How to Edit Your First Draft: A 4-Pass Method From Structure to Commas
You edit a first draft in four passes, in this order: structure (does the outline hold?), characters (are they consistent and do they want something?), scene (does each scene do its job?), then sentence (repetition, rhythm, typos). The order is not cosmetic: each pass makes the previous one pointless if you swap them. Polishing the sentences of a chapter you will cut in the next pass is a week thrown away. This article follows the four passes on a single fictional manuscript, shows what each coach in the workshop sees at its own level and what it leaves for you to decide, then puts realistic numbers on a timeline for 80,000 words.
Why the order of the passes changes everything
A first draft contains three kinds of problems, and they are not fixed at the same level. A sagging middle is an outline problem: no sentence, however good, repairs it. A character who changes his mind for no reason is a profile problem: his motivation was never set down. A flat scene is a scene problem: it has no stakes or no turn. A repeated word is a sentence problem.
The editing checklists you find online (Jericho Writers has one, Savannah Gilbo describes four phases) often mix the levels: "check your adverbs" sits next to "check that your antagonist has a goal". They are not wrong, they are unsorted. The classic result: you spend three evenings thinning the adjectives in chapter 9, then discover during the structure pass that chapter 9 has to go. Four passes sort the work from heaviest to finest, so that nothing you do gets undone by the next step.
The manuscript used as the example here is called "The Last Round". First draft, 84,000 words, 27 chapters. Solène, a relief postwoman in a mountain village, keeps receiving mail addressed to a man who died three months ago. Around her: Anselme, the tobacconist who also runs the post counter; Nadia, her sister who stayed in the city; Yann, a gendarme from the neighbouring station. The novel is invented for this article; the problems in it belong to almost every first draft.
Before the first pass: rest, then one straight read
Two prerequisites, without which the passes spin in place.
Let it rest. Four weeks minimum without opening the file. Three days is not enough: you are still reading what you meant to write, not what you wrote. In the meantime, write something else, a short story for instance.
Read it straight through, without correcting. Export the manuscript, read it on a different device (e-reader, printout, phone), over two or three days, with a notebook beside you. Correcting a sentence is strictly forbidden. You only write down reader impressions: "bored here", "I don't believe this decision", "who is this guy?". Those notes become the raw material of passes 1 to 3.
For "The Last Round", the straight read produced 41 notes. The five that came back most often: the middle drags (chapters 11 to 17), Anselme is likeable and then menacing with no transition, you never know what Solène wants, the letter from chapter 4 is forgotten until chapter 22, and Nadia shows up three times for nothing.
Pass 1: structure (what Dédale sees)
The question for this pass: does the outline, as it exists in the written manuscript (not the one you planned), stand up? You do not touch the text. You work from a bird's-eye view: a table of chapters with, for each one, what happens, what changes, and how many words it weighs.
Here is the extract of the table for the problematic middle of "The Last Round", rebuilt after the first draft:
| Chapter | What happens | What has changed by the end | Words |
|---|---|---|---|
| 11 | Solène does her round, thinks about the dead man | Nothing | 3,400 |
| 12 | Dinner at Anselme's, village memories | Nothing | 3,900 |
| 13 | Nadia calls, argument about the inheritance | Solène hangs up | 2,800 |
| 14 | Solène opens a letter (illegal) | She knows the dead man was expecting someone | 3,100 |
| 15 | Round, rain, description of the valley | Nothing | 3,600 |
| 16 | Yann drops by, coffee, flirting | Nothing | 2,700 |
| 17 | Solène goes back to the dead man's house | She finds the second key | 3,200 |
Three chapters out of seven change nothing in the state of the story. That is not an opinion, it is an empty column. The structure pass consists of looking at those columns and deciding: merge 11 and 15 (a single round, which leads into the discovery in 14), move 16 after 17 (the flirting means something else once Solène is hiding the key), cut 12 or give it a function (Anselme lets slip that he knew the dead man better than he said, which sets up his turn).
In the workshop, this is Dédale's job. He reads your outline and your word budgets, not your prose, and he reports what he sees: "Chapters 11, 12, 15 and 16: no story variable changes. The middle is 26% of the manuscript for a single event (the opened letter). The central pivot, if that is the discovery of the key in chapter 17, lands at 63% of the text; is that intended?" He does not rewrite the outline. He does not propose a new chapter. He shows you the empty column and leaves you to choose between merging, moving and cutting.
D Daedalus · Story architect Pre-recorded demo
Daedalus stress-tests the outline of "What the Sea Gives Back"
Daedalus
Your feeling is right, and the outline shows why. From chapter 9 to chapter 22, Maud looks for a buyer. Looking is not an event: it is a state. Nothing stops her from looking for thirteen chapters.
Three questions, and no answers on your behalf:
- Where is the midpoint? Something around chapter 14 has to make going back impossible. A buyer who signs, then pulls out? A discovery about the father?
- What does Jonas want? He appears in Act II with no goal of his own. A character who wants nothing does not create conflict, he creates a pause.
- Why does the cannery burn? If it is an accident, the climax comes from outside. If it is the consequence of a choice Maud makes in Act II, the whole middle tightens at once.
Daedalus
Then your Act II has just found its spine: Maud needs the money of the man who wants to destroy what she is trying to save. Your midpoint almost writes itself (she signs, or refuses to sign), and the fire in chapter 23 changes meaning depending on who holds the match.
I am not updating anything: it is your outline. Add the line "Jonas = buyer" to chapters 9 to 16 and look at what each chapter now has to contain.
Pre-recorded simulation. Daedalus never writes: it questions the structure of the novel and leaves the decisions to the author.
What you produce at the end of this pass: a new chapter table, with merges and moves marked, and a list of scenes to write (few) or to cut (usually more). For "The Last Round", the table goes from 27 to 24 chapters, and the total target drops to 78,000 words.
Pass 2: characters (what the Archivist sees)
The question: does every important character want something specific, and do their actions follow that desire from start to finish? This pass is done on the profiles, not on the text. If you have no profiles, now is the time to build one per main character, rereading only that character's scenes.
Solène's profile, after that targeted reread:
Solène, 34, relief postwoman. Wants: (first draft) nothing clear. (Revision) To prove she can hold a post on her own, after leaving hers in the city for a reason she does not say. Fears: being sent back to the city, to her sister's. Contradiction: to keep her job she must follow the rules; to understand the letters she must break them. Scenes where she acts against her desire: chapter 8 (she refuses to help Anselme for no reason), chapter 19 (she accepts Yann's invitation right after deciding to hide everything).
Once written down, the contradiction becomes the engine of the novel: each opened letter brings Solène closer to the truth and to professional misconduct. In the first draft she opened letters because the plot needed her to. After pass 2, she opens them because she has a reason to and a reason not to.
Anselme's profile fixes the switch flagged during the straight read. He does not need to become a different man in chapter 20; he needs to have, from chapter 3, an interest in the letters staying unread. One line in his profile ("has kept the post counter's accounts for fifteen years; the letters contain money orders") gives three lines of dialogue to plant in chapters 3, 12 and 14.
In Sagoræ, the Archivist keeps the profiles and flags the gaps between what they say and what the text does: "Solène is described as reserved (profile) but starts conversations with three strangers in chapter 6; Yann has grey eyes in chapter 5 and blue eyes in chapter 21; Nadia is called the elder sister, then the younger." He changes neither the profile nor the text. You decide, and often you find that the gap points to a real problem (a reserved Solène who talks to everyone may simply not be reserved, and the profile is what needs changing).
What you produce: updated profiles, a list of lines or gestures to plant in specific chapters, a list of factual inconsistencies to fix. Nadia, who appeared three times for nothing, gets a function (she is the one who knows why Solène left the city) or disappears. Here, she stays, and her three appearances become three phone calls that climb in tension.
Pass 3: scene (what Théo sees)
Only now do you open the text. Chapter by chapter, scene by scene, with a single question: does this scene do its job? A scene has a job when someone in it wants something, meets resistance, and leaves in a different state from the one they came in with. Scenes that fail those three conditions are either cut (pass 1 already flagged many) or redirected.
Chapter 14 of "The Last Round", the first opened letter, in the first draft:
Solène set the bag down on the kitchen table. It had been raining since morning. She put water on to boil, looked out of the window at the church roof, thought about Nadia, about the city, about everything she had left behind. Then she took out the letter. The envelope was blue, slightly warped by the damp. She turned it over several times. Finally she opened it.
The scene's job is there (she opens the letter), but nothing resists. No stakes are set before the gesture; we do not know what she risks. Théo, in the workshop, does not rewrite this passage. He asks questions: "What does Solène lose if someone sees her open this letter? Does the reader know that at this point? Why does the scene start with the rain rather than the bag? How much time passes between the bag being set down and the envelope being opened, and does that time serve the hesitation or the padding?"
The author's revised version, after those questions:
Solène set the bag on the table and pulled out the blue letter before she had even taken off her coat. Three months she had been setting it aside every morning and putting it back in the bag every night. The rule was one line long, pinned above Anselme's counter: undeliverable mail is returned within eight days. She was on day ninety. She slid a nail under the flap.
The scene is shorter, the stakes (the rule, the deadline already blown, Anselme who displays the rule) come before the gesture, and the hesitation has become a number rather than an envelope turned over several times. None of it was written by the coach. A complete example, on a whole chapter, with the passages the author deliberately kept, is worked through in before and after: a chapter revised with coaching.
What you produce: a manuscript where every scene has a goal, a resistance and a change. This is the longest pass, because it is the one where you really rewrite. Expect heavy rewriting on a quarter of the scenes, light rewriting on half, and nothing on the remaining quarter.
Pass 4: sentence (what the Checker sees)
The last pass works on what the reader sees: repetition, filter words ("she felt that", "he noticed that"), crutch adverbs, rhythm, tense consistency, dialogue punctuation, spelling. You do it last for a simple reason: it is the only pass whose work is lost if the text is still moving.
Two tools take turns here. The built-in checker, no AI, free and unlimited in Sagoræ, underlines while you reread: agreement, typos, punctuation. The AI Checker reads a chapter you have declared finished and flags what the first one cannot see: "stood up" appears eleven times in chapter 14; three consecutive paragraphs start with "She"; the past tense slips into the present halfway through a scene; a "suddenly" precedes every event. It explains why it is a tic and replaces no word.
On the revised chapter 14, its notes fit in four lines:
- "Set", "pulled", "slid": three simple past verbs, good rhythm. "Was one line long": stative, justified.
- "Every morning", "every night", "every": intended repetition? It works as a swing; keep it if it is a choice.
- Missing comma after "counter" in "pinned above Anselme's counter undeliverable mail".
- "Day ninety" is written as a word here; "day 90" appears in figures in chapter 9.
This pass is done with a list in front of you. The forty points to check, sorted by pass, are gathered in the self-editing checklist for fiction. Some are ticked for you by the checker; the others are yours.
How to know a pass is finished
The question that derails most revisions is not "where do I start" but "when do I stop". Without an exit condition, you stay on pass 1 for two months moving chapters around, or you reopen the structure in the middle of pass 4 because a doubt came back. Each pass therefore has a verifiable exit condition, and one rule: once a pass is closed, you only go back to it on outside feedback (a beta reader, an editor), never on a worry.
| Pass | Exit condition | What you stop doing afterwards |
|---|---|---|
| 1, structure | No row of the chapter table has an empty "what has changed" column; the central pivot lands between 45 and 55% of the budget | Moving or cutting a chapter |
| 2, characters | Every main character has a written desire, fear and contradiction; the list of factual inconsistencies is empty | Changing a profile |
| 3, scene | Every scene has a goal, a resistance and a change of state, noted in one line in the scene summary | Rewriting anything longer than a paragraph |
| 4, sentence | The built-in checker underlines nothing; the AI Checker has no remaining note on repetition or tense | Reopening the text before reader feedback |
On "The Last Round", pass 1 was closed while the author was still hesitating over where chapter 16 belonged (before or after the key). She wrote the hesitation in her revision log and moved on. In pass 3, while rewriting chapter 17, the answer came on its own: the flirting with Yann only has tension if Solène already has the key in her pocket. Going back to pass 1 would have cost a week; letting it ripen cost nothing.
That revision log is a plain file kept next to the manuscript, where you note at each session the current pass, what you decided and what you postponed. When you come back after ten days off, it tells you where you were without rereading fifty pages. And when the urge to start over from scratch arrives (around week six of pass 3, usually), it shows you the decisions already made and why you made them.
A realistic timeline for 80,000 words
The figures below assume someone revising alongside a job, about ten hours a week. They come from workshop experience, not from a statistic; adjust them to your pace, but keep the proportions.
| Step | Duration | What you produce |
|---|---|---|
| Rest | 4 weeks | Nothing (another project) |
| Straight read | 1 week | 30 to 50 reader notes |
| Pass 1, structure | 2 weeks | New chapter table |
| Pass 2, characters | 2 weeks | Updated profiles, lines to plant |
| Pass 3, scene | 8 to 10 weeks | Rewritten manuscript |
| Pass 4, sentence | 3 weeks | Clean manuscript |
| Total | 20 to 22 weeks | Draft 2 |
Five months, rest included. Pass 3 takes half the time, and that is normal: it is the one where you write. Passes 1 and 2 look short because they do not touch the text, but they are the ones that save weeks in pass 3, by cutting scenes before you rewrite them.
Two scheduling mistakes come up again and again. Doing pass 4 continuously (fixing sentences during pass 3): you lose half the work, because the scenes are still moving. Skipping the rest: you reread with the author's eyes, see nothing, and conclude the manuscript is ready.
What comes after the four passes
Draft 2 is not the final draft. It is the one you can let people read. Three outside readers, a precise questionnaire, and a method for sorting their contradictory feedback: that is the subject of how to find beta readers. Their feedback usually triggers a partial pass 3 (a few scenes) and a fresh pass 4.
Then comes the question of the professional. An AI coach is the first filter: it flags, questions, explains. It does not replace a human editor, who reads with a market in mind and puts their name on the line. What the three levels of professional editing cost, and which one to choose depending on whether you are aiming for a publisher or self-publishing, is laid out in developmental, line and copy editing: who does what.
Finally, formatting: line spacing, font, title page, synopsis, cover letter. A manuscript revised in four passes but sent in Comic Sans with unnumbered chapters starts with a handicap that has nothing to do with its text. The rules are in preparing your manuscript for submission.
What the coaches do not do
It is worth saying again, because the question comes up at every pass: Dédale, the Archivist, Théo and the AI Checker never write in your place. Dédale shows the empty column; he does not propose a chapter. The Archivist flags the gap between profile and text; he corrects neither. Théo asks questions about a scene; he does not write a line of it. The AI Checker underlines and explains; the word stays yours. Every pass ends with a decision you make, and that is why draft 2 is still your novel, not an average.
Frequently asked questions
How long does it take to edit an 80,000-word novel? About five months at ten hours a week, rest included: a month of rest, a week of reading, four weeks for structure and characters, two to two and a half months for scenes, three weeks for sentences.
Do I really have to let the manuscript rest? Yes, four weeks minimum. Without rest, you read what you meant to write; the straight read then produces almost no useful notes.
Can I fix typos as I go? You can let the built-in checker underline while you write, it costs nothing. But the sentence pass proper (repetition, rhythm, tics) comes last, otherwise it is lost when the scenes move.
Does an AI coach replace an editor? No. It is the first filter, at every pass, and it leaves the decision to you. A human editor reads with a market in mind, takes responsibility and negotiates with you; no tool does that.
Which pass do I start with if I never had an outline? Pass 1 all the same: rebuild the chapter table from the written text. It is often the first time you see your novel whole, and that is where the empty chapters show up.
If you want to run these four passes in a workshop where the outline, the profiles, the text and the checker sit in the same place, Sagoræ offers a free trial, no card required.
✒️ Read next: How to Write Your First Novel: A Guide That Shows Instead of Tells · How to Start Writing a Novel With No Experience: Your First Week, Hour by Hour · 12 First Novel Mistakes and How to Catch Them Before Your Readers Do
Write your novel in a real workshop: a focused editor, characters, story structure and AI coaches that guide you without writing for you. Start for free, no commitment.