Most working novelists write three to five full drafts before a book is done — a range that shows up consistently in writers' own documented accounts, from Ernest Hemingway's reported 39 rewrites of the ending of A Farewell to Arms to the more typical two or three passes many contemporary novelists describe in interviews and craft essays. The number matters less than what each draft is for. A draft is finished when it has done its one job, not when it feels perfect.
Daily Pencil writes about craft from documented practice; every habit and figure below belongs to its source, and the method that fits your book remains your call.
Why is there no standard draft count?
Because "draft" means different work at different desks. Some writers redraft from a blank page; others revise continuously, so their "first draft" already contains hundreds of local decisions. Stephen King, in his 2000 craft memoir On Writing, described a routine of a first draft closed away for at least six weeks, then a second pass cutting ten percent, then a third for polish — three drafts by his arithmetic, a dozen by a blank-pager's. The comparison only works when you know what each writer counts.
What do documented writers actually report?
The spread, from their own published accounts:
- Hemingway rewrote the last page of A Farewell to Arms 39 times, per his 1956 Paris Review interview with George Plimpton — though he framed it as getting the words right, not drafting 39 times.
- King's three-pass system, laid out in On Writing (2000): draft, rest, cut.
- Anne Lamott's famous "shitty first drafts" essay in Bird by Bird (1994) makes the first draft's job purely existence; revision is where the book happens.
- Novelists interviewed in The Paris Review's Writers at Work series over decades commonly describe two to five full passes, with more on endings.
What none of them describe is a single universal count. The documented writers treat the question as "what does this book need," not "how many am I allowed."
How do I know a draft is done?
Four signals recur in craft essays and editors' accounts: you are making only cosmetic changes (word swaps a copyeditor could make); the changes you make one week, you reverse the next; readers you trust stop flagging the same problems; and you can summarize what each scene does without hedging. When all four hold, the draft has done its job. One caution from editors' widely shared experience: the draft you can no longer see clearly is not done — it is just yours. Distance, whether King's six weeks or a weekend, is what restores vision.
What is each draft actually for?
Assigning each pass a single job keeps revision from becoming a loop. A documented shape, assembled from the accounts above:
- First draft — existence. Get the story down without judging it, per Lamott's essay.
- Second draft — structure. Cut, reorder, and fix the spine; King's ten-percent cut lives here.
- Third draft — scenes and texture. Dialogue, interiority, description, scene-level logic.
- Fourth and later — line edits and continuity, ideally with outside eyes: an editor, agent, or trusted reader.
When does redrafting become avoidance?
When the same manuscript has been "draft one" for years, or when revision prevents the book from ever facing a reader. Writers' accounts are honest about this: Lamott's essay is partly about permission to be bad on the page so the book can exist at all. The working distinction most craft teachers draw — and it appears across the Writers at Work interviews — is between revising toward clarity and revising toward safety. The book that would fail with readers cannot be saved by a fortieth private pass.
What about the book that never feels finished?
Some books get set aside, and documented writers say so without shame — their published interviews contain shelved novels described matter-of-factly as apprenticeship. The number of drafts a book needed to be ready and the number it received before you moved on are different facts. Deciding which one you are facing is the actual craft judgment, and it cannot be outsourced to a count.
What the documented record shows: three to five full drafts, each with a job, ending when the changes stop being structural. What it does not show is a rule — the writers quoted here disagree with each other, which is itself the lesson.
For more context, read How a Book Advance Actually Gets Paid Out.