Half the ad budget was buying grown-ups looking for their own costume

Costume search terms are a battleground between two completely different shoppers. One is a parent buying a dress-up outfit for a six-year-old; the other is an adult looking for something to wear to a party. They type nearly the same words and want entirely different products.

The brand's campaigns had grown around broad costume terms because the clicks were comparatively cheap — and they were cheap because the adult-intent traffic they attracted never converted. Nearly half of all paid clicks came from those terms, at a 3.9% conversion rate that dragged ACoS to 61%. The damage was not only financial: Amazon reads click-without-purchase as evidence that a listing does not answer the query, so the account was steadily teaching the algorithm that its own listings were a poor result for the very terms it wanted to rank on.

We cut the traffic that could not convert and reinvested in the traffic that does.

• Search-term audit by intent, separating every converting term from the adult-intent ones that had been absorbing budget.
• A negative-keyword layer blocking adult sizing, party and costume-for-men or women phrasing across every campaign.
• Child-qualified exact-match campaigns — age, size and “for kids” phrasings promoted into their own structure with their own budgets.
• Bid weight moved to occasion terms where child intent is unambiguous: school play, book day, birthday dress-up.

ACoS fell from 61% to 29% and revenue rose 75% on a smaller advertising budget than the account started with.

The organic gain matters more than the paid one. Once the listings stopped being served to shoppers who would never buy them, their conversion rate on the remaining traffic doubled, and Amazon's ranking algorithm responded exactly as expected: the main child-costume term went from #28 to #6. The cheap clicks had never been cheap; they had been paying to look irrelevant.