How to Know if a Scene Is Necessary (and What to Do)

A scene is necessary when something is different at the end of it and the book would break if it were gone. Different means a goal was won or lost, a relationship shifted, the stakes rose, or the reader learned something they will need later. If none of that happens, or if another scene already does the same job, the scene is a candidate for revision.

Cutting is only one of four answers. A weak scene can also be merged with a neighbour, compressed into a paragraph of summary, or strengthened so it finally does the job it was written for. The tests below tell you which problem you have, and the decision table tells you which fix fits it.

StoryWith AI is our own product, so weigh what we say about it accordingly.

Key takeaways

• Ask what changes between the first line and the last. "We learn more about her" is not a change. "She stops trusting her sister" is.
• Run the removal test: take the scene out on paper and list what breaks downstream. If nothing breaks, the scene is not carrying weight yet.
• A common fault is duplication: two scenes that do one job. Those often merge rather than die.
• A scene the reader needs the outcome of, but not the moment by moment of, should usually become summary.
• A scene with a real purpose and no conflict should be strengthened, not cut.
• Beta readers tell you where attention fell off. That location is evidence; their suggested fix is a guess.

What makes a scene necessary?

The classic tool here is Dwight V. Swain's scene and sequel. K.M. Weiland summarises it on Helping Writers Become Authors (post dated 12 December 2016) as two halves of three parts each: "scene: goal, conflict, disaster; sequel: reaction, dilemma, decision." She opens the post with a line worth keeping on the desk: "Stories live and die on their scene structure."

You do not have to write to the formula to use it as a test. Hold any scene up against these questions and answer each in one sentence.

The last row comes from the same Weiland post, which suggests asking "What question is this scene asking?" and checking that the question raised at the start is the one answered at the end.

A scene needs a clear yes on at least one row, and the strongest scenes hit three or four. A scene that only scores on "reader knowledge" is the one most at risk of turning into exposition; our guide to spotting and fixing info dumps covers that case in detail.

What is the "remove it and see what breaks" test?

It is the quickest honest check there is. A post on Jessica Brody's blog (dated 28 February 2023) calls it the "Lift Out Test" and puts it plainly: "Cut it and see if the rest of your story collapses like a house of cards." The same post lists five reasons scenes get cut in revision: duplicate information, scenes that kill the pace, repetitive actions or events, a tone that does not match the book, and scenes that do not move the plot forward.

Run it on paper:

1. Write the scene's number at the top of a page.
2. List every later scene that refers to something that happens only here: a fact, an object, a promise, an injury, a line of dialogue that gets called back.
3. For each one, note whether the reference could be planted somewhere else in a sentence or two.

If the list is empty, the scene is carrying nothing. If every item on it could move in a sentence, the scene is carrying very little. If even one item cannot move (a death, a betrayal, the moment two characters meet), the scene stays, and your job is to make it earn its length.

One caution: the removal test is blind to scenes whose job is emotional. A quiet reaction scene after a disaster (Swain's sequel half) may have no downstream references and still be why the next chapter lands. Before cutting one, ask whether the character's next decision makes sense without it.

What is the difference between a scene and summary?

A scene happens on the page in something close to real time: dialogue, action, the character's thoughts moment by moment. Summary tells the reader what happened over a stretch of time in a few lines. Both are legitimate, and choosing between them is a pacing decision; our guide to controlling the pace of your story treats scene versus summary as the first of its eight pacing tools.

If the reader needs the outcome but not the experience, summarise. If the reader needs to feel the moment turn (a choice, a confession, a loss), dramatise.

A hypothetical worked case: a 2,400-word scene where a detective interviews three witnesses, two of whom add nothing, can become about 200 words of summary for the dead ends plus a 600-word dramatised exchange with the third. That is 800 words in place of 2,400, so about 1,600 come back to the book.

Should you cut, merge, compress or strengthen the scene?

Match the symptom to the fix. Once you have run the tests above, a weak scene will often fit one row clearly.

If a scene carries a fact, plant the fact elsewhere before you cut. And if you choose "strengthen" for five or six scenes in a row, the trouble is probably structural, and it belongs in a structural pass first.

What does merging two weak scenes look like?

A hypothetical example, invented for this guide.

In a draft, Scene 14 has Maya visiting her brother's flat to borrow money. They talk about their late mother, and Maya notices an unopened letter from a bank on his table. The scene runs 2,100 words. Scene 16 has Maya and her brother at a café a week later. He admits he has lost his job, and they argue about whether to sell their mother's house. That scene runs 2,600 words.

Run the tests. Scene 14 has a goal (borrow money) but she never asks, so there is no conflict and no outcome. Its only lasting contribution is the bank letter. Scene 16 has conflict and an outcome, but the revelation lands flat because the only clue was dropped in a scene where nothing was at stake, so the reader barely registered it. Each scene is doing half a job.

The merged scene: Maya goes to the flat to ask for money. Before she can, she sees the bank letter. She asks for the loan anyway, testing him. He refuses with an excuse she can see through, she pushes, and he admits he lost his job. The argument about the house happens there, in his kitchen, with the letter on the table between them.

The goal now meets resistance, and clue and reveal land in one movement. The two scenes totalled 4,700 words; a merged version of around 2,800 returns about 1,900.

The general method:

• Keep the stronger scene's conflict, and choose whichever setting puts the most pressure on it (here, his flat, where the letter is).
• Move the weaker scene's one essential element (here, the letter) into it, early.
• Let that element raise the pressure on the goal.
• Check every later scene that referred to either original, so nothing points at a conversation that no longer happens.

How do you reverse outline scenes for necessity?

Our guide to editing a novel in order explains how to build a reverse outline for the whole book: what happens, what changes, plants and payoffs, and a verdict. For the necessity question, use a narrower variant with columns aimed at the tests above. Fill one row per scene, and do not fix anything while you fill it in.

Read the finished outline in one sitting and look for patterns:

• A run of "None" in the conflict column, which is where a middle starts to drag.
• Two rows with identical entries under "reader learns", which are your merge candidates.
• A blank "breaks if removed" column next to a long word count, which is where the cheapest cuts are.

What do beta readers tell you about unnecessary scenes?

Readers rarely say "scene 14 is unnecessary." They say "it got slow around the middle." Treat that as location data, and look a scene or two before the point where they stopped: attention fades before a reader closes the book.

Signs worth tracking across your readers:

• Several readers stop or skim in the same chapter.
• A reader's summary of the book leaves out a scene entirely, which suggests it left no trace.
• A reader asks a question the cut candidate was supposed to answer, which suggests the scene's one job is not landing.
• Readers remember the reveal but not the setup, which can point to a setup scene that should be merged into the reveal.

If three of five readers stall in chapter 6, chapter 6 has a problem. Whether the fix is a cut or a merge is still your call. If you do not yet have readers, our guide on where to find beta readers lists the usual places to look.

Where StoryWith AI fits

StoryWith AI reads the manuscript you wrote and adds its findings to one ranked list of fixes, linked to the exact scene. A scene flagged for weak purpose appears there as a "Weak scene purpose" item. It does not write the book, and it does not cut, merge or rewrite scenes for you. The decision about every scene stays with you.

The part that bears on this question is scene purpose analysis. For each scene it names a primary purpose (and optionally a secondary one) from a fixed set: plot, character, world, reveal, stakes, payoff, conflict, theme. It scores the strength of that purpose from 0 to 100, and the analysis is instructed to flag "Weak purpose" when the score is under 50, when the scene reads like exposition without movement, or when its purpose is unclear or overloaded. It also tags patterns such as redundant, low_movement, info_dump or transitional, and offers 2 to 4 strengthening options phrased as options, not instructions. Treat it as directional analysis, not a verdict: a scene flagged as weak may be a deliberate quiet beat, and you are the one who knows that.

Scene purpose costs one AI text action per scene and runs on any plan. The emotional arc graph plots a tension score per scene and marks flat stretches (runs of three or more scenes with little tension change), which is useful for spotting the runs of low-conflict scenes the reverse outline turns up. On paid plans you can also invite your own readers to excerpts through real reader feedback sessions and see where each one stopped; StoryWith AI does not supply the readers.

It is the wrong tool if you want someone to make the call for you, or if you need a human editor's judgment of taste and intent. Its AI reader personas are a simulated first pass, not evidence of how real readers respond. For a short draft, the paper method may be all you need.

Frequently Asked Questions (FAQ)

How do I know if a scene is necessary in my novel?

Check what is different at the end of the scene: a goal won or lost, a relationship changed, the stakes raised, or information the reader will need later. Then imagine the book without it and list what breaks. If nothing changes and nothing breaks, the scene needs revision, which may mean cutting, merging, compressing or strengthening it.

Should I delete a scene I love that does not move the plot?

Not straight away. Check whether it does a job other than plot, such as a character's reaction to a loss or a setup for a later payoff. If it does, strengthen that job. If it does nothing, cut it and keep the passage in a scraps file for lines worth replanting.

When should two scenes be merged?

Merge when two scenes do the same job, such as revealing the same information or hitting the same emotional beat, or when each does half a job, like a clue in one scene and the reveal it sets up in another, neither working on its own. Keep the stronger scene's conflict, choose the setting that puts the most pressure on it, move the weaker scene's one essential element into it, and check later references.

What is the difference between a scene and a sequel?

In Dwight Swain's model, as summarised by K.M. Weiland, a scene runs goal, conflict, disaster, and a sequel runs reaction, dilemma, decision. The scene is the action; the sequel is the character processing it and choosing what to do next. A sequel can look unnecessary under the removal test because nothing later refers to it, so judge it by whether the next decision would make sense without it.

Should a weak scene become summary instead?

Yes, when the reader needs to know what happened but does not need to experience it moment by moment. Travel, routine tasks, repeated attempts and conversations that only confirm what the reader knows are common candidates. Keep the beats where something turns as dramatised scene, and summarise the approach to them.

Can software tell me which scenes to cut?

Software can flag scenes whose purpose looks weak, repetitive or exposition-heavy. It cannot know what you intend a quiet scene to do, so treat any flag as a prompt to run the tests yourself.