Amazon Negative Keywords: Exact Vs Phrase And When To Block A Search
Use negative exact and phrase targeting carefully, review search-term evidence, and apply a practical decision log instead of a blanket no-sales rule.
Use negative exact when you want to exclude a specific shopping query and its close variations. Use negative phrase when the entire phrase represents demand you do not want across longer queries too. Decide whether the search is irrelevant before deciding whether it is merely underperforming.
What The Two Negative Match Types Do
For Sponsored Products, Amazon describes negative exact and negative phrase matching in its keyword-match documentation. Both include close variations. Phrase negatives exclude queries containing the complete phrase; exact negatives have a narrower scope. Do not assume an exact negative blocks only one literal sequence of characters.
| Decision | Negative Exact | Negative Phrase |
|---|---|---|
| Intended scope | A particular query and close variations | A phrase within queries, including close variations |
| Useful starting situation | One query needs exclusion; neighbouring demand may still fit | The phrase consistently describes an unsuitable product or use |
| Main review question | Could this query still deserve a controlled test? | Which valuable longer queries might this also block? |
The following examples are editorial illustrations of scope, not a guarantee of how every close variant will be classified. Suppose you sell an adult stainless-steel lunch container. A negative exact entry for “kids plastic lunch box” targets that query family. A negative phrase for “plastic lunch box” would also exclude longer queries containing that phrase. A single broad idea such as “lunch” would be a poor exclusion for a lunch-container campaign because it cuts across the product’s core demand.
Start With The Search Term, Not Just The Keyword
A bid keyword is a targeting instruction; a customer query is the demand that actually matched. Amazon’s Sponsored Products search-term report guidance says the report includes terms with at least one ad click. Report impressions therefore need not match every impression in campaign management. It also notes that a term can be inferred for non-search placements.
Download the report for a consistent period and retain campaign, ad group, advertised product, query or identifier, clicks, spend, attributed orders and attributed sales. Separate products with different margins. Review the attribution window and reporting maturity before judging recent clicks as failed sales.
Keep a copy before making changes. If a campaign later loses useful traffic, you need to know which exclusion was added, where it was applied and what evidence supported it.
Worked Example: Four Queries, Four Different Decisions
This is a hypothetical analysis, not an actual account result. Assume each attributed order is one $30 item and contributes $9 before advertising. The simple contribution break-even ACOS is 30%. Different purchased products, returns or fixed costs would change the analysis.
| Query | Clicks / Spend / Orders | Interpretation | Candidate Action |
|---|---|---|---|
| kids plastic lunch box | 5 / $4 / 0 | Wrong material and intended audience | Consider a targeted negative after checking the advertised SKU |
| stainless steel lunch container | 18 / $14.40 / 2 | $60 sales; 24% ACOS; $3.60 contribution after ads | Retain and monitor against the desired profit target |
| adult lunch container | 6 / $4.80 / 0 | Relevant query, limited evidence | Review listing fit and continue a bounded test if budget permits |
| metal lunch box | 40 / $32 / 1 | $30 sales; 106.7% ACOS; −$23 contribution | Investigate bid and conversion problems; reduce exposure or exclude if justified |
The first row is a relevance decision. The last is an economics decision. Treating both as “zero sales means block” would miss the distinction—and would not even describe the last row correctly. A relevant term can be too expensive at its current bid without being permanently unsuitable.
Why A Fixed Click Threshold Is Not Enough
Amazon’s targeting guide recommends evaluating performance after at least 20 clicks before deciding on a negative. That is guidance, not proof that every relevant term with 20 clicks and no orders is worthless.
For an illustrative term with a true 10% conversion probability, the probability of zero orders in 20 independent clicks is 0.920, approximately 12.2%. Real clicks need not be independent and conversion rates are not constant, but the calculation demonstrates why small samples can mislead. Conversely, an obviously incompatible query does not become relevant because it has not reached a click threshold.
Set a tolerable testing loss using your margin and budget. This is a business decision under uncertainty, not a universal number copied from someone else’s campaign.
A Safe Review And Change Log
- Confirm which product was advertised and whether the query’s intent fits it.
- Check report dates, attribution maturity, stock availability and listing changes.
- Calculate contribution using the appropriate product costs; see our ACOS and break-even guide.
- Choose between continued testing, bid adjustment and exclusion.
- If excluding, choose the narrowest scope that solves the observed problem.
- Record the query, match type, campaign or ad-group level, date, reason and review date.
- Check subsequent relevant traffic, orders and profit—not just the reduction in spend.
Amazon permits negatives at campaign or ad-group level. Its negative-targeting instructions also state that an existing negative’s match type cannot be changed. Check the current interface before replacing entries, and verify that the replacement has the intended scope.
Questions Worth Asking Before Bulk Changes
Should I block every expensive query?
No. First determine whether the bid is too high, the listing under-explains the product, or the query genuinely does not fit. Blocking demand will not repair an unclear listing.
Should I copy one negative list to all products?
Only after checking every affected product. A phrase describing an unwanted size for one SKU may describe the exact size another SKU sells.
Does a lower ACOS prove success?
No. A campaign can show a lower ACOS after losing valuable volume. Compare retained contribution and useful sales alongside efficiency. For image-related conversion questions, use our product photography shot list.