Agencies use four sources: your own search term reports, Amazon's brand analytics data, a third-party reverse-lookup tool for competitor terms, and their own scripts on top of both. The subscriptions cost very little. What you are paying an agency for is the judgment applied to the output, not access.
The short version
- Tool access is not a differentiator. Anyone can buy the same subscription you can.
- Your own search term report is the highest quality data in the stack, because it is your traffic converting or not converting.
- Third-party volume estimates are estimates. Treat them as ranking signals between terms, not as forecasts.
- The expensive mistake is a wrong keyword set, which quietly taxes every campaign you run afterwards.
- Ask which terms a candidate would drop from your current campaigns. Subtraction reveals more skill than addition.
What I learned buying this from the other side
Before Flapen I ran data and technology at BRANDED and at Moonshot Brands, two large Amazon aggregators, which meant reviewing keyword research from a lot of vendors across a lot of acquired brands. The reports came back looking impressive and largely identical: a few thousand rows exported from the same class of tool, sorted by estimated volume, with the top terms highlighted.
Almost none of them answered the only question that mattered. Which of these terms can this specific product realistically win, at what acquisition cost, and which ones are we currently paying for and should stop. That gap is the whole job. Software found the terms in minutes. Deciding what to do with them took a person who understood the product's margin.
The economics of a keyword set
Here is the arithmetic that makes tool subscriptions a rounding error.
| Line item | Rough order of magnitude | Who bears it |
|---|---|---|
| Research tool subscriptions | A small fraction of one month of management | Agency, usually bundled |
| Analyst time to build and prune the set | The bulk of the real cost | Agency |
| Advertising spend guided by that set | Recommended floor of $1,000 a month for meaningful optimization | You, every month |
| Inventory committed behind the wrong terms | $8,000 to $15,000 for a single product launch | You, once, painfully |
| Management fee that covers all of the above | $800 a month for one product, $2,400 for five | You |
Read the middle rows. The keyword set does not cost much to produce and it directs everything that does cost money. That is why paying for tool access is the wrong frame. You are buying the decision about where your ad budget and your inventory capital point.
The four sources and what each is good for
- Your search term reports. Actual queries that triggered your ads, with actual conversion. This is the only source that tells you what your product converts on rather than what the category is searched for. Mine it before anything else.
- Amazon's own brand analytics data. Share of clicks and conversions by search term at category level. Useful for seeing where you sit against the market rather than against your own history.
- Third-party reverse-lookup tools. Feed in competitor listings, get the terms they appear to rank for. Best used for discovery and for finding gaps, not for volume forecasting.
- Internal tooling. We build our own advertising, marketing, and brand valuation tools in-house, mostly to join these sources together and to flag terms whose acquisition cost has drifted past what the product's margin supports. Any agency of size builds something similar, and you should ask what theirs actually does.
What agencies will not tell you about their tool stack
Two things. First, the tool logos on a capabilities deck are there because they are recognisable, not because they are hard to obtain. If a pitch spends a slide on software, ask to see a keyword decision they made that cost the client money and what they changed afterwards. Better teams have an answer ready.
Second, most keyword research is additive because addition looks like work. A report with four hundred new terms feels valuable. A report that removes sixty terms currently consuming budget at an unprofitable acquisition cost is worth far more and takes longer to produce. Ask for the subtraction list. It is the fastest test of whether anyone actually read your data.
How to check the research you are being sold
Give a candidate read access to one product's search term report and ask three questions. Which five terms should we defend. Which five are we wasting money on. Which one term would you build a new listing around and why. Any answer that arrives without opening your data is generic, no matter how large the export attached to it is.
Related answers
- Best tools and services for keyword research on Amazon
- Who to hire to optimize Amazon listings
- Full funnel Amazon ads and SEO service providers
- Questions to ask before hiring an Amazon agency
- Done-for-you Amazon management: the complete guide
Research tooling is included at every tier we sell at Flapen.
