Blog

Amazon Search Terms: How to Mine Reports for Winning Keywords

Your customers are writing your keyword list for you — every Amazon search term in your reports is a real query a real buyer typed. This guide shows how to download your Amazon search term report , read the six columns that matter, promote winners into exact-match keywords, and cut losers with negative keywords . For deeper analysis, run the same data through our search term analyzer .

Table of contents
  1. Search term vs keyword
  2. How to download the report
  3. The 6 columns that matter
  4. Harvesting winners
  5. Negating losers
  6. How often to mine
  7. Search term FAQ
Search term report silicone spatula set heat resistant spatula cheap spatula free spatula for nonstick wooden spoon set spatula silicone 600F Winners ★ silicone spatula set ★ spatula for nonstick ★ spatula silicone 600F harvested Every converting search is a keyword you didn’t have to guess.
Mining a search term report: raw customer queries go in, proven winning keywords come out.

Amazon Search Term vs Keyword: What’s the Difference?

This distinction trips up most beginners, so let’s settle it first. A keyword is what you bid on — your targeting. An Amazon search term is what the customer actually typed into the search bar. They are rarely identical.

You bid on the keyword spatula set (broad match). Customers find you through the search terms silicone spatula set for nonstick pans , heat resistant spatula 600F , and — unfortunately — cheap plastic spatula free shipping . The search term report shows you all three. Your job is to promote the first two into keywords of their own and negate the third. That loop — bid broad, learn from real searches, refine — is the entire engine of profitable Amazon PPC .

Key takeaway: Keywords are your guesses; search terms are your customers’ answers. Mine the answers.

How to Download Your Search Term Report on Amazon

In the Amazon Ads console, go to Measurement & reporting → Sponsored ads reports , create a search term report for Sponsored Products (the same report type exists for Sponsored Brands and Sponsored Display), pick your date range, and download the spreadsheet. Seller Central’s advertising dashboard offers the same report under the campaign manager’s reporting tab.

Two practical notes: first, data has a 2–3 day attribution lag — don’t mine the most recent 3 days, or you’ll judge terms before their orders register. Second, pull at least 30 days the first time so you have enough history; weekly mining after that can use 7–14 day windows.

Anatomy of a search term report row Search term row One buyer query Customer search term What the buyer typed Match type How it triggered Clicks and spend Cost so far Orders and sales Result so far
Four columns tell the whole story — what they typed, how it matched, what it cost, what it earned.

Reading the Report: The 6 Columns That Matter

The raw export has dozens of columns. You need six:

  • Customer search term — the actual query. This is the gold.
  • Match type — which of your keywords (exact/phrase/broad/auto) caught it.
  • Clicks — your sample size. No clicks, no verdict.
  • Spend — where your money went, for prioritizing.
  • Orders — did it convert?
  • Sales — revenue, for computing the term’s ACoS.

From these you derive everything: term ACoS = spend ÷ sales, and the two verdicts that drive the whole workflow — winner (converts at or below your target ACoS) or loser (15+ clicks, zero orders). Sort by spend descending and work from the top; the first screen of rows usually holds most of the opportunity.

Harvesting Winners: Promote Converters to Exact Match

When a search term converts consistently inside a broad, phrase, or auto campaign, it has earned its own keyword. Harvesting means adding that exact search term as a new exact-match keyword — ideally in its own ad group or campaign where you control the bid directly.

The bar: at least 2–3 orders at an acceptable ACoS, not just one lucky sale. Set the opening bid from the term’s own data — its historical CPC is the most honest starting point you have. Then decide what to do with the original: many advertisers add the harvested term as a negative exact in the discovery campaign, so future impressions flow to the new exact keyword instead of competing with it.

Search term report WINNERS clicks + orders promote to exact match own ad group + bid LOSERS 15+ clicks, 0 orders add as negative keyword exact first, phrase if junk Mine weekly: harvest winners, negate losers
The mining loop: every search term becomes either a harvested exact-match keyword or a negative keyword.

Try it: Search Term Harvester Paste your report rows — it separates winners worth promoting from losers worth negating, with suggested bids. Free, in your browser.

Harvest my search terms
Harvest vs negate: making the call Do this: Harvest when Orders at target ACoS Repeat converters Exact-match worthy Not this: Negate when Spend with zero orders Irrelevant to product Repeat offenders
The call is binary and the data makes it — converters get harvested, spenders get negated.

Negating Losers: Stop Paying for Non-Buyers

The other half of mining is pruning. Any Amazon search term with roughly 15+ clicks and zero orders gets a negative keyword — exact match by default, phrase match when the whole word family is junk. This is also where you catch the structural waste: wrong-gender terms, "free"/"used"/"DIY" modifiers, and sizes you don’t sell.

Don’t stop at keywords: product-targeting campaigns accumulate losing ASIN placements the same way. Negate the ASINs that spend without converting. Our negative keyword analyzer flags every 15-click/0-order offender in seconds, so the weekly prune takes minutes.

Weekly search term mining checklist Green = done Orange = improve Red = fix now Report downloaded Done Sorted by spend Done ! Winners harvested to exact Three pending Losers negated Do today ! Decisions logged Start a log
Mining without a log is re-mining — write down every harvest and negate decision.

How Often Should You Mine Search Terms?

Weekly for active accounts — the report is a living feed, and a week of new data is plenty to act on. New launches deserve twice-weekly checks in the first month, because discovery campaigns find traffic fast and waste compounds just as fast. Mature, low-spend accounts can go biweekly .

Whatever the cadence, keep it ritual: same day, same 20 minutes, same two outcomes — winners harvested, losers negated. Sellers who mine on schedule compound small wins into dramatically lower ACoS over a quarter; sellers who “get around to it” fund Amazon’s Q4 instead of their own.

Key takeaway: Download monthly for history, mine weekly for profit. Harvest the converters, negate the bleeders, repeat.

Amazon Search Term FAQ

A keyword is what you bid on (your targeting); a search term is what the customer actually typed. You might bid on the broad keyword “spatula set” and appear for the search terms “silicone spatula set”, “heat resistant spatula”, and “cheap spatula”. The search term report reveals the real queries so you can promote converters and negate waste.

In the Amazon Ads console: Measurement & reporting → Sponsored ads reports → create a search term report for Sponsored Products, Brands, or Display. Choose your date range (mind the 2–3 day attribution lag) and download. Seller Central’s campaign manager offers the same report.

Wait for at least 2–3 orders at an acceptable ACoS — one sale can be luck. Use the term’s own historical CPC as the opening bid for the new exact keyword, and consider negating it as exact in the discovery campaign so traffic consolidates on the new keyword.

Usually not immediately. Keep the discovery keyword running so it keeps finding new terms, but add the harvested term as a negative exact there — that routes its future impressions to your new dedicated exact keyword instead of splitting data between two places.

You can pull up to roughly 60–90 days per report depending on the console. For mining purposes, 30 days of history is the sweet spot for the first pass — enough data to judge, fresh enough to act on. Remember the last 2–3 days are incomplete due to attribution lag.