Amazon assigns every book it sells a bestseller rank, recalculated hourly, inside categories subdivided narrowly enough that a few dozen purchases in a week can carry a title to number one in something like Teen & Young Adult Humorous Fantasy eBooks. The badge that follows — a ribbon on the product page, a screenshot dropped into a newsletter — gets forwarded to us constantly as proof a book has arrived. It proves the algorithm noticed a sales spike. Wikipedia's guideline for books is asking something else entirely.
WP:NBOOK tests a book against five things, and sales isn't one of them: two or more substantial, independent reviews in a trade outlet that reviews books as a matter of editorial practice, such as Kirkus Reviews, Publishers Weekly, or Library Journal, or in a newspaper's book section; a major literary award; assignment as a text in multiple college courses; or having become the basis for a notable film or other adaptation. A sales figure, however it's framed, answers none of the five.
The test excludes press-release reprints, flap copy, and other statements by the author, the publisher, or anyone with an interest in promoting the book.
The rank fails for a specific reason: nobody read the book to produce it. An algorithm counted purchases and assigned a position; no editor judged the book worth describing, no critic weighed it against anything else in its genre. A real bestseller list — the kind a newspaper or trade publication compiles nationally, with a stated methodology — is a different object, and the guideline's own list of qualifying coverage includes exactly that. A retailer's internal subcategory ranking was never built to answer the same question.
Two self-published novels came through recently and make the gap concrete. The first spent a week at number one in a Kindle subcategory for frontier-town mysteries; the author's own site called it a bestseller, and the draft cited that site as its source. Nothing else existed outside retailer pages and the author's social accounts. It was declined. The second sold a fraction as many copies but drew a three-paragraph notice from Kirkus and a longer piece in a regional newspaper's books section, both assessing the prose rather than reporting a number. That draft cleared review on its first submission.
Finding the second kind of evidence means searching trade-press and newspaper archives by title and author, not a retailer's category pages. A one-paragraph notice that announces a release without assessing it fails the same test a press release fails everywhere else in Wikipedia's sourcing rules: acknowledgment, not analysis. Two notices of that kind don't add up to one qualifying review any more than two press releases add up to one independent source.
A sales rank answers a retailer's question: did people buy this, this week, inside this category. Wikipedia's guideline for books answers a different one — did anyone outside the publisher sit with the text long enough to say something specific about it. The badge on the product page was never built to answer that, and changing categories won't make it start.
