What are Negative Keywords?

Negative keywords are exclusions you add to a campaign or ad group so that matching shopper searches no longer trigger your ad there. They are how you stop paying for clicks from people who want a different product. Used carelessly, they are also how you quietly switch off traffic that was selling.

Definition

A negative works forwards only. It stops future impressions on matching queries in the scope you choose. It does not refund spend already made, and it does not remove history from your reports.

Two choices define every negative:

Negative exact is the precise tool and negative phrase is the leverage tool, because one phrase entry also blocks queries you have not seen yet. Product targeting has its own equivalent, the negative product target, which stops an ad serving against a specific ASIN.

The only good reason to negate is irrelevance: the shopper wanted something you do not sell. A term that produces orders is working. If it converts at a poor ACoS, the bid is wrong, not the term.

The formula: when zero orders becomes evidence

Some irrelevance is obvious from the words alone. The rest shows up as clicks without orders. The question is how many clicks it takes before zero orders stops being bad luck. That depends on how often your own campaign converts, so the threshold is something you derive, not a number to borrow.

Chance of 0 orders = (1 − CVR) ^ clicks
Clicks needed      = ln(0.10) ÷ ln(1 − CVR)

CVR is the conversion rate of the campaign the term ran in, over the same window. The second line gives the click count at which zero orders would happen by chance only one time in ten.

A single click count applied to every campaign is wrong in both directions. At a 5% conversion rate, ten clicks without an order happens by chance about 60% of the time.

Worked example

Suppose you sell adult cycling gloves, the campaign converts at 10% of clicks, the average click costs $0.80, and the average order is $25. All of these are example figures. The search term report shows:

• kids cycling gloves. 0.9 ^ 25 = 7.2%, so zero orders is unlikely to be chance, and the words confirm it: a child is not your customer. Negate it.
• winter cycling gloves. 0.9 ^ 8 = 43%. Eight clicks say nothing yet. Wait.
• padded cycling gloves. Two orders make $50 of sales on $24 of spend, an ACoS of 48%. If your break-even ACoS is 40% and your target 30%, the bid is too high. Revenue per click is $50 ÷ 30 = $1.67, so a bid near $1.67 × 0.30 = $0.50 fits the target. Negating it would delete the sales along with the spend.

For the first term, the match type follows the reason. If every query containing "kids" is wrong for an adult product, a negative phrase on "kids" removes the whole theme. If only this one query is wrong, use negative exact.

How to use negatives to make decisions

1. Find themes before terms. Waste tends to spread across many cheap queries that share one wrong word. Group the search term report by single words, then pairs, and total spend and orders per group.
2. Ask the relevance question first. Is this shopper looking for a different product? If the answer is no, it is not a negative, whatever the numbers say.
3. Check zero-order terms against your own threshold from the formula above, and leave out clicks from the last seven days on Sponsored Products, because attributed orders for those clicks can still arrive.
4. Choose the narrowest fix that works. Ad group scope and negative exact first. Widen to phrase or to the campaign only when every query containing the words is wrong everywhere in that scope.
5. Repeat on a cadence. The first pass on a newly managed account usually finds the most. After that, review monthly at first, then quarterly, plus an extra pass a few weeks after a launch or a large bid change, because both change which queries you match.

Common mistakes

• Negating on performance. A converting term with a poor ACoS needs a lower bid. Check it against your break-even with the break-even ACoS calculator before reaching for a negative.
• Judging too early. One or two clicks without an order is noise at almost any conversion rate.
• Phrase on an ambiguous word. "Gift" can be wrong in one query and your best customer in another. Exclude a reliably wrong pair, or exclude at exact.
• Campaign scope in a mixed campaign. A word that is waste for one ad group can be core business for another.
• Conflicts. A negative that matches a keyword you bid on in the same ad group switches that keyword off, and you keep managing a target that can no longer serve.
• Negating a harvested term at its source by default. The source campaign often still wins some of those clicks more cheaply. See keyword harvesting.

How negatives relate to adjacent metrics

Negatives move ACoS only through spend: removing zero-order clicks lowers spend without lowering sales, so ACoS falls and conversion rate rises. Removing a term that converts lowers both, which is why the two cases need different tools. Bid changes are covered in our bid optimization guide. Your own brand terms are never waste, though keeping them out of non-brand campaigns is a structure choice covered under search term isolation. Harvesting is the mirror image of this work: the same report, read from the other end.

How Shurq helps

In Shurq's Ad Manager, the Search Terms view marks each term as targeted, harvested or already negated. Add Negatives lets you pick ad group or campaign scope and negative exact, negative phrase or negative product target, then shows a preview before anything is sent to Amazon.

N-gram analysis splits selected search terms into word groups of the length you choose and totals spend and orders per group, so a wrong theme shows up as one row. AI suggestions apply fixed checks first: fewer than ten clicks is marked too early, and your own brand, harvested terms and terms already negated are skipped. The model then judges only whether the shopper wanted a different product, and never sees spend or ACoS. Suggestions default to ad group negative exact, and you widen to phrase when you mean it.

Ask Shurq AI to find negative keywords and it judges zero-order terms with the test above, at each campaign's own conversion rate. It never proposes your own product terms, a term that converts anywhere, a negative that already exists, or one that would switch off an exact keyword in the same ad group. Campaign Map flags any enabled negative that blocks an enabled keyword inside its scope.