The same product competing in two different markets at once

The same box of craft materials is bought by a parent looking for an activity and by someone stocking a classroom or a childcare room. Those are not two segments of one market; they are two markets. The first compares the product against toys, buys one, and is influenced by how fun it looks. The second compares it against bulk supplies, buys eight, and cares about quantity and washability.

Everything ran through one campaign structure with one set of bids and one creative. The result was a blended ACoS of 44% that described neither audience: bids too high for the low-value toy queries and too low to win the supply ones, and 31% of spend going to queries the listing was not written for.

We split one market back into the two it had always been.

• Query classification — every search term coded as toy intent, supply intent or ambiguous, then reconciled against what each actually ordered.
• Two campaign structures with separate budgets and bid logic, so the value of a classroom order is not averaged against a single impulse buy.
• Two creative treatments — the activity and the outcome for the toy pool; quantity, durability and cleanup for the supply pool.
• Negatives run in both directions, so each structure stops paying for the other's traffic.

Blended ACoS halved from 44% to 22%, and revenue grew 104% to $180k a month.

The supply pool did most of the work — it nearly quadrupled, because it had been systematically underbid by an average that the much larger toy-query volume was setting. Wasted spend on mismatched queries fell from 31% to 7%. Neither product nor price changed; the account simply stopped pretending that two audiences were one.