Amazon Keyword Research: A Practical Workflow Without Guesswork
Where Amazon keywords actually come from, how to rank them on relevance, volume and conversion likelihood, and why the same list is deployed differently in the listing and in ads.
Amazon keyword research is a workflow, not a tool: pull terms from four sources, group them into topics, score each topic on relevance, volume and conversion likelihood, then deploy the same shortlist two different ways — once in the listing, once in the campaigns. Most brands skip the middle and go straight from a tool export into the title. The export is not the deliverable; the grouped, prioritized list is.
Amazon keyword research starts with four sources
Collect from all four sources before judging any of them — each one is blind in a different way.
1. Your own search term report (Amazon Ads console)
The Sponsored Products and Sponsored Brands search term reports show the actual queries shoppers typed that led to a click on your ad, with spend, orders and ACoS attached. This is the only source that tells you what a term is worth to your product rather than to the category.
Two limits worth knowing. It only covers paid traffic, so terms you rank for organically but never bid on are invisible here. And the console does not keep report history indefinitely — export on a fixed schedule into your own sheet, or you will rebuild the same dataset from scratch every quarter.
2. Brand Analytics (Seller Central, Brand Registry required)
Two dashboards do the heavy lifting, and they answer different questions.
Search Query Performance covers up to 1,000 of your most relevant queries and reports the full funnel per query — impressions, clicks, cart adds, purchases — plus your brand's share at each stage. It combines organic and paid, which is a feature for keyword analysis and a trap for ad reporting. The diagnostic value is in the gaps: healthy impression share with thin click share is usually a main-image or title problem, not a keyword problem.
Top Search Terms is marketplace-wide rather than brand-specific. It gives each query a Search Frequency Rank (lower number = more searched) and the click and conversion share of the top three products for that term. That combination is how you find demand you are completely absent from, which the search term report can never show you.
3. Competitor listings and the terms they rank for
Reverse-ASIN lookups on three or four genuine competitors — not the whole category, just the products a shopper would realistically cross-shop against yours. Read their titles and bullets directly too; brands often name a use case you have not thought to claim. The weekly routine for this is in Amazon competitor analysis.
Volume numbers from third-party tools (Helium 10, Jungle Scout and the rest) are modeled estimates, not Amazon's own figures. Use them for discovery and relative ranking, then let Brand Analytics arbitrate anything that matters. What each tool is genuinely good at is covered in our seller tools guide.
4. How customers actually describe the product
Reviews, answered questions, return reasons and your support inbox. This is where you find the phrasing no keyword tool surfaces because nobody types it into a search bar yet — "doesn't leak in a backpack", "for a dog that swallows pills whole". It matters more now that a share of discovery happens through conversational queries, where the assistant reads listing copy to answer a question rather than matching a phrase.
Group into topics before you score anything
A raw export is thousands of rows and perhaps eighty real topics. Collapse them: word order, plurals and connector words do not need separate slots, because Amazon indexes the words in your copy rather than the exact phrase string. "Organic dog joint supplement" and "dog joint supplement organic" are one topic.
You should end this step with a list of topics, each holding its variants, its best volume estimate and your current position — see how Amazon ranking works for the mechanics underneath.
Prioritize on three axes, not on volume
Volume alone is how brands end up ranked for a head term that never converts. Score each topic on all three:
| Axis | Read it from | Red flag |
|---|---|---|
| Relevance | Does the query describe this product, not the category? | A shopper landing here would need a different size, format or use case |
| Volume | Search Frequency Rank in Top Search Terms; tool estimates as a cross-check | Head term where the top three products own most of the click share |
| Conversion likelihood | Your own cart-add and purchase rates in Search Query Performance and in the search term report | Clicks arrive, cart adds do not |
Then sort into three tiers, because the tier decides where a term is allowed to go:
| Tier | What qualifies | Where it goes |
|---|---|---|
| Core | High relevance, proven conversion, meaningful volume | Title, first bullets, exact-match campaigns |
| Secondary | Relevant, lower volume or unproven conversion | Bullets, description, A+ copy, phrase-match campaigns |
| Discovery | Plausible but unvalidated | Backend search terms, broad and auto campaigns only |
Keep the tiers small at the top. Five to eight core topics per ASIN is a realistic ceiling; anything more and the title stops reading like a sentence a human would trust.
The same list lands differently in the listing and in the ads
This is the step most keyword projects get wrong: one list, two completely different deployment rules.
| Dimension | Listing | Campaigns |
|---|---|---|
| Goal | Get indexed, then convert the click | Buy traffic you can measure |
| Coverage | Each core topic once; repetition adds nothing | Exact for proven, phrase/broad and auto for discovery |
| Where variants go | Backend search terms — synonyms, misspellings, alternative names not in the visible copy | Broad and auto campaigns, harvested into exact as they prove out |
| Failure mode | Stuffing — indexed for everything, persuasive to nobody | Bidding on discovery terms at core-term bids |
One mechanical rule on the listing side: Amazon's Search Terms field is capped at 250 bytes in the US and the penalty is all-or-nothing — go over and none of it is indexed — so do not spend that space repeating words already in your title. Write the visible copy for the shopper first; the listing optimization checklist walks through the order we work in. This is the core of how we run Amazon listing SEO, and it compounds: a pet supplements brand grew organic sales 187% through listing work alone.
On the ads side, discovery terms are cheap experiments. Harvest what converts into exact match, negate what does not, and let the auto campaigns keep feeding the pipeline — that loop is what keeps the list alive between formal research rounds.
Refresh cadence
Rebuild fully once or twice a year, or whenever the catalog, category or a major competitor changes. In between, a weekly search term report review and a monthly Search Query Performance pull are enough: you are watching for topics where your click or purchase share is sliding, which is the earliest warning that a listing or a competitor moved.
FAQ
How many keywords should an Amazon listing target?
Five to eight core topics per ASIN, plus their variants. Amazon indexes the words in your copy, so a topic covered once in the title or bullets is indexed — repeating it adds no ranking benefit and costs readability. Secondary and unvalidated terms belong in the description, A+ copy and the backend field.
Do I need to repeat keywords in the title and the backend search terms?
No. Words already in your title, bullets or description are indexed, so repeating them in the 250-byte Search Terms field wastes the only space you have. Reserve the backend field for synonyms, alternative product names, common misspellings and secondary use cases that do not fit naturally into customer-facing copy.
How accurate is Amazon keyword search volume from third-party tools?
Treat it as a modeled estimate, useful for ranking terms against each other but not as an absolute number. Amazon's own first-party figures come from Brand Analytics — Search Query Performance for your queries and Search Frequency Rank in Top Search Terms for the marketplace. Where the two disagree, Brand Analytics wins.
How often should I redo Amazon keyword analysis?
A full rebuild once or twice a year, triggered earlier by a new product variation, a category shift or a competitor launch. In between, review the search term report weekly to harvest and negate, and pull Search Query Performance monthly to catch declining click or purchase share before it shows up in revenue.
Start here this week
- Export the last 90 days of search term data plus a Search Query Performance pull for your top three revenue ASINs, and put them in one sheet.
- Collapse the rows into topics and tier them — core, secondary, discovery — before touching any listing copy.
- Fix the mismatches first: core topics missing from your title, and discovery terms currently bid at core-term prices. A free Amazon audit covers the same two gaps if you would rather have them found for you.
