Negative keyword strategies to reduce high ACoS work on exactly one of the three things that cause high
ACoS. The other two they cannot touch. So before building a negative list, spend ten minutes
establishing which problem you actually have, because negating your way through a conversion problem burns
a quarter.
The three causes, and what fixes each:
| Cause | What it looks like | What fixes it |
|---|---|---|
| Irrelevant traffic | Clicks from searches your product does not serve | Negative keywords |
| Poor conversion | Relevant clicks that do not buy | The listing, or the price |
| Bids above break-even | Relevant clicks that convert, at a cost you cannot afford | Bid math from your margin |
The diagnostic is simple. Open your search term report and read the terms that spent the most. If they
describe your product, negatives are not your answer. If they describe something else, they are. I’d do that before reading further.
In our experience roughly half the accounts that arrive convinced they need a negative keyword project
actually have a conversion problem, and negating traffic in that situation reduces spend and sales
together while leaving ACoS unchanged.
First, set a target ACoS from your margin
Not from a benchmark, and this is where most ACoS conversations go wrong immediately.
Your break-even ACoS is your contribution margin as a percentage of price. If a product carries a 30%
margin after all costs, advertising at 30% ACoS breaks even. Above that you are buying sales at a loss, and in our experience nobody has calculated it before we ask.
An earlier version of this page said most sellers aim for 15 to 25%. That figure is meaningless without
a margin attached. A 45% ACoS is profitable on a high-margin product and ruinous on a thin one, and I’d
throw out any benchmark that arrives without your own numbers.
Decide what you are buying. Break-even ACoS is the ceiling for profit-focused advertising. You may
deliberately exceed it during a launch to buy rank, which I’d call a decision rather than a mistake, provided
somebody made it on purpose and wrote it down.
Then convert it to a bid. Break-even cost per click is margin per unit multiplied by conversion rate.
Our guide to Amazon bid types and strategies
covers the arithmetic properly.
The diagnostic, in ten minutes
Four steps. It decides everything that follows.
1. Pull the search term report for the last 60 days, sorted by spend descending. I’d not use a shorter window.
2. Read the top 20 terms and mark each one. Does it describe your product, or something adjacent, or
something unrelated? Nothing more sophisticated than that. This single pass tells you more than any
metric, and I’d do it by hand rather than by rule.
3. Count the spend in each bucket. If most of your spend sits on terms that genuinely describe your
product, stop reading about negatives. Your problem is conversion or bids, and in my experience that is where it usually is.
4. Check conversion on the relevant terms. Clicks with no sales on a term that describes your product
exactly is a listing signal, not a targeting signal. Our guide to
improving your product detail page covers
diagnosing which element loses the sale.
When I have run this on accounts with a stated ACoS problem, the split has been genuinely mixed, and
the sellers who assumed irrelevant traffic without checking were wrong about as often as they were right.
When negatives genuinely are the answer
Three patterns. If you see these, negate.
Wrong product entirely. You sell a dog bed and you are paying for “dog bed stairs”. Different product,
same words. The commonest genuine case, and broad match is usually the source.
Wrong intent qualifier. “Free”, “cheap”, “used”, “wholesale”, “DIY”, “how to make”. The shopper is not
buying a new one from you at your price, and I’d negate the whole family at once.
Wrong attribute. You sell the 12-inch and you are paying for “18 inch”. Specific, high-volume, and
entirely wasted. I’d hunt these first because they are unambiguous.
A pattern worth naming beyond individual terms. Look at the words that recur across your worst terms
rather than the terms themselves. If “for cats” appears in nine losing search terms, negate the phrase
once rather than the nine terms separately. That is n-gram thinking, and it scales in a way term-by-term
negation does not.
The thresholds I would actually use
Because “negate the losers” is not an instruction.
Ten clicks minimum before any judgment. Under that, zero sales means nothing. A keyword paused on
three clicks is a keyword you never tested, and in our experience impatient negation quietly shrinks
accounts.
Then compare spend against break-even, not against a fixed dollar figure. An earlier version of this
page suggested negating after $20 to $50 with no sale. That number has no relationship to your product. If
your break-even cost per click is $0.40, ten clicks with no sale is already twice your acceptable
acquisition cost. If it is $3.00, the same ten clicks are within tolerance.
Judge relevance separately from performance. An irrelevant term should be negated on the first click,
because no amount of data will make it relevant. A relevant term needs data before you touch it, and in our experience conflating the two is the root error.
Review monthly, not weekly. Search term data accumulates slowly, and I’d rather act once on four weeks
than four times on noise.
Where to add them, and at which level
Placement matters more than sellers expect.
Campaign level for anything universally wrong. “Free”, “used”, a competitor’s product type. It applies
everywhere and belongs once. I’d keep a standing list to paste into every new campaign.
Ad group level for anything specific to that product. The wrong size, the wrong color, the wrong
variation.
Negative exact for a single term you have judged. Precise and safe.
Negative phrase for a pattern. This is where the n-gram work pays, and I’d use it more than most
accounts do.
Negate harvested terms in the source campaign. When a term graduates from an auto or broad campaign
into its own exact-match group, add it as a negative in the campaign it came from. Otherwise you bid
against yourself, and if somebody managing your account does not mention this step, they have not run a
clean account. Our guide to
optimizing negative keywords covers the
mechanics in full, and
finding negative keywords covers the
discovery workflow.
Mistakes that make ACoS worse
Five, and the first two are common enough to be worth naming loudly.
Negating relevant terms because they had a bad month. You remove converting traffic and ACoS stays
flat while sales fall. The worst outcome available here and it looks like diligence.
Negating so broadly that you block variations you want. A negative phrase is a blunt instrument. I’d
read it aloud and ask what else it catches, because when I have seen this go wrong it took a month to notice.
Negating instead of fixing bids. If a term is relevant and converts but costs too much, lower the bid.
Do not remove the traffic.
Negating instead of fixing the listing. Same logic. If relevant traffic does not convert, the listing
is the problem and advertising was only exposing it.
Building the list once and never revisiting. Search behavior shifts, and a list built last year is
blocking terms that may now be worth having. I’d re-read the negatives annually.
What to do if negatives are not your problem
Because for half of you they are not.
If relevant traffic does not convert: fix the listing first. Main image, then the specification
shoppers came for. Advertising amplifies whatever conversion rate you have.
If relevant traffic converts but costs too much: lower bids to break-even and accept less volume. Our
guide to
Amazon PPC campaign structure covers the structure
that makes bids readable per product.
If the product itself is uncompetitive on price: no advertising decision fixes that, and I’d rather say
so plainly than sell you a keyword project. In our experience that conversation saves more money than any optimization.
If ACoS is high because you are launching: that may be correct. Buying rank costs money on purpose.
If you would rather have the diagnosis done 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
Will negative keywords reduce my ACoS?
Only if irrelevant traffic is causing it. High ACoS has three causes: irrelevant clicks, poor conversion on
relevant clicks, and bids set above break-even. Negative keywords fix the first and cannot touch the other
two, so diagnose before building a list.
What is a good ACoS on Amazon?
Whatever sits below your break-even, which is your contribution margin as a percentage of price. A 45%
ACoS is profitable on a high-margin product and ruinous on a thin one, so published benchmarks such as 15
to 25% are meaningless without your own margin attached.
How do I know if a search term should be negated?
Judge relevance and performance separately. An irrelevant term should be negated on the first click since
no data will make it relevant. A relevant term needs at least ten clicks before you judge it, then compare
its cost against your break-even cost per click rather than a fixed dollar threshold.
How many clicks before pausing a keyword?
At least ten. Below that, zero sales is statistically meaningless, and pausing on two or three clicks means
you never tested the keyword. Impatient negation shrinks accounts while looking like careful management.
Should negative keywords go at campaign or ad group level?
Campaign level for anything universally wrong, such as “free”, “used”, or an unrelated product type. Ad
group level for anything specific to that product, such as the wrong size or color. Also negate harvested
terms in the campaign they came from, or you bid against yourself.
What is n-gram analysis for negative keywords?
Looking at the words that recur across your worst-performing search terms rather than at the terms
individually. If “for cats” appears in nine losing terms, one negative phrase handles all nine, which
scales far better than negating each term separately.
Last updated: August 31, 2026. Amazon’s advertising console, match types, and reporting change over time;
the Amazon Ads console carries the current options for your account. An earlier version of this page
quoted a 15 to 25% target ACoS and a $20 to $50 spend threshold for negating a keyword. Both are
meaningless without your own margin and conversion rate, and both have been replaced with the arithmetic
that actually applies.
