Broad, phrase, and exact are not three levels of "good keywords." They are ways to control what a Sponsored Products keyword can match. The useful question is what your next click should teach you—and how much you can afford to learn.
This is a weekly review method for sellers with access to their own Amazon Ads reports. It is not a substitute for account-specific profitability data.
Start with a product truth check
Before a bid decision, write down the actual size, material, function, pack count, and unsupported claims. A query for "waterproof packing cubes" should not be pursued by a product without a substantiated waterproof claim, regardless of apparent search demand.
| Evidence | Starting action | What to review next |
|---|---|---|
| Relevant search term with mature, affordable orders | Test an exact target with its own cost limit | Does it receive traffic and remain affordable? |
| Relevant phrase with multiple useful modifiers | Test phrase and inspect each resulting query | Which modifiers merit separate control? |
| Customer vocabulary is unclear and a test budget exists | Use bounded broad discovery | Which actual queries fit the product? |
| Query conflicts with the product's essential feature | Exclude the mismatch | Does the negative also block valuable traffic? |
Amazon's match-type documentation describes these controls for Sponsored Products. Even exact match permits close variations, so the search term report—not the keyword label—is the evidence of traffic actually received.
Separate discovery from scaling
Export the search term report with clicks, orders, spend, and attributed sales. Let recent clicks pass the relevant attribution window before declaring a zero-order term a failure. Calculate CPC as spend/clicks and ACoS as spend/attributed sales; when attributed sales are zero, ACoS is N/A, not zero.
Next, derive a cost ceiling from the product's own contribution margin. If a $30 product has $12 left after included variable costs before ads, the illustrative break-even ACoS is 40%. A target below that ceiling leaves room for uncertainty and excluded costs. This is arithmetic for a hypothetical product, not a recommended universal bid.
Keep discovery spend visibly separate from established targets. One low-cost order is a lead for another bounded test, not proof that a new keyword is repeatably profitable.
Promote a term without cutting off its source
When a search term has relevant and sufficiently mature account evidence:
- Add it as an exact target with a bid and budget aligned to the cost ceiling.
- Confirm the new target is eligible and actually receiving impressions and clicks.
- Only then consider a source-campaign negative if overlap impairs control.
- Record the change date, placement/bidding changes, stock, price, and next review date.
Adding exact does not automatically route all eligible traffic to that target. Check actual delivery before blocking broad or phrase traffic. A negative phrase has wider consequences than a negative exact; inspect the queries it could remove.
For research outside your own Ads account, Nexscope Amazon data APIs can provide structured keyword and competitor context. Keep external demand signals separate from your account's actual clicks, orders, and costs.
Next step: Explore Nexscope Amazon Data API →
Originally adapted from How to Choose Amazon Keyword Match Types for Better Ads. Disclosure: This adaptation was drafted and fact-checked with AI tools; it does not claim live seller results.
Top comments (0)