Start, sustain, and finish

How to know if your book idea is good

August 16, 2026 · The StoryProp team

Yes, you can tell — and three of the four tests you can run this afternoon. A book idea is good when it makes a promise you can state in one sentence, sits in a category readers are measurably buying, and can be delivered at the length that category buys. Circana BookScan counted 44 million US romance print units in calendar 2025, up 3.9%; science fiction units grew 22.1% and fantasy fell 8.7%. Your idea's neighbourhood is public information. The fourth test — whether you will still care in a year — is yours alone, and it decides most books.

Test one: is it a premise, and what does it promise?

The most common reason an idea feels alive in your head and then wobbles under pressure is that it is a situation, not a premise. A situation is a circumstance: a woman inherits her family's house after her mother dies. Evocative, but nothing in it has to happen. A premise adds a want that meets resistance. Mara comes back to sell the house and be done with it, and finds in a cupboard the chipped cup that should not exist — the one her mother swore was broken the night Daniel left. Now something has to happen, and either path, pursuing the truth or pretending she never saw it, costs her.

Most advice on developing an idea amounts to add more: more characters, more twists, more backstory. The fix is usually the opposite — add force. Ask who wants something, what stands in the way, and what it costs to keep going or to quit. Then finish this sentence out loud: by the last page, the reader will have… If the blank fills easily — witnessed a reckoning, solved the crime, watched a marriage survive on new terms — the idea is making an offer a reader can accept. If it will not fill, the idea is not bad. It is unmade, and the next hour of work is obvious.

Test two: is anyone buying books like it?

Taste is unmeasurable. Demand is not. Americans bought 762.4 million print books in 2025, up 0.3% and a second consecutive year of growth, on Circana BookScan figures reported by Publishers Weekly in January 2026. BookScan covers about 85% of US trade print retail, according to the American Booksellers Association, and counts no ebooks or audio at all, so read it as the shape of the market rather than its full size. Underneath that almost-flat headline, categories moved very differently — enough to matter to anyone choosing where to spend a year.

Romance is the story of the decade. Circana's June 2025 release, quoting analyst Brenna Connor, put romance at 51 million print units in the rolling twelve months to mid-2025 — more than double four years earlier — and named it the leading growth category of the entire US print market, with sales up 24% year to date. The two windows are not directly comparable, but they point the same way. The same release recorded psychological thrillers up 29%, dark fantasy up 23%, horror up 13%, and triple-digit growth in romantasy and sports romance. If your idea lives near those shelves, demand is not your problem; keeping the promise the genre makes is.

Fantasy's 8.7% fall is the sharpest among the big fiction categories, and it is not an instruction to stop: 24.1 million units is still an enormous readership, and BookScan has no romantasy category, so some of that reading is counted under romance. Science fiction grew fastest at 22.1%, but off a base of just over six million units. In nonfiction the split is sharper still — self-help rose 14.7% in 2025 while business and economics fell 3.9% and biography and memoir fell 7%. The UK Publishers Association's 2025 figures show the same asymmetry: consumer fiction revenue of £1.1bn, up 8%, against non-fiction at £1.0bn, down 3%.

Test three: can you deliver it at the size the category buys?

An idea that passes both earlier tests can still stall in the query pile for being the wrong length — and length is the one variable you can fix before writing a word. The Manuscript Academy's 2026 guide, sourced to agent Noelle Falcis Math of Transatlantic Literary Agency, publishes current target ranges genre by genre. Convert them using the Steve Laube Agency's working rule of 300 words to a printed page and you can see the object you are proposing to build — usually the moment an idea stops being a mood and becomes a project.

Now do the arithmetic on your own idea. An 85,000-word mystery at 1,000 words a session is 85 sessions — about four months at five sessions a week. NaNoWriMo's published target was 50,000 words in thirty days, an average of 1,667 words a day, and 50,000 divided by 300 is roughly 167 printed pages: below every adult trade fiction target in the table above, but squarely inside middle grade and category romance. That is not a failure of the month. It is a useful fact about what 50,000 words is, and about which shelf your idea already stands on.

Length also tells you which door to knock on. Harlequin states exact lengths line by line on its submission manager — Love Inspired at 55,000 words, Harlequin Desire and Harlequin Presents at 50,000 — and those lines accept submissions without an agent. SFWA's Nebula rules, effective January 2025, draw the boundaries with no rounding: under 7,500 words is a short story, 7,500 to 17,500 a novelette, 17,500 to 40,000 a novella, and 40,000 or more a novel. So if your idea is 30,000 words of story, it is not a failed novel. It is a novella, with awards categories of its own.

What the sales data actually says about ideas

The odds are worth knowing precisely. Kristen McLean, a books industry analyst at NPD/Circana, reverse-engineered sales data from the DOJ's case against Penguin Random House and found that about 66% of titles from the ten largest US trade publishers sold fewer than 1,000 print copies in a 52-week window, and fewer than 2% sold more than 50,000. Fifteen percent of big-publisher frontlist sold fewer than twelve copies in a calendar year. That last figure is the sourced version of the statistic that went viral in 2024 in a far more dramatic form; Lincoln Michel documented in Slate that April that the viral version misreads the trial exhibits, and a BookScan analyst could not reproduce it.

The supply side looks equally alarming until you look closely. Bowker registered more than four million US ISBNs in 2025, up 32.5%: 642,242 traditionally published titles, up 6.6%, and more than 3.5 million self-published, up 38.7%. But the two largest self-published subject categories that year were games and activities, at 354,684 titles, and travel, at 246,615 — an enormous low-effort tail, not 3.5 million rival novels. The number of books genuinely competing for a reader's evening is far smaller than the headline, and the number written with real care smaller still. That gap is what a good idea, properly executed, walks through.

Where a good idea actually earns

A book is not a one-week event, which is the most encouraging structural fact in publishing. Berrett-Koehler's Steven Piersanti estimates that new titles are now roughly a quarter of US retail book sales, down from 48% in 2005 — meaning about three-quarters of everything sold is backlist. The bestseller list proves it: in the first half of 2025, BookScan's US print top ten included Atomic Habits, from 2018, at 439,849 copies; Oh, the Places You'll Go!, from 1990, at 436,000; and Fourth Wing, from 2023, at 433,116. A good idea keeps earning for decades. A quiet launch week means much less than it feels like.

One format shift is worth knowing before you draft. The mass market paperback — the format that built genre fiction — fell 29.2% in 2025 on the Association of American Publishers' December StatShot panel, against a 2.4% rise for hardback. If your idea is a mass market thriller in the 1990s sense, the appetite is intact but the object has changed.

The test only you can run

The fourth test has no dataset behind it and predicts more than the other three. A book is months of your attention — the eighty-five sessions of the worked example above, and a revision pass after them — most of it unspectacular labour with no applause. The question is not whether the idea excites you tonight but whether it sits close to something you have not finished thinking about — an argument you are still having, a place you cannot leave alone, a person you never quite understood. Ideas rooted in your own unfinished business have deep fuel tanks. Ideas adopted because they resembled a market report tend to run dry right about where the real work starts.

It helps to know that no career rests on idea number one. Written Word Media's 2025 survey of 1,346 mostly self-published authors — self-selected, so read it for shape, not population — found 44% earning under $100 a month and 13% earning over $5,000. The difference tracked catalogue size: about 80% of authors with one to three books sat in the lowest band, while those over $10,000 a month averaged 25 or more titles. Its 2023 survey of over 2,000 authors put the median backlist at eight books, eighteen for romance authors. Even Onyx Storm, which sold about 1.3 million US print copies in its first week in January 2025, arrived after the Empyrean series had already sold nearly 6.5 million.

So write it. The cheapest decisive experiment on the idea you are holding is to draft the single scene it most clearly implies and read it cold the next morning; AI novel writing software that takes plain-language direction turns that into one evening's work. Then keep going: how to start writing a book picks up where the testing stops, which should be the moment you can say what you promised the reader, what question your ending answers, and which word-count band you are aiming at.

The fastest way to settle whether a premise carries is to write the scene it most obviously implies and read it cold. StoryProp takes that direction in plain language, drafts the scene, and writes down the decisions you keep — the promise, the question your ending answers, the length you are aiming at — so the test leaves you holding the first pages of the book instead of another page of notes.

← All posts