To find negative keywords for Amazon FBA, one report does almost all the work: the search term
report, which shows the actual phrases shoppers typed before clicking your ads. Everything else on this
page is a way of reading it faster.
The mistake that wastes the exercise: sorting by ACoS. Sort by spend. A term that burned real money
matters; a term with one click and no sale is noise, and a list built from noise takes hours and changes
nothing.
And one rule before you negate anything: relevance and performance are judged differently. An
irrelevant term goes on the first click. A relevant term needs at least ten clicks before you touch it.
The workflow, in about thirty minutes
Six steps, and step four is the one that saves the most time.
1. Pull the search term report from the Amazon Ads console, last 60 days, all campaigns. Not the
keyword report. The search term report shows what shoppers actually typed, which is the whole point.
2. Sort by spend, descending. Read the top 30 rows. That is where your money actually went, and I’d ignore everything below it on the first pass.
3. Mark each one: relevant, adjacent, or wrong. By hand, by judgment. No metric can do this because
only you know what your product is. In our experience this pass takes fifteen minutes and produces most
of the value.
4. Extract the recurring words, not just the terms. This is the shortcut. If “for cats” appears across
nine losing terms, one negative phrase handles all nine. Look at the vocabulary rather than the strings,
and I’d do this before writing a single negative.
5. Check the relevant terms for conversion before negating anything on that side. A relevant term with
clicks and no sales is a listing problem wearing a keyword costume.
6. Add the negatives at the right level and re-check in four weeks. Our guide to
optimizing negative keywords covers exact
versus phrase and where each belongs.
Reading the report without drowning in it
The search term report for a mature account runs to thousands of rows, which is why most sellers open it
once and never again. Three habits make it tractable.
Filter to spend above a floor before you start. Anything that has consumed less than a few times your
break-even cost per click is not yet evidence. Filtering it out can shrink a five-thousand-row export to
perhaps eighty rows worth reading, and I’d set that filter before the file finishes downloading.
Work campaign by campaign, not across the whole account. A term that is irrelevant in one product’s
campaign may be perfectly relevant in another’s. Judging in aggregate produces negatives that damage a
campaign you were not looking at. In our experience this is the error that makes people distrust the
exercise.
Keep a running vocabulary sheet rather than a term list. Two columns: words that qualify a buyer in,
words that qualify them out. After a few months that sheet becomes the most valuable artifact in the
account, because it transfers to every new product you launch. I’d maintain it deliberately rather than
rediscovering the same qualifiers every quarter.
One measurement worth taking before and after. Note total spend on terms you marked wrong. That figure,
not ACoS, is what your negative keyword work is actually reducing. When I have shown sellers that number,
it reframed how seriously they treated the exercise.
The five sources, ranked by what they actually give you
Not all equally useful, and one is nearly worthless.
Your own search term report. The only source that reports reality. Free, specific to your product,
already in your account. This is 80% of the job and I’d exhaust it before anything else.
Your auto campaigns. Deliberately broad, so they surface mismatches quickly. Worth operating one
permanently on a modest budget purely as a discovery instrument, in our experience, even in a thoroughly
mature account.
Amazon autocomplete. Type your seed term and read the suggestions. You will encounter qualifiers you do not
serve: “for kids”, “under $20”, “refurbished”. Cheap pre-emptive negatives, available before you have spent a penny, and I’d harvest them during setup.
Your own reviews and returns. Customers who bought the wrong thing tell you what they thought they
were buying. I’ve found this produces negatives nothing else surfaces.
Third-party tools. Useful at catalog scale for pattern-finding across many campaigns. They cannot
tell you what is irrelevant to your product, which is the actual judgment, so I’d treat them as a
speed-up rather than a decision-maker. We name none and publish no ratings; pricing and features change
constantly.
What a wasted term actually looks like
Four shapes. Recognizing them is most of the skill.
Wrong product, same words. You sell a dog bed, you pay for “dog bed stairs”. Broad match is usually
the culprit. The commonest genuine case by a distance.
Wrong attribute. You sell the 12-inch, you pay for “18 inch”. Specific, high volume, entirely wasted,
and unambiguous. I’d negate these first.
Wrong intent. “Free”, “cheap”, “used”, “wholesale”, “DIY”, “how to make”, “review”, “instructions”.
The shopper is not buying a new one at your price.
Wrong audience. “Commercial”, “professional”, “for salons” when you supply a domestic version. Or precisely the
reverse, which I’d say is equally frequent and slightly harder to spot.
A shape that is not wasted, and gets negated anyway: a relevant term having a bad month. Removing it
cuts sales along with spend and leaves ACoS where it was. In our experience this is the most damaging
mistake in negative keyword work, precisely because it feels productive.
The thresholds
Concrete numbers, because “negate the losers” is not an instruction.
Ten clicks minimum on a relevant term before any judgment. Under that, zero sales is statistically
meaningless.
One click on an irrelevant term is enough. No amount of data will make “dog bed stairs” relevant to a
dog bed.
Then compare against break-even cost per click, not a fixed dollar figure. Break-even is your margin
per unit times your conversion rate. If yours is $0.40, ten clicks with no sale is already double your
acceptable cost. If it is $3.00, the same ten clicks are fine. A universal dollar threshold is
meaningless across products, and I’d distrust any guide that offers one.
Monthly cadence. Search term data accumulates slowly. I’d intervene once on four weeks of evidence rather than four
times on weekly noise, and in our experience the monthly discipline outperforms the attentive alternative.
Pre-emptive negatives, before you spend anything
The cheapest negatives are the ones added on day one.
Competitor brand names, unless you are deliberately conquesting. Expensive traffic with poor
conversion, and if you are not doing it on purpose you are doing it by accident.
Price qualifiers below your price. “Under $20” if you sell at $45.
Condition words you do not offer: “used”, “refurbished”, “open box”.
Instructional intent: “how to”, “DIY”, “tutorial”, “instructions”. Somebody researching rather than
buying.
Adjacent products you do not stock but which share vocabulary with yours. I’d enumerate 5 before launching, and it
occupies roughly two minutes.
A caution on being too clever here. Pre-emptive negatives are guesses. Add the obvious ones, then let
the search term report correct you rather than blocking traffic you never tested.
When to stop looking
Worth saying, because this work can expand to fill any amount of time.
When the top 30 terms by spend are all relevant. You are done. The remaining waste is smaller than the
time cost of finding it.
When your negatives outnumber your keywords several times over. That usually signals a structure
problem rather than a targeting one. Our guide to
Amazon PPC campaign structure covers the shape that
prevents it.
When the problem is not actually traffic. If relevant terms are not converting, negatives are the
wrong tool entirely. Our guide to
negative keyword strategies for high ACoS
covers that diagnosis, and it matters more than the discovery work on this page.
If you would rather have the search term reports read and the account managed against margin, that sits
inside our Amazon PPC service. We are an Amazon
Ads partner and an Amazon SPN Verified Partner.
FAQ
How do I find negative keywords for Amazon FBA?
Pull the search term report from the Amazon Ads console for the last 60 days, sort by spend descending, and
read the top 30 terms, marking each as relevant, adjacent, or wrong. Then extract the words that recur
across the wrong ones, since one negative phrase can replace nine individual negatives.
Should I sort the search term report by ACoS or by spend?
By spend. A term that burned real money is worth acting on, whereas a term with one click and no sale is
statistical noise. Sorting by ACoS surfaces mostly noise and produces a long list that changes nothing.
How many clicks before I negate a keyword?
At least ten on a term that is genuinely relevant, because zero sales below that is meaningless. An
irrelevant term can be negated on the first click, since no amount of data will make it relevant to your
product.
What is the best source of negative keywords?
Your own search term report, which reports what shoppers actually typed rather than an estimate. Auto
campaigns surface mismatches fastest, Amazon autocomplete gives cheap pre-emptive negatives, and your
reviews and returns reveal what buyers thought they were purchasing.
What negative keywords should I add before launching?
Competitor brand names unless you are deliberately conquesting, price qualifiers below your price,
condition words such as “used” or “refurbished”, instructional intent such as “how to” and “DIY”, and
adjacent products that share your vocabulary. Keep it to the obvious ones and let the data correct you.
Can a tool find negative keywords for me?
A tool can find patterns across many campaigns faster than you can, which matters at catalog scale. It
cannot decide what is irrelevant to your specific product, which is the actual judgment, so treat it as a
speed-up rather than a decision-maker.
Last updated: August 31, 2026. Amazon’s advertising reports, match types, and console layout change over
time; the Amazon Ads console carries the current version for your account. An earlier version of this page
recommended a fixed dollar threshold for negating a keyword and listed specific third-party tools; the
threshold has been replaced with break-even arithmetic and the tool recommendations removed.
