How to Choose KDP Categories: A Step-by-Step Method
A repeatable method for choosing your three KDP categories by relevance, buyer fit and competition, with a worked example and a ve…
What a BSR Actually Is
Best Sellers Rank (BSR) is a number Amazon prints on a listing. Lower means the book is selling better relative to other books in the same category. Amazon does not publish the formula, and it does not publish unit sales for any book that is not yours.
As commonly reported by third-party sellers and analytics vendors, the rank refreshes roughly every hour and weights recent sales more heavily than older ones. Amazon's own Best Sellers Rank help page is the place to check the current wording before you rely on either detail.
Two facts matter for everything below. First, each format has its own rank: a Kindle eBook can sit at a strong rank while the paperback edition of the same title ranks far lower. Second, the rank is relative. A book at rank 100,000 sells a different number of copies on a slow Tuesday in February than in the week before Christmas, because the crowd it is being compared with sells a different amount.
A keyword extension therefore starts from a number with no unit attached. Converting it into copies per day is a modelling step, and modelling steps can be wrong.
How Extensions Collect the Data
Most browser-based tools follow the same basic pattern. Vendors differ in the details, and none of them publish full technical documentation, so read the following as the common mechanism rather than a description of any one product.
1
Read the page you are on
A content script runs on the Amazon search or product page in your browser and reads what is already in the HTML: title, price, page count, publication date, review count, and the BSR line in the product details. Nothing secret is involved. If you can see it, the extension can read it.
2
Fetch neighbouring pages
To fill a table of the top 10 or 20 results, the tool requests each product page in the background and parses it the same way. This is why some extensions make a results page feel slow the first time.
3
Send identifiers to a server
Many tools pass the book identifier (ASIN) and the rank to the vendor's server, which returns the sales estimate, a competition score or a historical chart. The heavy modelling lives there, not in your browser.
4
Ask Amazon's suggestion box
For keyword ideas, tools query the same type-ahead suggestions you see when you start typing in the Amazon search bar, often once per seed phrase plus each letter of the alphabet.
The consequence: the rank, price, page count and review count are direct observations. Sales per day, monthly revenue, search volume and competition scores come from the vendor's own model. When a dashboard mixes both kinds of number in one row, it rarely says which is which.
How to read both rank lines yourself
You do not need an extension to see the raw rank. Open the book's product page, scroll to the Product details block, and find the Best Sellers Rank line. It shows an overall figure such as "#N in Books" (or "in Kindle Store" for an eBook) followed by one or more subcategory lines.
The page shows the rank for the format you have selected. To compare the Kindle and paperback ranks of the same title, click the other format on the listing; it opens a separate page with its own rank. Write down both, and note which one any tool quotes.
How a BSR Becomes a Sales Number
There is no public table that maps rank to copies sold. Vendors build their own by tracking how a book's rank moves when it sells, and by comparing against sales figures they can see.
As of September 2026, Kindlepreneur's page for its Publisher Rocket sales-rank calculator describes an algorithm built from tracked rank changes compared with real sales data, and says it now runs on Publisher Rocket's continuously updated data. It calls its outputs estimates. Amazon publishes no formula, so any such model remains an approximation.
Free calculators and paid extensions differ in what they hold constant. The table lays out the data points you will meet and how much weight each deserves.
| Data point | Where it comes from | Direct or modelled | How far to trust it |
|---|---|---|---|
| BSR | Amazon's product page | Direct | Accurate at the moment of the snapshot; it moves through the day |
| Price, page count, publication date | Amazon's product page | Direct | High |
| Review count | Amazon's product page | Direct | High, but reviews lag sales by weeks |
| Daily or monthly sales | Vendor model applied to the rank | Modelled | Order of magnitude only |
| Monthly revenue | Modelled sales times list price | Modelled twice | Ignores discounts, Kindle Unlimited page-read income and your actual royalty |
| Search volume | Vendor model, often from third-party panel data | Modelled | Use for ranking phrases against each other, not as a count |
| Competition score | Vendor formula (reviews, ranks, listing text) | Modelled | Unaudited; the weights are the vendor's opinion |
Amazon itself now shows a purchase signal on many search results: a "bought in past month" line with rounded buckets such as 50+ or 1K+. As of September 2026, third-party summaries describe it as a rolling 30-day window that rounds down to a threshold and excludes returns and cancellations, and sellers report that not every category shows it. It is the closest thing to a first-party sales figure a competitor's book can display, and it is coarse by design: a 100+ label could sit near the bottom or well up the range. Confirm the current rules on Amazon's own help pages before quoting the badge as data.
A worked example with a stress-test range
Take a hypothetical US paperback at BSR 10,000. Octozia's free BSR calculator, using its print curve, returns about 4.5 copies per day for that rank (as of September 2026). The calculator is itself a model: it interpolates between reference points that were derived from published sales-rank charts (Kindlepreneur's among them) and author self-reports, and its print curve is an approximation. Treat 4.5 as one estimate among several, because different categories and marketplaces pair ranks and sales differently.
Run the numbers. 4.5 × 30 = 135 copies a month. Suppose your per-copy royalty on a comparable book is $2.50 (a made-up figure; pull yours from KDP's royalty calculator or your reports). The implied income is 135 × $2.50 = $337.50 a month.
Now stress-test it by moving the estimate half down and half up. This 50 percent band is an arbitrary range for illustration, not a measured error.
- Low case: 2.25 per day, 67.5 copies, 67.5 × $2.50 = $168.75.
- High case: 6.75 per day, 202.5 copies, 202.5 × $2.50 = $506.25.
The single headline figure of $337.50 hides a three-to-one spread between the two cases.
Real error is not fixed. It tends to grow at worse ranks and with the failure modes in the next section. Once you have tested a tool on your own book (step 1 below), replace the 50 percent with the gap you actually measured.
How Search Volume Gets Estimated
KDP gives authors no store-wide search-volume report; check Amazon Ads and Brand Registry help pages for what your own account can see. Suggestion-expander tools say it outright: autocomplete returns phrases, not counts. So the volume figure beside a keyword has to come from somewhere else.
Three sources are commonly cited by vendors:
- Autocomplete position and presence. A phrase that appears early in the type-ahead list is probably searched more than one that never appears. That supports a ranking of phrases, not a monthly count.
- Clickstream panels. As of September 2026, vendors such as Keyword Tool say their estimates are based on clickstream data: browsing behaviour from a sample of users, scaled up. A sample that skews toward one country, one browser or one shopper type skews the estimate.
- Your own advertising data. Amazon Ads shows impressions and clicks on search terms inside your campaigns. This is the only volume evidence you generate yourself, and it covers only terms you bid on.
A useful test for any volume figure: check whether the tool gives a number for a phrase nobody would type. A tool that returns 300 monthly searches for a phrase you invented is telling you something about the model, not the store.
Long-tail phrases are the weakest. Panels are thin on rare queries, so the difference between 10 and 50 monthly searches is mostly noise. Rank long-tail candidates by how they behave in the store (do the top results share the phrase, are they recent, are they selling) instead of by the count column. Related reading on how POD sellers vet keywords sits in the guide to keyword research tools for print on demand.
Where the Numbers Go Wrong
Most errors come from a small set of causes. Knowing them lets you decide when to discount a figure before you have spent any money.
| Failure | What happens | Sign to look for |
|---|---|---|
| Snapshot timing | The rank you see is one moment in a moving series, so a launch-week spike or a holiday dip gets read as normal | The listing is under three months old, or you are viewing it in Q4 |
| Wrong format | A Kindle rank gets used to judge a paperback, or the reverse | The same title has a Kindle listing and a paperback listing with separate ranks; check which format's page or ASIN the tool read |
| Category mismatch | Rank 20,000 in a tiny subcategory means far fewer copies than 20,000 overall | Both an overall rank and a niche rank are shown; the tool quotes one without labelling it |
| Kindle Unlimited borrows | Sellers report that KU borrows count toward rank without appearing as unit sales, so unit estimates for KU-enrolled eBooks run high (reported, not confirmed by Amazon) | Reported pattern for eBooks enrolled in KDP Select; check KDP's Kindle Unlimited help page for how borrows are counted |
| Price promotions | A temporary discount lifts rank, and revenue estimates use the list price | A strike-through price or a coupon on the page |
| Bundles and variants | One rank reflects several formats or a series read-through | Many near-identical titles from one author |
| Stale cache | The tool shows a saved value from its own server, not today's rank | A last-updated date, or a rank that differs from the page in front of you |
Category mismatch deserves a second look because it flips niche decisions. Overall rank compares a book against millions of titles. A niche rank compares it against a few hundred. Two books can both read "top 5" in their subcategory while selling wildly different amounts.
A Five-Step Sanity Check Before You Trust a Tool
You can test any extension in an afternoon without paying for anything beyond its own trial.
1
Test it on a book you own
Open your own listing, read the BSR, and compare the tool's daily sales estimate with your KDP sales report for the same day. If you have no book, ask a publishing friend to share one week of numbers. One book gives you a rough calibration point and your own error figure.
2
Cross-check with a second estimator
Enter the same rank into the free BSR-to-sales calculator and any other estimator you can find. All of them are models, so hold each to the same test. Three sources within a factor of two give you a working range, and a tenfold disagreement means at least one model is guessing. Estimators built on the same public rank charts tend to agree with each other, so agreement here is weaker evidence than a match against your own KDP report in step 1.
3
Re-check the rank at three times of day
Look at the same listing morning, afternoon and late evening. A rank that swings by tens of thousands is a book with a handful of sales per day, and a single-day figure from it is meaningless.
4
Compare volume against the store
Type the phrase into Amazon. If the tool says high volume but the results show old, irrelevant titles and no purchase badges, the estimate is probably inflated.
5
Read the tool's methodology page
A vendor that states where the estimate comes from, what it holds constant and how often it refreshes gives you something to audit. One that says only that the numbers are accurate gives you nothing.
Choosing a KDP Keyword Research Tool and Using Its Output
Tools fall into a few categories: browser extensions that read Amazon pages as you browse, desktop or web apps such as Kindlepreneur's Publisher Rocket that combine suggestion data with a sales model, and standalone suggestion expanders such as Keyword Tool. Check each vendor's pricing page for the current price and any trial; prices change.
Octozia has no KDP keyword or BSR extension. Its Chrome extensions cover Redbubble and TeePublic, and its KDP-side free tools are the calculators at /calculators, including the BSR-to-sales one used above.
Whichever tool you pick, its output has three jobs:
- Backend keywords. KDP gives you seven keyword slots when you set up a book. Fill them with distinct phrases a buyer would type, not repeats of your title. Check KDP's help pages for the current character limit and rules.
- Title and subtitle phrases. The strongest phrase, the one the top results share and that shows real purchase signals, belongs in the title or subtitle where a shopper reads it.
- Category choice. Read the subcategory ranks of the top sellers and pick the category where a rank like yours would sit near the top.
One plain limit: your own KDP reports and your Amazon Ads search-term data beat any paid extension, because they show real customers and real sales. A tool is a starting point until you have your own data. Free routes, typing a seed phrase into Amazon and reading the suggestions, or reading ranks by hand as described above, cover a lot of ground before you spend anything.
Using the Numbers for a Niche or a Launch Price
Treat the tool as a filter, not a verdict. It is good at eliminating things: a niche where every top result sits above a rank of a million is not producing much. It is bad at telling a 5-copy-a-day niche from a 15-copy-a-day niche.
For niche picking, use ranks as bands. Group results into rough tiers (strong, middling, dead) and ask how many of the top 10 are in the strong tier and how old they are. Three young books doing well beats one aged bestseller propping up the average.
For pricing, borrow the price, not the sales figure. Read what the top sellers charge, then work out your own margin with the KDP royalty rules. Amazon's royalty rates and printing-cost tables change, so pull the current ones from KDP's help pages, and remember large-trim books such as 8.5 × 11 puzzle books use a different printing-cost table than regular trim. The free KDP royalty calculator models regular trim only.
If your idea also extends to shirts, stickers or other print-on-demand products, the same rank-as-a-band habit applies. Two guides cover that side: best print on demand niches and the POD niche research workflow.
One more habit worth adopting: keep a log. Each time you consult a tool, write down the phrase, the reported figure and the date. After a few launches you will know how far your tool of choice drifts from what you actually sold, which is worth more than any vendor's accuracy claim.
FAQ
Where does a KDP keyword extension get its BSR?
From Amazon's product page. The extension reads the Best Sellers Rank line in the same HTML you see, sometimes after loading neighbouring pages in the background. The rank itself is Amazon's number. What the tool adds on top, such as sales estimates, is its own model and not data from Amazon.
Do I need a paid KDP Amazon BSR & keyword research extension?
Not to start. The rank is visible on every listing, Amazon's suggestion box is free, and a free calculator gives a rough sales range. A paid tool saves time by collecting these into one table. Your own KDP reports and Amazon Ads search-term data are more reliable than any of them once you have sales.
How accurate is a BSR-to-sales estimate?
Good enough for a range, not for a single figure. Vendors calibrate their models against tracked rank changes and some real sales, but Amazon publishes no formula. Treat any estimate as plus or minus a wide margin, and test the tool against a book whose sales you know.
Can a tool show real search volume for Amazon keywords?
Not directly. Amazon's type-ahead suggestions show phrases without counts, so volume figures come from vendor models, sometimes based on clickstream panel data. They are useful for comparing phrases against each other. Your own Amazon Ads search-term reports are the best first-party evidence you can get.
Why does the same book show a different BSR in two tools?
Timing and format are the usual causes. Rank changes through the day, and one tool may show a cached value while another reads the live page. A second cause is quoting different ranks: Kindle versus paperback, or overall versus subcategory.
Is the "bought in past month" badge better than a BSR estimate?
It comes from Amazon, so it is a first-party signal, but it is bucketed and rounded down, and not every category shows it. Use it to confirm that a book is selling, then use your own margin maths to decide whether the niche is worth entering.
If your niche is puzzle books, KDP Creator builds the pages
KDP Creator is an interior tool, not a research tool: it generates puzzle pages as numbered 300-DPI PNG images in a ZIP that you assemble into your interior; 3-day trial, card required.
