Amazon Negative Keywords: How To Find The Bleeders

There is a column in the Sponsored Products search term report called Customer Search Term, and it is the only file in an Amazon account where wasted money is written out in plain English. Every advertising audit our team runs opens there, sorted by spend, reading the rows that took clicks and returned nothing.

Those rows are what we call bleeders, and they are rarely hiding. They have been spending small amounts nobody flags, for months, which add up to a bad quarter, and Amazon negative keywords are how they get shut off, one search term at a time.

We understand why they survive that long, because nothing looks broken. ACoS drifts up a point or two a month and the campaign still returns something overall, so it feels like the market got more expensive. Underneath that average there is usually a short list of terms doing most of the damage.

What a bleeder actually is

A bleeder is a search term rather than a keyword, and that distinction is the whole reason the problem exists. A keyword is what you told Amazon you want to bid on. A search term is what a shopper actually typed.

Broad and phrase match, automatic campaigns and product targeting all exist to find queries you did not think of, so by design they hand your budget to searches you never wrote down. Sometimes that means a joint supplement shown to people looking for joint compound. The clicks are real, the intent is not.

Amazon reports both sides of that gap. The search term report puts the customer search term next to the keyword or target that triggered it, with impressions, clicks, spend and attributed orders, which makes it a bank statement rather than a dashboard. Source: Amazon Ads help, search term report for Sponsored Products, Amazon primary.

The left column is where most sellers go wrong, because a term with no orders is not automatically a bleeder. A product converting at 10% needs roughly ten clicks before one order is likely, so calling that term dead at four clicks is a guess, and the negative is permanent in a way the data was not.

How to tell if this is you

1. Compare campaign ACoS against the worst ten terms inside it. When the campaign looks acceptable and ten terms are four times worse, the average is hiding the thing you needed to see.
2. Count the distinct search terms that took spend last month. On a tight keyword list this stays modest. Once it runs into the thousands, automatic and broad match are doing your discovery, which is where bleeders come from.
3. Look for the same term in three campaigns at once. That is not diversification, that is your own campaigns bidding against each other on a query you may not want.
4. Check when the negative list was last touched. A list built at launch and never revisited describes a catalogue and a market that have both moved on.

What to check before you add a single negative

The urge to open the report and start excluding is worth resisting for an hour, because a negative is easy to add and easy to forget, and six months on nobody remembers why a good term is blocked.

1. Work out your break-even clicks per order. Average order profit divided by average cost per click gives the number of clicks a term is allowed before it has genuinely failed.
2. Confirm the listing is not the actual problem. If relevant terms convert badly across the board, negatives cut spend without fixing conversion, and you will have cut traffic to a page that needed it.
3. Separate brand terms from everything else. Your own brand name, and the near misses shoppers type instead, belong in their own campaign so the numbers stop flattering the rest of the account.
4. Save a copy of the report first. This data does not stay available forever, and a saved file is the only way to judge a negative later.

The search term workflow, step by step

1. Pull the Sponsored Products search term report for the last 60 days. The path is Amazon Ads, then Measurement and Reporting, then Sponsored ads reports, with Search term as the report type.
2. Sort by spend, descending. Money leaves in order of size, so the top of that sort holds the recoverable spend.
3. Filter to rows with clicks above your break-even number and zero orders. That one filter turns a four thousand row export into a list you can act on.
4. Sort what remains into wrong product, wrong intent, or right term and wrong page. The first two become negatives. The third is a listing job, because excluding a relevant term to fix a page problem loses the term permanently.
5. Choose the match type per term rather than by habit. A single wrong product query becomes a negative exact. A recurring modifier such as free, used or refill becomes a negative phrase, since one entry catches every query it sits inside.
6. Apply at the level the term is costing you, and record the reason. Amazon stores neither the date nor the reasoning, so your own sheet is what lets you unwind a bad call later.

We will not promise a percentage of spend recovered, because it depends on how much discovery your campaigns are doing and how long the list has gone untended. On a supplements account, more than $800 of ad spend was going to one SKU that was not converting, and it surfaced in a weekly report rather than a quarterly one. The cadence found it.

Negative exact and negative phrase, and why the choice matters

Sponsored Products offers negative exact and negative phrase, and there is no negative broad. Negative phrase stops your ad appearing on queries containing the complete phrase or a close variation. Negative exact stops it only on queries that are the exact phrase or a close variation, which in practice means singular and plural forms. Source: Amazon Ads, negative keyword targeting, Amazon primary.

The practical difference is reach. A negative exact on "free protein sample" blocks that one query. A negative phrase on "free" blocks every query the word sits inside, which is how people accidentally block "free from parabens" on a product built around exactly that. Length matters too, since a negative phrase entry runs to four words and a negative exact to ten. Reported by Karooya, 2018, and Ad Badger, 2025, both third party, in agreement.

Structure is what stops bleeders coming back

Cleaning the report once is a task. Not needing to clean it constantly is a structure decision. The version that holds up keeps discovery and harvesting apart: automatic campaigns and broad match find on a small budget, terms that convert graduate into their own exact match campaign where the bid can be set properly, and terms that do not become negatives back in the discovery campaign.

Without that split, discovery keeps rediscovering the same failures. The demand a negative removes does not disappear either, and reaching those shoppers away from the search results is Amazon DSP advertising work rather than Sponsored Products.

Placement is the other half. The placement report splits one campaign into top of search, product pages and rest of search, and those three behave like different channels. A term that looks like a bleeder in aggregate is sometimes only losing money in one of them, and a placement adjustment is the cheaper fix there.

When the whole account runs hot rather than a few terms, that is a different diagnosis, and our high ACoS work starts from the structure.

The weekly routine that keeps it from rebuilding

1. Weekly, read the last seven days filtered to spend with no orders. Seven days cannot judge a term, and it is enough to spot a new one climbing.
2. Fortnightly, run the 60 day version and apply negatives. This is where the decisions get made, because the click counts finally mean something.
3. Monthly, read the negative list itself. Lists accumulate entries that made sense against an older catalogue, and one blocking a product you now sell is an invisible cost.
4. Quarterly, check the discovery campaigns are still discovering. When one stops surfacing anything new, its budget does more good elsewhere.

What we would do first if this were our account

Work out the break-even clicks per order before opening the report, because without it every judgement is a feeling, and feelings negative too aggressively after a slow week.

Then pull 60 days, sort by spend, and read the top of the list only. A first pass over the twenty highest spending zero order terms usually accounts for most of the recoverable money.

After that the question becomes whether the terms that do convert are pointed at pages that deserve them, because a good term on a weak listing looks identical to a bad term in the report. That is why our team runs the listing side alongside the campaign side rather than after it.

We cannot promise a specific saving, because we would have to read your account first, and anyone quoting a figure before that is guessing with your money. What we can tell you, before you spend anything, is where the spend is leaking and whether the cause is the term list, the structure or the listing. That is what the free, no-obligation audit covers, and if your campaigns are tight already, we will say so.

Related guides

• Ads getting clicks but no sales
• Reading TACOS instead of ACOS
• Title limits, banned wording and truncation

Common questions about Amazon negative keywords