Seen by everyone, bought by almost nobody

The brand's whole plan was to be seen more. It bid harder every quarter, it added keywords, and on its twenty most valuable terms it had built an impression share of 24% — roughly one in four shoppers searching those words saw the brand somewhere on the page.

Those same terms returned a purchase share of 6%. That gap is not a visibility problem and no amount of extra budget closes it; it means shoppers were looking at the brand and choosing something else. And because the leak was different on every term — some lost people at the click, some lost them in the cart, one lost them because it had been out of stock for three weeks — the account's single blended ACoS number hid all of it behind one average that looked acceptable.

We stopped treating keywords as one pool and read each one as its own funnel.

• A per-term funnel — impressions, clicks, cart adds and purchases for every top term, so each leak is located rather than averaged away.
• Click leaks fixed at the search result — the terms where the brand was seen and not clicked got title, price-point and main-image work, since that is all a shopper sees at that stage.
• Cart leaks fixed on the page — the terms with strong clicks and weak conversion got the content the query implied, usually an unanswered question about age, contents or delivery.
• Terms abandoned honestly — four head terms where the product genuinely is not the best answer were dropped, and their budget moved to terms with a winnable gap.

Purchase share on the twenty terms that matter went from 6% to 19%, and revenue grew 52% on an advertising budget that rose 4%.

Every point of that came from converting attention the brand was already paying for. The diagnosis was the whole job: the account had been buying more impressions for two years to fix a problem that was happening after the impression, and a blended ACoS could never have shown which of twenty terms was broken or how.