The search term report: where your ad spend goes, and when to negate a search

The keyword you bid on and the search a buyer clicked are two different things. Broad and phrase match, and automatic campaigns, show your ad on many different searches. The search term report lists, row by row, exactly what the buyer typed (or which product page the ad appeared on), what you spent on it, and whether it sold.

Every claim below points at a number in that report. We checked them on 493 Sponsored Products campaigns in 12 advertising accounts, on European and US marketplaces, with daily data from March 28 to October 3, 2026. That is 42,657 daily rows.

The report is what you paid for

Only 5 of the 42,657 rows had no click. So in practice the report does not show every place you appeared, only the ones where someone clicked, which means what you paid for.

Counted per ad group, 18,457 different searches appear in the report. Of those, 16,923 never produced a sale over the whole period, and 12,679 of them got exactly one click.

The money goes into the long tail

For each account, we looked at how much of the spend on searches went to searches that never sold. In the median account it was 65%. The best account was at 24%, the worst at 99%.

The surprise is how little of that money sits in the obvious searches. Searches with at least 10 clicks and no sale took 10% of the spend in the median account. The rest of the no-sale spend is spread over thousands of searches with fewer than 10 clicks, and 14,747 of those 16,562 searches got one or two clicks. You cannot judge those one at a time: a single click tells you nothing.

Sales, meanwhile, concentrate in a few searches. In the six accounts where at least 100 searches sold, the top 5% of searches by sales value brought at least 66% of the sales value.

The click-count rule misleads

A common rule says to negate a search after 10 clicks without a sale. In our data, 496 searches reached 10 clicks with no sale. Of those, 135, or 27%, sold later. Their median final ACOS was 36%, and 56 of them ended below 30%. The 10-click rule would have cut all of them off.

Twenty clicks is not a safe line either. 141 searches reached 20 clicks without a sale, and 44 sold later.

Searches that sell usually sell quickly. Among the 1,534 searches that sold, half had at most 2 clicks by the end of the day of their first sale, and 90% had at most 11. So a slow starter is rare, but not rare enough to judge safely by click count.

Judge by money, not clicks

The better measure is how much you spent without a sale, relative to the product's price. If your target ACOS is 30% and your product costs 20 euros, you can spend 6 euros per sale. Once a search has spent 6 euros without a sale, even its first sale lands at 30% ACOS or above. It can only get back under your target with further sales.

In the data, that rarely happened. We used the account's average order value as the price. 233 searches spent at least 30% of it without a sale. 73 of them sold later, but only 11 finished under 30% ACOS. At half the average order value there were 85 searches, 30 sold later, and 4 finished under 30% ACOS.

In your own account, use the product's actual price and your own break-even ACOS, not our 30%.

The long tail needs structure

The thousands of one-click searches are not controlled by negating, but by structure.

• Harvest what sells. 161 searches brought at least two sales from a broad, phrase or automatic campaign. 112 of them never got a click through an exact-match keyword in the same account. Add these as exact keywords, and add them as negative exact in the source campaign, so the two campaigns do not bid against each other.
• Negative phrase for a recurring word. If one word keeps appearing in searches that do not sell (a different product type, a different size, a competitor's brand name), one negative phrase excludes all of its variants at once.
• An ASIN row means a product page. Many rows in the report are not text but an ASIN (a code starting with B0): a product page where your ad was shown. Our data had 6,492 of them. A negative keyword does not exclude these, you need negative product targeting.

Twenty minutes a month

Download the search term report for the last 60 days and sort it by spend. Make three lists from it.

The first is harvesting: searches with at least two sales that are not yet exact keywords. The second is negating: searches that have spent, without a sale, what you allow for one sale (price times target ACOS). The third is everything else: leave it, and look again next month.