Skip to content

The revision pass on a finished draft

Guide 22 of 22 For fiction writers 8-minute read

You finished. Ninety thousand words, a beginning and an end, and the advice from everyone is the same: put it in a drawer, wait, then revise. What nobody says is what revising is. Reading it again produces wincing and line edits — a better verb on page 40, a cut adjective on page 41 — and after a month of that the book is smoother and has exactly the problems it had before, because the problems were never in the sentences.

They were in the shape, and the shape cannot be seen from inside the prose. The revision pass is the act of getting outside it. The tool is the oldest one editors know, the reverse outline, and the reason to do it on cards rather than in a notebook is that a notebook holds the outline still and cards let you move it.

Everything in this guide works on paper. If you use software, the steps are the same, and the last section says what changes.

Step 1: Cut the draft into scenes

Not chapters — scenes. A chapter is a unit of reading; a scene is a unit of story, and most chapters hold two or three. Go through the draft and mark every place the story moves in time, place, or point of view. Each stretch between marks is a scene. Ninety thousand words is usually forty to seventy scenes.

One card per scene, its name at the top: The letter, Down to the harbor. If a stretch has no name — if you cannot say what it is about in three words — write ? and keep going. Those are the first finds.

Step 2: Answer the five questions from the prose, not from memory

This is the step that makes the reverse outline different from the outline you started with. For each scene, read it — actually read it, the whole thing — and answer, on the card, the five questions of guide 17: whose eyes; what she wants here; what is in her way; how it ends; what is different for her by the end.

Answer from what is on the page. Not what you meant. Not what the outline said. If the scene as written has Ada wanting nothing in particular, the card says nothing in particular, however clearly you remember what she was supposed to want. The gap between the outline you started with and the reverse outline you are making now is the revision. Every place they differ is a place to work.

Two hours for forty scenes, if you are honest. It is the most useful two hours of the whole process.

Step 3: The four reads

Lay the cards in a row, in the draft's order. Now read down the cards four times, a different line each time. Each read finds a different kind of problem.

Read one: the endings. Down the how does it end? line. Count the she gets it cards. A first draft is usually a third of them, and each one is a scene the reader could have skipped. Mark them; not all will go, but each needs either a cost or a cut.

Read two: the changes. Down the what's different by the end? line, reading each card against the one before. Safe → watched. Watched → afraid. Afraid → afraid. Afraid → afraid. Two in a row that do not move is the stretch where readers stop, and it is almost always in the second quarter, and it is almost always invisible from inside the prose, because the prose in those scenes is often the best in the book — you were comfortable, and it shows.

Read three: the threads. Build the grid of guide 19 — threads across, scenes down, a mark where each thread moves. Read the columns. The locket's long silence; the strike that stops; Ada and Ned resolved at scene 22 of 40. And the tension down the side: the twelve scenes in the middle that are all a five.

Read four: the promises. This one needs a second pass through the cards with a different question: what does this scene promise? A locked drawer, a name mentioned and not explained, a gun on the mantel, a character who says I'll tell you when this is over. Each is a promise to the reader, and each gets a small card of its own, clipped to the scene that made it. Then go through again asking what does this scene keep? — where the drawer opens, the name is explained, the gun goes off — and clip those to their scenes. Now match them. Every promise should have a keeping, later; every keeping should have a promise, earlier. A promise with no keeping is the thing readers mean when they say what happened to the locket? A keeping with no promise is worse: a payoff for a setup that was never made, which reads as a coincidence, and it is usually a setup you cut in the second draft without noticing what it was holding up.

If the book has secrets, there is a fifth read — the ledger of guide 20, who knows what and when — and for a mystery it is the one that matters most.

Step 4: The pile

After the four reads, some cards have marks on every line: no want, no change, no thread, no promise kept or made. Those are the cut. Take them out of the row and put them in a pile at the side, with the reason on the back — does nothing, duplicates scene 9, belongs to a different book — and the date. Do not throw them away. Guide 16 is about why: a cut on cards is a move, and the scene you cut in January is the scene you will want a piece of in March.

Then look at the row with the gaps in it. Some gaps close on their own — the scenes on either side were always adjacent in story time and the cut scene was between them by accident. Some do not, and those are the scenes to write: the collision of want and need that guide 18 said had to exist and the reverse outline just proved does not; the middle beat of the locket; the end of the strike. Blank cards, with a name and the five questions answered, in the row where they go. You now have an outline for the second draft that is made of the first draft's real scenes plus the ones it was missing, and not one word of prose has changed yet.

Step 5: Move things

The row is on a table for a reason. Pick up the second quarter — the flat stretch — and try it with scene 17 after 23. Read down the changes again. Try the strike's beats a scene earlier each. Try the book opening at scene 3, with the first two folded in as memory. Each experiment costs thirty seconds and tells you whether the shape improved, before you touch a sentence.

Only when the row is right do you go back to the prose, and then the prose work is different from wincing: it is the specific work each card says it needs. Give this scene a cost. Ada learns about Fell Street here, not in scene 7. Cut the last page; the change already happened. Line edits come last, and they come to scenes that have earned them.

Step 6: Continuity, which is a separate list

One more read, unrelated to the shape: the facts. Eye color, the day of the week, how long the walk to the harbor takes, how old Ada was when her mother left. First drafts contradict themselves on these constantly, and readers who miss a sagging middle will write to you about a brown-eyed character who was green-eyed in chapter two.

On paper this is a list per character and per place, filled in as you read, and a note every time the draft disagrees with the list. It is tedious and it is not optional.

Doing this in software

Software built on cards can do the first step for you: bring the draft in, split at scene breaks or headings, one scene card per stretch, with the prose on the back of each card and a check that every character of the original is accounted for. The reverse outline is then a set of cards with the five questions empty, which is exactly where step 2 starts — and step 2 is still yours, because answering from the page is the point.

The four reads, and the fifth, become a pass the software runs whenever you ask, over the cards you have answered, in plain sentences that point at scenes: "Tomas's room" changes the same thing in the same direction as "The letter", the scene before it. The thread about the locket goes quiet for twenty-three scenes. Readers will forget it. You promised something here and never came back to it. This pays something off, but nothing earlier is linked as the promise it keeps. This scene is through Ada's eyes, but it mentions something Ada doesn't know yet. And the sixth read, the tedious one, becomes the software's whenever the cards carry the facts: a character card that says green eyes, against every scene that says brown; a birth year, against every age the text states; a scene that opens the next morning whose date is four days after the scene before; a character in two places at the same story time; a full moon on a night that had none. Your forty-one drafted scenes are all about two thousand words long — length is one of the ways pace changes.

None of it is a verdict. Each is a fact about the cards, and the fix is yours. What the software changes is that the reverse outline is never stale: move a scene and its prose moves; fork the order to try the book two ways and compile both; cut a scene to the Cut, with its reason and date, and restore it in March. And when the second draft is done, the record of how it was made — what was cut, when, and why; what moved; what was written new — is kept in the cards' own dates, so that the question every editor asks, what did you change?, has an answer that is a list rather than a memory.

That is the revision pass. Not reading it again. Getting outside it, on cards, and moving the cards until the shape is right — and only then going back in.

The same method, on screen

Everything here works with a pack of paper cards. Leyden & Beam does the same steps on screen, keeps every order you try, and never writes a word for you.

Start your 14-day trial

Keep reading

Fiction · Guide 17 The scene card, and why a synopsis is not one Fiction · Guide 18 What she wants and what she needs Fiction · Guide 19 Threads: the plot grid on a table