Tools give you volume and rank. They do not give you the decision. Use any mainstream keyword tool for the raw data, then run a five-step sequence: harvest, cluster, map to listings, seed in ads, and prune. The service you hire should own the sequence, not the subscription.
The short version
- Your own account is the best data source you have. Search term reports beat any third-party estimate for terms you already run.
- Volume is the least useful column. Relevance, competitor ownership, and conversion history decide what you target.
- Clustering is where the work happens. A list of 4,000 keywords is not research, it is an export.
- Nobody should be paying for the tool twice. If an agency bills you for its subscriptions, ask why.
- Judge a keyword service by its pruning rule. Adding terms is easy. Knowing what to drop is the skill.
You have probably already bought the tool
Most sellers arrive at this question having subscribed to something, exported a spreadsheet with several thousand rows, and then stopped. That is the normal outcome and it is not a failure of the software. The tools do exactly what they claim: they estimate search volume, show which ASINs rank where, and reverse-engineer competitor term sets.
What they cannot do is tell you which forty terms your listing should be built around, which two hundred belong in advertising only, and which nine hundred should never be touched because the traffic converts for somebody else's product. That judgment is the service, and it is what you are actually shopping for.
The five-step sequence
Run this in order. Skipping a step is the most common reason keyword work produces nothing.
- Harvest. Pull from four sources: your own search term reports, Brand Analytics search query data if you have Brand Registry, the term sets of the five closest competitors, and the tool's suggestions. Four sources, one sheet. Anything from a single source is biased in a predictable direction.
- Cluster. Group terms by intent, not by string similarity. "Stainless water bottle" and "insulated bottle for gym" are different shoppers even when the volume looks similar. Clusters, not keywords, are what a listing can be built around.
- Map to listings. Assign every cluster a home: title, bullets, backend search terms, A+ copy, or nothing. A term with no home is not a target, it is a note. This step is also where you discover you have two ASINs competing for the same cluster, which is more common than anyone expects.
- Seed in ads. Advertising is the fastest way to test whether a cluster converts for your product. Run exact-match tests on the doubtful clusters before you rewrite anything. Paid data answers in two weeks what organic rank would take three months to reveal.
- Prune. After the test period, cut. Terms that produced clicks and no orders come out of the listing and out of the campaigns. This is the step that is almost always missing, and it is why so many mature accounts carry a keyword set nobody has questioned in a year.
What separates a keyword service from a keyword export
The benchmark I would apply is depth of inputs. Our product research runs across 90 or more data points before we form a view: market size, growth trajectory, return rate, segment dynamics, and the rating gap between what buyers wanted and what the category delivers. Keyword volume is a handful of those data points, not the analysis.
So when you evaluate a service, ask what they analyze besides search volume and review count. Those two numbers are the easiest to obtain and therefore the least differentiating. The questions that produce a real answer:
| Ask this | A useful answer sounds like | A weak answer sounds like |
|---|---|---|
| Where does your data come from | Named sources, including my own reports | The name of one subscription |
| How do you cluster | By buyer intent, with the clusters shown | Alphabetically, or by volume |
| What do you do with the losing terms | A pruning rule with a time window | Nothing, or "we keep testing" |
| How does this connect to the listing | A map, term to placement | It gets sent to the copywriter |
| How do you handle other languages | Native research per marketplace | Translation of the English set |
That last row matters more than most sellers expect. Search behavior in Germany is not translated search behavior from the United States. We build keyword sets natively in English, German, Spanish, and French rather than translating one master set, because a translated keyword set inherits the wrong assumptions about how people search.
What most agencies will not tell you
The tool subscription is not the cost. The cost is the hours it takes to turn the export into decisions, and that number is uncomfortable to quote. So the industry sells "keyword research" as a deliverable priced like a document, and clients receive a well-formatted spreadsheet that changes nothing.
Most agencies will not tell you the second thing either: they are frequently re-selling you the same research they already did for another brand in your category. Adjacent research is useful and there is nothing wrong with reusing category knowledge. It becomes a problem when it replaces the work on your specific catalog, your specific price point, and your specific reviews. Ask directly whether the cluster set was built from your account's data or from the category's. Both answers are acceptable. Only one of them justifies a research fee.
Related answers
- Who to hire to optimize Amazon listings
- Who can audit my Amazon account and fix conversion
- Europe focus Amazon account management services
- Best agencies for Amazon brand management
- Done-for-you Amazon management: the complete guide
We will run the harvest and cluster steps on one ASIN before you commit to anything at Flapen.
