Amazon Ads

Amazon Search Term Report: The Weekly Harvest and Negate Routine

A weekly routine for the Amazon search term report: thresholds from your own numbers, a promote-or-negate decision table, brand negatives in auto, and how to stop your own campaigns competing.

Cover: Promote winners, negate waste. a search term list splitting into two arrows, one to an exact-match campaign box and one to a negatives box.

Where to get the report

  • Console: Measurement and Reporting > Reporting. Amazon made this unified reporting page generally available on 8 June 2026 and says the older Sponsored ads report pages are sunset by 31 December 2026 (Amazon), so re-create any scheduled search term reports there.
  • Bulksheets: tick "Sponsored Products search term data" when you create a custom spreadsheet. The tab includes the campaign and ad group IDs you need to act on it, but Amazon fixes its lookback at 7 days (bulksheets search term report).
  • API and tools built on it: contains only search terms with at least one click, while the console report also lists terms with impressions only, so row counts differ. Amazon keeps 65 days of Sponsored Products and 60 days of Sponsored Brands search terms in the API (search term report).

Rows where the search term is an asterisk are product page placements with no search behind them. Rows that look like a 10-character product code are ASINs: your ad showed on that product’s page, usually from auto substitutes or complements.

Set your thresholds before you open the file

Every rule below comes from three numbers you already have:

  • Allowable cost per order = average order value × target ACOS. The most you can pay for one sale.
  • Clicks per order = 1 ÷ conversion rate. How many clicks a sale normally takes.
  • Judgement window: the last 30 days, minus the most recent 3 days (starting point).

Why leave out recent days? Amazon’s reporting FAQ says conversion data is restated 1, 7 and 28 days after the conversion event. A term that "has no sales" from yesterday’s clicks may simply not have them reported yet.

Starting point, adjust to your account

Search term decision table

 RuleAction
Converts at or under target2+ orders, ACOS ≤ targetAdd as exact in the exact campaign; add negative exact in the campaign it came from
Converts above targetOrders, ACOS well above targetKeep; lower the bid on the source target, or promote to exact with a lower bid
Spends without sellingSpend ≥ allowable cost per order, or clicks ≥ 2 to 3 × clicks per order, 0 ordersNegative exact in that campaign
Irrelevant root wordSeveral wasted terms share one word (a colour, a different product type)Negative phrase on that word, after checking it never converts elsewhere
Too early to judgeFew clicks, no ordersWait for next week
ASIN that convertsProduct page row with ordersAdd as a product target in a product targeting campaign
ASIN that wastesProduct page row, spend, no ordersNegative product target
Your brand in auto or genericBrand name or spellingNegative phrase in auto and generic; make sure the brand campaign has it as exact
Competitor brandAnother brand’s nameDecide with the client; move to the competitor campaign or negate

The weekly routine

How to

Weekly Amazon search term harvest

The weekly routine an agency team runs on each Amazon profile.

⏱ About 45 minutes

  1. 01

    Pull the report

    Sponsored Products and Sponsored Brands search terms for the last 30 days, ending 3 days ago, with campaign, ad group, match type, clicks, spend, orders and sales.

  2. 02

    Add your thresholds

    Add columns for ACOS, allowable cost per order and clicks per order, and label each row with an action from the decision table.

  3. 03

    Check word patterns

    Group wasted spend by single words. One word in many losing terms is a candidate for a negative phrase.

  4. 04

    Promote winners

    Add winners as exact keywords at a bid near their cost per click, in the right exact campaign and language ad group.

  5. 05

    Negate at the source

    Add each promoted term as negative exact in the auto, broad or phrase campaign it came from, and the wasted terms as negatives.

  6. 06

    Clean brand traffic

    Make sure brand terms are negated in auto and generic campaigns and covered in brand campaigns.

  7. 07

    Log and upload

    Keep a change log with date, term, action and reason, then apply it through the console, a bulksheet or your assistant.

Avoid self-competition

Amazon will happily let two of your campaigns enter the same auction. The usual causes are a promoted term left active in auto, the same exact keyword in two campaigns, and two auto campaigns advertising the same products. The last one is the most expensive: in our experience the campaign with the higher default bid wins your own traffic, and its cost per click climbs while the cheaper campaign starves. Keep one auto campaign per product group, and give each winning term one owner.

On bilingual marketplaces, treat English, Arabic and transliterated spellings as separate keywords. Harvest each one that converts, and negate each one that doesn’t, rather than assuming one covers the others.

Where this goes wrong

  • Negating on too few clicks. A term with a handful of clicks and no order is noise. Use your clicks-per-order threshold.
  • Negating at the wrong level. A campaign-level negative applies to every ad group in it. Use ad group level when only one product is irrelevant.
  • Broad negative phrases. A negative phrase on a common word can block searches that convert elsewhere. Check the word across all campaigns first.
  • Promoting without negating. The exact keyword and the source campaign then compete, and the report can’t tell you which one deserves credit.
  • Product names that attract the wrong searches. Auto campaigns read your listing. A product named after a colour or flavour can pull searches for that colour or flavour in other categories. Negative phrase on that word.

Do it faster with Whathead

Through the Amazon Ads MCP, Claude or ChatGPT can pull the search term report, apply your thresholds and draft the changes. Use the read-only connection (https://www.whathead.com/mcp/read) for the analysis and the write connection only when you apply changes. Whathead is our product.

  • “Pull the Sponsored Products search term report for the UAE profile, last 30 days ending 3 days ago. My AOV is [X], target ACOS [Y]%, conversion rate [Z]%. Label every term with an action from my decision table and total the wasted spend.”
  • “Add the promote rows as exact keywords in the exact campaign and as negative exact in their source campaigns. Show me the preview.”

Whathead shows each keyword and negative as a dry-run preview (campaign, ad group, text, match type and bid) and sends nothing until you confirm. For a standing weekly view, the Google Sheets add-on has a Sponsored Products search terms report for Amazon Ads that can refresh on a schedule.

Related: Amazon PPC audit, campaign structure, bulksheets and brand defence.

Amazon search term report: common questions

How far back does the Amazon search term report go?

In the API, 65 days for Sponsored Products and 60 days for Sponsored Brands. The bulksheet search term tab always covers 7 days. Amazon’s unified reporting page keeps up to 15 months of daily data, so schedule a regular export if you want long history.

Why does my search term report show ASINs?

Those rows are product detail pages where your ad appeared, mostly from auto targeting or product targeting. Treat them like search terms: target the ones that convert and negate the ones that waste.

How often should I review search terms?

Weekly for accounts with steady spend, and at least monthly for small ones, before the data ages out.

What is the difference between the search term report and Search Query Performance?

The search term report covers only searches that triggered your ads. Search Query Performance, in Brand Analytics for brand owners, covers searches that led shoppers to your brand, organic and paid.

When should I add a negative keyword?

When a term has spent more than your allowable cost per order, or had two to three times your usual clicks per order, with no sales, after excluding the most recent days.

Sources

Written by Ahmed Shahin, founder of Whathead. Last reviewed 19 November 2026.