Start, sustain, and finish

What should you write a book about?

August 16, 2026 · The StoryProp team

Write the one whose middle you can stand. That middle is the overwhelming majority of the job: the Manuscript Academy's 2026 word-count guide, sourced to agent Noelle Falcis Math of Transatlantic Literary Agency, puts a debut literary novel at 80,000 to 95,000 words, commercial fiction at 80,000 to 100,000, and epic fantasy at 90,000 to 120,000. Whichever idea you choose, that is the quantity you are agreeing to produce. The premise you fell in love with accounts for about a paragraph of it.

You are probably not short of ideas. You have a notes app full of premises, four abandoned first chapters, a folder called IDEAS, and a growing suspicion that having many ideas is functionally identical to having none. You are not looking for inspiration. You are looking for a decision procedure.

The trap is that choosing feels like a judgment about quality, so people try to rank their ideas by how good they are. But you cannot evaluate a book you have not written, and the assessment you make before you start has almost no relationship to the book that exists when you finish. Ranking by quality is a way of stalling that feels like diligence. What you need to know is which idea you can sustain, and that is a different measurement entirely.

Run the energy audit first

Take your top four or five candidates and, for each one, write down what you would have to do in the middle. Not the premise. Not the opening image, which is usually the part you fell in love with. The middle: at Reedsy's working convention of 2,000 to 4,000 words a chapter for adult fiction, an 80,000-word novel is somewhere between twenty and forty chapters, and most of them are the long stretch where the novelty has worn off and you are just a person doing a job.

For a heist novel, the middle is choreography — who is where, what they know, what the reader knows that they don't, and the tedious pleasure of keeping all of it straight. For a multigenerational family saga, the middle is inhabiting people whose historical moment you will have to research until you stop noticing you are researching. For a quiet domestic novel like the one about Mara and Daniel and their long unspoken argument, the middle is sitting in a kitchen for three chapters and making a chipped cup carry more weight than it reasonably should.

Now read what you wrote and notice your body. One of those descriptions will produce a small forward lean. Two will produce nothing. At least one will produce a distinct sinking feeling that you may have been mistaking for ambition. The heist premise might be your cleverest idea and still be the one whose middle reads like homework. That is not a verdict on the idea. It is information about you, and it is the most reliable information you have.

Writers routinely choose the idea with the best logline over the idea with the best middle, then spend a long time confused about why they can't sit down at their desk. The logline gets you through week one. The middle gets you to the end.

The dinner-party test

The second signal costs you nothing to collect, because you have been collecting it for years. Which of these ideas have you already told people about, unprompted, without being asked what you're working on? Not the pitch you rehearsed — the one that comes out sideways when someone says something adjacent and you say oh, that reminds me, I've been thinking about this thing.

That impulse to retell is the closest thing to a naturally occurring measurement of your own interest. A story you keep telling is a story that has not finished working on you. And notice what happens across the retellings: you drop the parts that bored the listener, you add the detail that got a reaction last time, you find the moment you always land on. You have been quietly drafting.

There is a corollary that stings. If an idea has lived in your notes for three years and you have never once told a human being about it, that is a signal too. Sometimes an idea is one you admire rather than one you want. You like being a person who might write it. That is a different relationship than the one required to actually write it, and it can persist indefinitely without ever converting.

Know what each idea is asking you to build

Every premise on your list arrives attached to a category, and categories have published dimensions. Choosing between a mystery and an epic fantasy is choosing between roughly 85,000 words and roughly 105,000 — a difference of about a quarter of a book, and all of it middle. Before you decide on the strength of the premise, look at what the finished object is contracted to be.

Length is only the visible part of the contract. If your candidate is a romance, readers expect the central relationship to be the engine and the ending to deliver on the promise the opening made. Category romance states its terms even more exactly: Harlequin's own submission listings ask for 55,000 words for Love Inspired and 50,000 for Desire, Presents and Harlequin Romance, and take unagented submissions at those lengths. If it's a mystery, the reader expects to be playing fair, which means committing to a structure of concealment and revelation that holds up on a second read. If it's literary fiction with no genre scaffolding, the expectation is that the prose itself will keep someone in the chair, which is a real and demanding job.

Some categories also expect more than one book. Written Word Media's 2023 author survey, the largest in its series at more than 2,000 respondents, found a median backlist of 8 published titles among working indie authors — and a median of 18 among romance authors specifically. That is worth knowing at the choosing stage, because a romance premise is often an invitation to a shelf rather than to a single volume. So the real question is not whether people buy these. It is: do I want to make the thing this idea is asking me to make, at the size it is asking for, as many times as it is asking? That framing disqualifies candidates fast. Several of yours will turn out to be premises you like attached to forms you have no interest in serving.

Use mashups to break a tie

Sometimes the audit and the dinner-party test leave two survivors and no way to separate them. Don't split the difference and don't flip a coin. Try collision. Take the strongest element of each and see whether they belong in one book — not a genre blend for its own sake, but a specific graft. The family saga's inherited house plus the mystery's unreliable timeline. The heist's ensemble plus the quiet novel's refusal to explain anything. Write a paragraph as though the combined book already exists and you are describing chapter nine to a friend.

One of three things happens. The combination is inert, and you have learned these were genuinely separate projects. Or it produces a third thing with more heat than either parent, and your choice just made itself. Or — most usefully — the paragraph reveals that one idea was never a book at all. It was a strong element looking for a home, and it has now found one inside the other project. A great deal of what looks like too many book ideas is one book and several of its future scenes, filed separately because they arrived on different days.

Commit without killing the others

The reason people cannot choose is rarely that the ideas are equally good. It is that choosing one feels like sentencing the rest to death, so they hedge, keep all five alive, and write none. The fix is structural rather than emotional: make the unchosen ideas a visible parking lot rather than a graveyard. Give each one a page with the premise, the two scenes you can already see, and the reason you didn't pick it this time.

Then keep the lot open while you draft. New material arrives constantly during a first draft, and a good portion of it does not belong to the book you're writing. When a scene shows up that clearly belongs to the family saga you set aside, you file it there and go back to work. That is the actual function of the parking lot — not consolation, but a place to put things so they stop competing for the front of your mind. This is exactly what a per-project workspace is for, and it is how StoryProp is built: each book is its own project, with its own canon, its own characters, its own research kept separate with sources and dates, and every draft retained. The saga you parked keeps its own records while you work on the one you chose.

The parking lot is for material, not for permission to switch. Set a floor before you start and make it specific: twenty chapters, or 30,000 words, or the last Sunday in March. NaNoWriMo's published rules asked for 50,000 words between 1 and 30 November — an average of 1,667 words a day — and that quota moved an enormous number of drafts precisely because it was arbitrary, public and countable. A third of the way in, every book becomes harder than the shiny unwritten alternative, and the lot will suddenly look like an escape hatch. It isn't. That difficulty is what drafting feels like, and it will arrive in the same place in whichever project you flee to.

Once you've chosen, the question stops being which of these and becomes how do I build this one out — which is outlining work, and it is only answerable about a single idea at a time. That is why it has to come second: you cannot pressure-test five things at once. Pick the one with the sustainable middle and the retelling instinct behind it, park the rest, and start. The idea you can stand for 90,000 words is the right one, and you already know which it is.

Once you have picked, the whole job is the middle — and it goes better when the decision lives somewhere. StoryProp keeps each project separate, with the canon, characters and style rules you agree to written down before it drafts a word, so the book you chose gets your full attention and the ones you parked stay intact.

← All posts