01
1. Index the material before you look for a story
Start by making the material findable. Number the notebook pages, list each interview with its date and the subjects covered, and index the documents. Then build one table with a row for every usable piece of material: a scene, a quotation, a figure, a document, an observation.
Mark each row with its type and its status. A scene you observed, a scene described to you, a quotation on the record, a quotation unattributable, a figure with its source, an inference of your own. This labelling is the safeguard that stops reconstructed material presenting itself as observation three drafts later.
The index takes two or three hours on a week's reporting and it is not optional work. Every long piece that collapses in editing collapses because the writer could no longer say which material they actually had.
02
2. Find the argument, then the shape
A feature needs a single sentence stating what it shows. Not the subject, not the cast, but the thing the reader will know afterwards. Write it and put it at the top of your outline. If you cannot write it, you are not ready to draft, and more reporting is a better use of the day than more drafting.
The argument determines the shape. If the meaning is in a sequence, use chronology. If it is in a comparison between one experience and one institution, braid the two. If it is contained in a single loaded moment, frame the piece with that moment and use the middle for everything else.
Test the argument against your index. Every element of it should be supportable by specific rows, and if a clause has no supporting material you must either report it or drop it. This is the point at which a good feature quietly becomes a smaller, true one instead of a larger, shaky one.
03
3. Choose the scenes
Scenes are the expensive material: three hundred words each, sometimes more. A two thousand-word feature supports three or four. Choose the moments where something changed, where a decision was taken, or where the gap between an institution's account and a person's experience becomes visible.
For each chosen scene, check the index for enough material to build it: place, time, people, action, sensory detail, dialogue. A scene assembled from three thin notes will require invention to work, and that is exactly where features go wrong. If the material is thin, use summary instead and keep the scene for something you witnessed properly.
Order the scenes before you write any of them. Their sequence is the spine, and reordering three finished scenes is far more painful than reordering three lines in an outline.
04
4. Write the opening and the nut graf together
The opening is usually a scene or a specific detail, short and concrete, that raises a question without answering it. It should be true in every particular and should not promise a piece other than the one you are writing. Avoid the opening that is more dramatic than anything that follows.
The nut graf then states why this piece exists and why now, in three to six sentences: the scale of the thing, its timeliness, and what the piece will show. Place it between the third and eighth paragraph. Earlier and it summarises an article the reader has not begun; later and the reader has already left.
Write both, then read them together as if you were a reader who arrived by accident. Do they tell you what kind of piece this is and why you should stay? If the nut graf is doing all the work, the opening is decorative. If the opening is doing all the work, the piece has no argument yet.
05
5. Set a word budget and hold to it
Decide the length before drafting and divide it. In a two thousand-word feature a workable division is around three hundred words of opening, one hundred and fifty of nut graf, three scenes of roughly three hundred, four hundred of summary and context, and a two hundred-word ending. That is a plan, not a rule, and having one prevents the commonest long-form failure.
That failure is a piece where the first scene is superb and everything after it is compressed, because the writer spent forty per cent of the words on the first ten per cent of the material. Allocating in advance makes the choice visible while it is still cheap to change.
Track the budget as you draft. When a section exceeds its allocation by more than a quarter, stop and decide deliberately whether it deserves the extra words and which section will give them up. Deciding later means cutting your own favourite paragraphs under deadline.
06
6. Keep the line between reporting and writing
Long-form writing invites the habits of fiction, and journalism cannot borrow them. Do not invent people or dialogue. Do not merge two people into one. Do not attribute a thought or a feeling that was not described to you. Do not move an event in time. Do not present a reconstructed scene as one you witnessed.
Where you need an inner state, attribute it: the person said they felt, the person recalled deciding. Where a scene comes from an account, say so once, clearly, near the start of the scene. Where a name has been changed for protection, say that too, with an editor's decision behind it.
The test is simple and worth applying to every paragraph: could you show the sentence to the person it describes and defend where it came from? If the answer depends on them not noticing something, the sentence is wrong.
07
7. Draft in sections, out of order if necessary
Long pieces do not have to be written from the top. Draft the section you have the most material for first, because a finished section changes what the rest needs to do and gives you a real sense of pace rather than an imagined one. Keep each section in its own file or clearly separated block.
Leave marked gaps rather than stopping to check. A bracketed note recording the figure you need, the spelling you must confirm or the quotation you have not yet transcribed keeps the draft moving, provided every bracket is closed before anybody else reads it.
Between sections, reread the argument sentence at the top of your outline. Long drafts drift, and the drift is invisible from inside the paragraph you are writing. One reread every section costs a minute and prevents a wholesale restructure.
08
8. Do the outline test on the draft
When the draft is complete, write one sentence describing what each paragraph does and put those sentences on a single page. This is the most productive hour in long-form editing, and almost nobody does it voluntarily.
The page exposes what reading cannot: three consecutive paragraphs doing the same job, a strand of a braid dropped for eleven paragraphs, context arriving after the reader needed it, a scene that advances nothing. Fix at outline level by moving and deleting, before you touch a sentence.
Then check the three long-form specifics: the nut graf arrives before a reader would give up, every scene advances something, and every strand is picked up again within a readable distance. Reordering at this level takes an hour and saves a day.
09
9. Verify the long piece
A feature holds far more assertions than a news item, and they are harder to see because they sit inside description. Number them all, including every specific in every scene: the weather, the number of people, the time, the sign on the door, the colour of a document. Details in scenes are checkable facts and are checked exactly like any other.
Trace every scene detail back to your index. Anything that cannot be traced comes out, however good it is, because a detail remembered rather than recorded is the one that turns out to be wrong and undermines everything around it.
Then run the legal and ethical read: identification of people who were promised protection, adverse claims about identifiable people and whether a response was sought, third-party material and its permissions, and any detail about a child, a bereavement or a health matter. Refer anything doubtful to somebody qualified before publication.
10
10. The final read, and the file
Read the whole piece aloud. It is slow and it catches what silent reading does not: sentences that cannot be spoken, repeated constructions, a paragraph whose rhythm reveals that it is padding. Mark rather than fix as you go, then fix in one pass.
Check the furniture last, once the piece is stable: headline, standfirst, captions, any pull quote. Each must be supportable by the text, and a pull quote in particular must not change the meaning of the sentence it came from.
Then archive properly: the index, the notebooks, the recordings, the transcripts, the document folder, the verification notes and the version published, with a retention period and the terms agreed with any confidential source. A feature attracts questions months later, and the file is what answers them.
Allocating a feature's words
| Element | Words | Job | Sign that it is over budget |
|---|---|---|---|
| Opening scene or detail | 250-350 | Raise a question with something concrete | More dramatic than anything that follows |
| Nut graf | 120-180 | Why this piece, and why now | Reads as a summary of the whole article |
| Scenes (three) | 900 | Show change, decision or contradiction | A scene that advances nothing but atmosphere |
| Summary and context | 350-450 | Scale, history, figures, institutional account | Context arriving after the reader needed it |
| Ending | 150-250 | Show where things now stand | Tells the reader what to conclude |
Word-budget arithmetic
words per scene = scene allocation ÷ number of scenesscene capacity = total words × 0.45 ÷ 300over budget = section words ÷ allocated words × 100Worked example: a two thousand-word feature that gives forty-five per cent of its words to scenes has nine hundred words of scene, which supports three scenes of three hundred. Planning four scenes means either shorter scenes or fewer of the other elements.
Worked example: an opening drafted at 480 words against a 320-word allocation is at 150 per cent. That is the moment to decide which section gives up the difference, not after the whole draft exists.
Ten working days from notebook to filed feature
- Day 1Index everything: numbered notebook pages, interview list, document folder, one row per usable piece of material.
- Day 2Write the argument sentence. Test each clause of it against the index and cut what has no support.
- Day 3Choose the structure and the scenes. Order them in an outline before any prose is written.
- Day 4Fill the reporting gaps the index exposed: the missing figure, the second account, the document you never obtained.
- Day 5Set the word budget and draft the section with the strongest material.
- Day 6Draft the opening and the nut graf together, then read them as a stranger would.
- Day 7Draft the remaining sections, leaving bracketed gaps rather than stopping to check.
- Day 8Outline test on one page. Reorder and delete at outline level before touching sentences.
- Day 9Verify every assertion, including every specific inside every scene. Close every bracket.
- Day 10Legal and ethical read, read the whole piece aloud, write the furniture, file and archive the file.
Checklist before a feature is filed
- Material indexed with one row per usable item and its status recorded.
- Argument written in one sentence and supported clause by clause.
- Structure chosen deliberately and scene order fixed in an outline.
- Word budget allocated and tracked while drafting.
- No composite people, invented dialogue, moved events or unlabelled reconstruction.
- Nut graf placed between the third and eighth paragraph.
- Outline test completed on a single page after the draft was finished.
- Every scene detail traced back to the index, with untraceable details removed.
- Legal and ethical read done, with doubtful passages referred onward.
- Whole piece read aloud, furniture written last, file archived with a retention period.
Questions readers ask
How much reporting does a two thousand-word feature need?
Far more than the piece uses. A workable ratio is three to five times the material you publish, because selection is what makes a feature specific and a piece that uses everything gathered reads as a transcript.
Can I use a detail I remember but did not write down?
No. Remembered details are the ones that turn out to be wrong, and a wrong detail in a scene undermines the factual material around it. If it is not in the notes or on the recording, leave it out.
Is it acceptable to change the order of events for a better story?
No. You may choose where to start the narrative and you may cut, but presenting events in an order other than the one in which they happened misleads the reader about the record.