data comes from a market tool such as Helium 10 Xray across the top parent listings. Some categories split cleanly by format or technology. In those, cross-format benchmarks mislead. When your product competes inside one format, take its competitors and ad targets from that format only.
What the category pays for. Which claims carry revenue? A revenue index is a claim's share of category revenue divided by its share of offers. Above 1.0, the claim earns more than its presence suggests. Watch for one signal. Your product delivers a high-index claim. Your title leaves it out.
How shoppers search the category. What is the clean search universe? Which pools of demand sit inside it? A raw keyword export overstates demand, because adjacent products share the same words. Keyword Mastery is my system for this layer: intent and relevance over volume. Every search term goes into exactly one pool by intent. Each pool gets a status: owned, contested, closed or not relevant.
When the category buys. Where are the peaks? What does each peak want? Some seasonal categories have two peaks with different buyers. The signal is a listing with no cue for the peak that fits its price tier.

Layer 2: the customer

What earns loyalty and what breaks it. Amazon Product Opportunity Explorer splits review topics into praise and complaints. Read both. Ask which drivers are now baseline expectations, the things every shopper assumes. Failing one shows up fast in complaints.
Why customers return the product. Returns and complaints are two different lists. A return at unboxing points to a gap between the listing and the box. A complaint after weeks of use points to the product. A page-driven return reason is the cheapest fix in the audit. It costs photography and copy.
Who the buyer is. Which segment buys at this price? Which segment does the listing speak to? The signal is a mismatch. The copy and imagery speak to one segment. The price sits in another segment's band.

Layer 3: the competition

The competitive set a shopper sees. Who are the real rivals for this shopper? The real set is the handful of products one shopper compares. That shopper has this budget and this intent. Read revenue per single listing and per parent listing, because variations inflate parent revenue.
How competitors write and sell. Which conventions do the rivals in the lane share? A function count in the title is one. A stated time or capacity is another. Box contents and third-party validation are common too. A listing following none of them asks the shopper to do extra work.
The product constraint. This is the one part about the product itself: a size or setup fact the listing has to state plainly. The signal is creative claiming the opposite of what the product delivers. Shoppers find out at unboxing. The return follows.

Layer 4: current position

Performance and trajectory. Is execution the constraint? Check lane rank and review velocity against the category median. A young listing already ranking well in its lane tells you the team executes. Then look for the constraint in the other parts.
Price position. Where does the price sit against the lane average and against the two offers a shopper sees beside it? Check how the price reads to a shopper from a category search. That shopper needs proof before paying a premium.
The listing: copy and imagery. Read the title against the current character cap. Read the bullets against the objections in the returns data. Compare the main image with the lane. Then read the gallery slot by slot. For each slot, ask whether it gives a reason to buy. Also ask whether it raises a doubt too early.

The audit opens the traffic and advertising data after these thirteen parts. By then you know which search pools the page should win. You also know which buyer it should speak to.

The measurement rules

Three rules keep the audit honest.

Measured or modelled. Every claim carries a tag. Measured means it came from the data. Modelled means it rests on an assumption. Write the assumption next to the claim.
Amazon-native conversion only. Report conversion on shoppers who arrived through Amazon. Keep external traffic in its own line. A blended rate turns a traffic fix into a fake conversion gain. It also leaves the creative work with a moving baseline.
One segment per search term. Every term in the search term report goes into exactly one segment: brand only, product line, brand plus category, technology or format, generic category, competitor conquest. Then read conversion and ACOS per segment. A blended campaign ACOS hides the answer.

What did the audit find in a real account?

The audit found a page converting brand searchers at 7.40% and category searchers at 1.22%. It found external traffic diluting the reported conversion rate. The listing's own promises drove the returns. The audit also found a lane-wide title problem. The listing was a premium Home and Kitchen product, 3 months old on Amazon US, already ranking well in its lane.

The brand team asked the question every agency hears: why does the advertising cost so much? The audit ran on about one quarter of daily trading data, plus market, keyword, review and returns data for the category.

The market said: the money sits in the top price tier

In this niche, the top price tier held 64.6% of revenue on only 9.9% of clicks. A plan that follows click share would under-invest in the tier carrying the revenue.

Review count worked as a threshold in this niche. For a 3-month-old listing, review velocity had to sit inside the plan from day one, next to the ads.

The customer said: the page promised the wrong thing

43.25% of returns in the category were expectation failures: the product arrived different from what the shopper expected. Those same failures were only 10.30% of complaints. Shoppers who felt the page misled them sent the product back at unboxing, before any review.

In the returns read, 29.76% of returns gave the reason "larger than expected". What surprised the shopper was the size of the product itself. Photography and stated dimensions fix that, at zero unit cost.

The buyer read showed a second gap. The copy and lifestyle imagery spoke to a family and kids segment. The price sat in the band where premium enthusiasts buy. The copy sold to a family buyer at an enthusiast's price.

The competition said: the whole lane is out of date

Every title in the competitive lane ran 156 to 198 characters. Amazon now caps product titles at 75 characters in every category except media. Its AI rewrites longer titles gradually, on Amazon's own schedule. Brand owners get up to 14 days to review those rewrites. Every long title in this lane faced one. The first brand to write a 75-character title holding its key claims would move ahead of every rival at once.

The listing also followed none of the four conventions its lane used. It carried no seasonal cue for the peak that fit its price tier.

The traffic said: the reported conversion rate was wrong

About 545,000 page views came from external traffic and converted at 0.068%, one order per 1,466 visits. Those visitors pulled the recorded conversion rate down to 0.42%. Shoppers arriving from Amazon itself converted at 0.81%, nearly twice as well.

This matters beyond the report. My own expectation is that conversion history feeds organic rank. I tag it as modelled. If that holds, half a million low-intent visitors made the page look weak to Amazon. Inside the account, I expect the external campaign hurt the organic rank the ads were building.

The ads said: the account harvested brand demand

Now the audit opened the search term report and split it into segments.

Segment	Spend	Sales	Conversion	ACOS
Brand name only	$436	$3,684	7.40%	11.9%
Brand plus category	$1,498	$21,244	3.73%	7.1%
Generic category terms	$21,051	$24,051	1.22%	87.5%
Competitor conquest	$9,913	$11,078	1.31%	89.5%

The table shows four of the six segments. The 45% share below counts all six.

Generic category terms took 45% of ad spend. About $20,000 of spend returned nothing or less than nothing. Off-intent targeting ran at 117.5% ACOS.

After the first four layers, this table points to the page. Shoppers arriving through generic terms did not know the brand. The page gave those cold shoppers no reason to pay a premium price. It stated no dimensions in text and made no price argument.

What the plan put first

The audit ended in a plan in dependency order. It follows my brand-level Fix Order: positioning, then offer, then retail readiness, then conversion stability, then PPC scale. This listing's positioning and offer passed. The break sat at retail readiness. The copy and images spoke to the wrong buyer. So PPC scale waits until conversion holds. Phase 1 cleans the measurement first. Conversion stability needs a clean read:

Phase 1, days 0 to 14, traffic and spend integrity. Route external traffic off the detail page. Cut cross-category conquest by 75%. Pause off-intent targeting running at 117.5% ACOS. Keep the brand term. It has the best conversion rate in the table above.
Phase 2, days 14 to 45, copy. Rewrite the title to 75 characters, with the high-index claims in it. Rewrite the other copy fields to match.
Phase 3, days 30 to 75, imagery and A+. Dimensions at human scale. Images show the product doing its job.
Phase 4, days 45 to 120, compounding. Head-term rank, review velocity toward 1,000 and peak-season creative.

Traffic integrity goes first for a measurement reason. Until the external traffic leaves the detail page, the conversion rate cannot show whether the new copy and images worked.

This paper reports audit findings only. The audit also modelled an opportunity range. Those figures are projections. I left them out of this paper.

What should an agency owner check in an audit like this?

An agency owner should check that the audit answers thirteen questions with data, each with a measured or modelled tag, from category structure through to the plan order. Use the list below on your own accounts. Also use it to judge an audit from a vendor or a new hire, before it reaches your client.

Does it size the category from a clean universe, without adjacent products?
Does it define the competitive set by format, price and intent?
Does it show which claims the category pays for, with a revenue index or an equal measure?
Does it split search demand into pools by intent, with a status for each pool?
Does it name the peaks and match the product's price tier to one?
Does it read returns and complaints as two separate lists?
Does it name the buyer the price fits? Does the copy speak to that buyer?
Does it state the product constraint? Does the creative tell the truth about it?
Does it report conversion on Amazon-native traffic only?
Does it put every search term into exactly one segment before reading ACOS?
Does it check the title against the current 75-character cap?
Does it order the plan by dependency, with traffic integrity before creative?
Does it keep projections apart from findings?

If an audit misses more than two of these lines, treat it as a campaign review.

How should an agency start?

Start with one account where the client blames the ads. Run two steps before the next bid change: split the search term report into segments, then read the category's returns against its complaints. In my experience they take about a day. They tell you whether the constraint sits in the campaigns or on the page.

Download the last 60 days of the Sponsored Products search term report. Put every term into one segment. Read conversion per segment. If brand searchers convert three or more times better than category searchers on the same page, the constraint sits on the page.
Open Product Opportunity Explorer for the category. Pull the returns and review topics. If the top return reasons point to what the listing promised, the fix is photography and copy. The ads wait.

Then run the full audit before you present a plan. I run these audits for Amazon brands and agencies to find held-back revenue before anyone touches a bid. You walk into the next client call with the real constraint on that account.

The client called about an ad problem. They need someone who finds its cause and sets the order of fixes. That person keeps the bids on the generic terms until the page fixes are live. The ad console shows where the money went. The category audit shows why it did not come back.

How to read the numbers

Account findings come from one anonymised client account. This is an audit. It reports findings from one point in time.
Figures that say "in this niche" are category market data from research tools.
The audit's projections stay out on purpose.
No exact price appears next to a rating or review count. That keeps the listing anonymous.

FAQ

What is an Amazon category audit?

It is a structured read of a listing's market and customers before any ad change. It sizes the category and reads returns and reviews. Then it defines the competitive set. Last, it judges the account's own page and ads.

How is this audit different from a PPC audit?

A PPC audit reads campaigns for wasted spend. This audit looks first for the reason behind the waste, upstream of the campaigns. The PPC read comes last, as a symptom check.

Why does one listing convert differently for brand and category searches?

A category searcher arrives cold and needs the page to give a reason to buy. When the brand rate runs three or more times the category rate on one page, the page is losing the cold shopper.

Should external traffic go to the Amazon detail page?

Send it to a storefront or landing page, or track it in its own report. Low-intent external visitors convert at a fraction of the native rate. They also pull down the conversion rate in Business Reports. In the case here, external traffic converted at 0.068%. Shoppers from Amazon converted at 0.81%.

How long does the audit take?

In my experience, the first answer, from a segmented search term report and a returns read, takes about a day. The full thirteen-part read takes longer, depending on the category and the data access. Run it before the plan goes to the client.

Sources

Case figures: one anonymised client account audit, agency-side.
Category figures: market research tools (Helium 10 Xray, Helium 10 Cerebro, Amazon Product Opportunity Explorer), quoted as market data.
Amazon 75-character title cap, scope, gradual AI rewrites and the 14-day review window: Canopy Management, "Amazon 75-character title limit and Item Highlights", updated 13 August 2026, https://canopymanagement.com/amazon-75-character-title-limit-item-highlights/
Same change: WisePIM, "Amazon title character limit", updated 11 October 2026, https://wisepim.com/guides/amazon-title-character-limit