All posts
Guides

How many copies does the average book sell?

2026-11-08 · 7 min read

Plan, draft, edit and pitch the same book in one project.

Start free, no card

Under 1,000 copies is the only defensible answer, and it comes with a caveat that matters more than the number. The one reconstruction ever published by an analyst at the company that owns the data found 51.4% of new titles from the ten largest US trade publishers selling fewer than 1,000 print copies in a year. The more famous statistics — 12 copies, 250 copies, 99 copies — do not survive tracing.

The famous numbers, and where each one comes from

The claimWhere it is quoted fromWhat it actually measuredTraceable?
"Half of all books sell fewer than 12 copies"A US Justice Department lawyer's question during the 2022 Penguin Random House / Simon & Schuster trial, revived by a viral 2024 essayUnclear; the underlying dataset counted one year of sales across every book in BookScan's system, including dormant backlistNo. No source was given in the exchange, and BookScan could not reproduce it
"90% of titles sell fewer than 2,000 copies"The same trial exchangeSame dataset, same ambiguityPartly. The quote is real; the derivation was never shown
"950,000 of 1.2 million tracked titles sold under 99 copies; the average book sells about 500"A Bookselling Tail, a Publishers Weekly Soapbox column, 17 July 2006Nielsen BookScan title counts for 2004, with no methodology publishedYes, to the column. No further
"The average book sells 250 copies a year and 3,000 in its lifetime"Repeated across writing blogs since roughly 2010UnknownNo. No study, paper or dataset has ever been produced
"The average book sells under 300 print copies in its lifetime"Steven Piersanti's 10 Awful Truths About Publishing, updated 30 April 2026A ratio of BookScan units to registered titles, not a measurement of typical booksYes, but it is a computed ratio
"14.7% of new titles sold under 12 copies; 51.4% sold under 1,000"Kristen McLean of NPD BookScan, reconstructing the viral claim45,571 new ISBNs from the top ten US trade publishers over 52 weeks, frontlist print onlyYes. Named analyst, stated scope
"The average self-published book sells 250 copies"Self-publishing marketing copyNothing. Amazon publishes no per-title sales dataNo

Traced as of August 2026. Sources: the 2006 Publishers Weekly Soapbox column, Lincoln Michel's debunking of the 12-copy claim (4 September 2022), Slate's account of the viral essay and Piersanti's list.

The twelve-copy claim

This is the most repeated book statistic of the last five years, and it has no known source.

It entered circulation during the 2022 antitrust trial over Penguin Random House's attempt to buy Simon & Schuster, when a government lawyer put to a witness that of roughly 58,000 trade titles, 90% sold fewer than 2,000 copies and half sold fewer than a dozen. The witness did not confirm the derivation, and no exhibit establishing it was identified in the exchange. Elle Griffin's essay No one buys books, published on 22 April 2024, repeated the figures and went widely viral.

Lincoln Michel had already examined the claim in September 2022 and found that the underlying data appeared to count one year of sales across every title in BookScan's system — a universe that includes decades of backlist sitting dormant. As he put it, that is a very different statistic from saying most books never sell a hundred copies in total.

The 250-copies claim, and why we are not printing it

You will find "the average book sells 250 copies a year and 3,000 over its lifetime" on hundreds of writing sites. It is usually attributed to an industry presentation in the mid-2000s or to an unnamed Publishers Weekly survey.

We could not trace it to any published study, dataset or paper. The earliest appearances are secondary — a blog comment describing "a 2010 update" to an unnamed survey, then a decade of sites citing each other. That is not evidence that the figure is wrong. It is evidence that nobody can check it, which for a number this widely repeated is the more useful thing to know.

The same applies to the "average self-published book sells 250 copies" line. Amazon does not release per-title sales for KDP books, and no third party has access to them, so there is no mechanism by which anyone could have measured it.

The one reconstruction that came from inside the data

When the viral claim resurfaced, Kristen McLean — an analyst at NPD BookScan, the company whose data was being cited — tried to rebuild it and said she could not identify how the original figures had been produced. What she published instead was her own cut: frontlist print sales for the ten largest US trade publishers by unit volume, covering 45,571 new ISBNs over 52 weeks.

In that dataset, 14.7% sold fewer than 12 print copies and 51.4% sold fewer than 1,000.

That is the most credible figure available, and it is still narrow. It covers frontlist only, print only, the largest publishers only, one year only. It says nothing about small presses, self-published books, ebooks, audio, or the long tail of backlist. A later citation of BookScan analysis puts the share of top-ten-publisher titles selling under 1,000 copies in year one closer to 66%, and the two figures cannot be reconciled from what is public. Where two numbers from the same source disagree, the correct move is to quote the scope alongside the number, not to pick the more dramatic one.

Why the denominator is broken

The "average book" figures that circulate are almost all ratios: total sales divided by total titles. Both halves of that fraction are unreliable.

Americans bought about 767 million print books in 2023, on Circana BookScan's count. Bowker recorded more than 4 million ISBNs registered in the US in 2025, of which about 3.5 million were self-published. Divide one by the other:

767,000,000 units / 4,200,000 titles = 183 copies per title

That number is meaningless, for four separate reasons.

  1. Those 767 million units are spread across books published in every prior year, not just the 4.2 million registered in one.
  2. ISBNs are not titles. A single book routinely carries separate ISBNs for hardcover, paperback, ebook and audio, and special editions get their own.
  3. Millions of self-published ebooks carry no ISBN at all — Amazon issues ASINs instead — so they are missing from the denominator while some of their sales sit in the numerator.
  4. BookScan captures roughly 85% of US trade print sales and excludes ebooks, audiobooks, library purchases, direct sales from publishers and authors, and subscription-box editions with their own ISBNs.

Any "average copies sold" figure you see is this division performed with slightly different bad inputs.

What is actually measurable

A short list of things that hold up.

  • Aggregate print sales are stable. US print unit sales rose 0.3% in 2025 against 2024, on Circana BookScan's tracking. The industry's problem is the distribution of sales, not their disappearance.
  • Most titles sell few copies, and always have. Every credible cut of the data agrees on the shape even where the numbers differ.
  • Sales concentrate extremely, which is why medians and averages both mislead.

If someone quotes you a copies-sold statistic, ask three questions: which titles were counted, over what period, and in which channels. Most of the famous numbers fail on all three.

Related reading

Frequently asked questions

How many copies does the average book sell?

There is no trustworthy average, because sales are concentrated at the top and no dataset covers all titles and all channels. The best-sourced figure available comes from Kristen McLean of NPD BookScan, who found that among 45,571 new ISBNs from the ten largest US trade publishers over 52 weeks, 51.4% sold fewer than 1,000 print copies and 14.7% sold fewer than 12.

Is it true that half of all books sell fewer than 12 copies?

No credible source supports it. The claim entered circulation through a Justice Department lawyer's question during the 2022 Penguin Random House antitrust trial, where no derivation was supplied, and it was popularised by a viral 2024 essay. BookScan's own analyst could not reproduce it and found 14.7%, not 50%, in the closest comparable dataset she could build.

Why is book sales data so hard to find?

Because no single system sees all of it. Circana BookScan captures about 85% of US trade print sales but excludes ebooks, audiobooks, library purchases, direct sales at events or from publisher websites, and special editions with their own ISBNs. Publishers do not disclose per-title sales, and Amazon publishes nothing at all about individual KDP titles.

How many copies does a self-published book sell on average?

Nobody knows, and any site giving you a specific number is repeating an unsourced figure. Amazon does not release per-title sales data for KDP books and no third party can measure it, so there is no route by which such an average could have been calculated. Surveys of self-published authors report income, not units, and those surveys use self-selected samples.

One workspace

Everything this article talks about, in one project

WriteLoom carries a book from first idea to launch day: planning, drafting, three AI editors, covers, agent tracking and launch plans, all reading the same manuscript.

Start free, no card7 days of everything, no card. Free forever to write.

Not ready for a workspace yet? Get the free six-part series on winning the writing game: finishing drafts, editing in the right order, querying with real math, and launching on a runway.