Phrase Match
Phrase match is the middle Amazon PPC keyword match type. It triggers when the shopper's query contains the keyword's tokens in the same order, with additional words allowed before or after. Useful where word order carries semantic meaning.
Phrase match sits between Broad and Exact in Amazon's match-type hierarchy. A phrase-match keyword triggers when the shopper's search query contains the keyword's tokens in the same order, with additional words allowed before and after (and the usual close-variant tolerance for plurals and typos).
Bid keyword: stainless steel water bottle. Will match:
best stainless steel water bottle 1L(extras around)stainless steel water bottles for kids(extras after, plural variant)
Will not match:
water bottle stainless steel(order changed)steel water bottle(token removed)
Where phrase actually earns its keep
Most modern Amazon accounts run a heavy broad/exact split with phrase as a thin bridge layer. The cases where phrase outperforms either alternative:
- Compound keywords where word order carries semantic meaning.
women's running shoes≠shoes for women running. Phrase match preserves the order constraint without paying the broad-match expansion premium. - Brand + product modifier combinations.
nike air maxshould not matchmax nike air(which is gibberish). Phrase keeps the brand-modifier-product structure intact. - Long-tail intent capture without exact restriction.
wireless earbuds for sleepingas phrase capturesbest wireless earbuds for sleeping on flightswithout the broad-match risk of matchingwireless headphones for music.
CVR profile
Phrase-match CVR typically lands at 70–85% of the equivalent exact-match CVR — closer to exact than to broad. The bid follows the same math:
Phrase Bid = ASP × CVR_phrase × Target ACOS ≈ 70–80% of exact bid
When phrase is the wrong tool
Two cases where running phrase adds operational complexity without measurable upside:
- Single-token keywords.
headphonesas phrase match is functionally identical toheadphonesas broad match — the order constraint is vacuous when there's only one token. - Categories where Amazon's close-variant matching is loose. In some categories the platform's relevance model effectively expands phrase match into broad match, defeating the purpose.
For most accounts, the rule of thumb is: skip phrase by default, add it when a specific compound keyword needs the order constraint.
Common mistakes
- Running phrase as the primary match type. Some templates default to phrase-only campaigns. The result is mid-range CVR with neither the discovery breadth of broad nor the precision of exact.
- Bidding the same on phrase and exact. Phrase CVR is lower; the bid should be ~20–30% lower.
- No negative-exact on the phrase parent for terms harvested into exact. Same leak as broad — the phrase parent keeps bidding on the harvested winner.
- Mixing phrase keywords into exact ad groups. Destroys ad-group CVR consistency. Keep one match type per ad group.