How to Use Amazon Ads Search Term Reports

Amazon describes the Sponsored Products search term report as a view of the search terms customers used to look for products on Amazon, including terms inferred as a best match in some non-search contexts. The report is available to supported Sponsored Products and Sponsored Brands advertisers. It is query evidence from clicked ads, not a complete record of every impression or every organic search that occurred during the period.

Amazon notes that the report includes search terms that produced at least one ad click. That means report impressions may not match the impressions displayed in campaign manager. Preserve that boundary when reconciling totals. A missing term does not prove that no shopper used it, and a row in the report does not prove that the term is profitable or strategically important without the campaign, product, cost, and attribution context.

Export a stable period and retain the account, marketplace, campaign, ad group, advertised product, target, match type, search term, clicks, spend, attributed orders, and attributed sales fields that are available. Keep the raw export unchanged. Perform normalization and decision tagging in a working copy so the team can trace every action back to the source row rather than reconstructing a decision from a later campaign state.

A search term is the shopper expression associated with the ad interaction. A keyword is an advertiser-selected target used in manual keyword targeting. Product targets and automatic targeting can also make an ad eligible. Amazon explains that Sponsored Products can use automatic and manual targeting, and it recommends using search term evidence from automatic campaigns to identify terms that may be useful in manual targeting after the campaign has collected observations.

Do not collapse these fields into a single keyword column. One target can match multiple customer queries, and one query can appear through different targeting routes. Grouping them without the originating campaign and target can hide whether performance came from a broad discovery campaign, a controlled exact target, or a product target. Preserve the relationship so a new manual target does not accidentally duplicate spend across several campaigns.

Normalize only what is necessary for review. Trim spacing, standardize case in the analysis layer, and tag obvious brand, category, competitor, use-case, and irrelevant patterns. Keep the exact reported query beside the normalized value. The exact text remains necessary for negative match decisions, auditability, and later comparisons because phrase and exact controls do not behave like an analyst's broad topic label.

Review relevance first. Confirm that the query describes a need the advertised product can satisfy and that the product detail page supports the promise implied by the term. Then review the campaign objective and the observed metrics. A term intended for discovery should not be judged only by the same threshold as a mature efficiency campaign, but the team should still document what outcome would justify keeping or promoting it.

When a relevant term shows repeatable evidence that fits the objective, consider adding it to a manual campaign or ad group with a clear product and bid boundary. Amazon says search term information in campaign manager can be copied into new ad groups to turn performance insights into targeting expressions. Before copying, check whether the term already exists elsewhere and whether the new target would create internal competition or make ownership harder to interpret.

When a term is irrelevant, or when sufficient campaign evidence shows that it does not meet the objective, consider negative keyword or product targeting. Do not delete the source row after taking action. Record the action date, level, match type, reason, owner, and affected campaigns. That decision log lets the next review distinguish a real demand change from a campaign that simply stopped entering certain auctions.

After a targeting change, allow the campaign to collect new observations before judging it. Compare a clean post-change period with the decision record, while accounting for product availability, price, promotions, budget limits, placement changes, and other campaign edits. If several controls changed at once, state that the result cannot be assigned confidently to the search-term action alone.

Build a recurring review with distinct queues for discovery, promotion, exclusion, and watch. A discovery row needs more evidence. A promotion row has a supported manual-targeting action. An exclusion row has an explicit relevance or performance reason. A watch row stays unchanged because the evidence is mixed. Require every row to have one current state so the same term is not promoted by one operator and excluded by another.

Close the review by reconciling source rows, actions, and current campaign state. Confirm that the intended target or negative exists at the documented campaign or ad-group level and that no duplicate action was created. Retain rejected actions and reversals. Search behavior, inventory, product detail pages, and campaign goals change, so the durable asset is the evidence trail and review discipline, not a permanent list of good and bad terms.