How the KDP Niche Finder works

Where the phrases come from

We maintain an index of search phrases observed on Amazon over time. Phrases that describe the same kind of book are grouped together, so “coloring book adults” and “adult coloring book” count as one niche typed two ways.

The index is rebuilt and re-checked on a schedule. It is not a live scrape of Amazon, which is why a lookup answers instantly and why the page always tells you the date the data was served.

What is counted, and what is estimated

Counted: how many distinct phrasings cluster into a niche, how many words are in your phrase, and the month we last observed it. These are facts about our index. If the number is twelve, we are holding twelve phrasings.

Estimated: everything in the paid suite that ranks one niche against another — demand, competition, and the score built from them. Those are our estimates from observed search suggestions and public listing data. They are evidence of demand, and they are not a guarantee of sales. A book is not a search query, and nothing here knows whether yours will sell.

This free page deliberately shows you none of the estimated figures. It shows what we counted.

Why there is no search volume

Amazon does not publish keyword search volume. There is no such numberto be accurate about, so any bare figure presented as one would be a guess wearing a fact’s clothes. What can be honestly measured is how broadly and how high a phrase appears across Amazon’s own search suggestions, and that is what the suite measures.

On this page that measurement usually reads not measured yet. Taking it needs a live check, which the suite runs and this free page does not. We would rather say we have not measured it than show you a number that means something else.

Evidence depth: seed, probed, verified

Niches sit at one of three depths. A seed estimate is built from the index alone — no live checks have been run, so it is the weakest thing we will show you and it says so. A probed niche has had its search evidence collected. A verified niche has had its live page-one listings read as well.

Depth changes how much a figure is worth, so it is stated rather than averaged away. In the suite, a score built from thin evidence is marked provisional and rendered with a wider band; a factor we could not measure at all is reported as insufficient, with the reason, instead of being counted as a zero. A missing measurement and a measurement of zero are different facts, and a niche with no competition data is not a niche with no competition.

Flags are prompts, not verdicts

Some phrases carry a flag — it reads like a brand name, or it describes a format we cannot help you produce. Each flag says two things: what we detected, and what you should verify yourself. They are the start of your own check, never the end of it.

We do not tell you whether a title may be used. That is a legal question about your specific book in your specific market, and no automated check can answer it. “Nothing detected” is a statement about our checks and their limits, not a clearance. For trademarks specifically, run the Book Trademark Checker, which screens against live records and is equally explicit about what it does not cover.

Where this is thin

Grouping is imperfect: two phrases that mean the same thing to a shopper can land in different niches, and an unusual phrasing can end up in a broader niche than it deserves. Coverage differs by marketplace: a niche we track well on Amazon.com may be sparse on Amazon.de simply because we have observed fewer phrases there. And the index has a date — a niche that appeared last week may not be in it yet.

None of that is hidden on the results card. The served date is printed under every answer, an unknown phrase says it is unknown rather than returning an empty niche, and a phrase we have seen but not yet grouped says that too.