Formats and attention

1. Attention is a constraint, not an audience metric

Treating attention as something to be captured leads to production choices that borrow against trust: manufactured suspense, withheld conclusions, headlines that promise more than the article contains. Treating it as a constraint leads somewhere more useful. A reader has a finite amount of it, they have chosen to spend some on you, and the editorial obligation is to deliver the substance efficiently and let them leave when they have what they came for.

That reframing has practical consequences. It argues for putting the finding at the top rather than at the end, for structure that allows selective reading, for honest labelling of length, and for accepting that a briefing read in forty seconds has succeeded. It also argues against measuring every format by dwell time, which rewards inefficiency and penalises clarity.

2. Match the format to the central evidence

The reliable question at commissioning is: what is the core evidence of this story? A document argues for text with excerpts and links, because the reader may want to check the wording. A place argues for photography and mapping. A voice — testimony, an accent, a hesitation — argues for audio, because a transcript loses precisely what makes it evidence. A dataset argues for charts and a filterable table. An unfolding sequence argues for live coverage followed by a settled account.

Getting this wrong is expensive in a specific way: the format fights the material. A document story told as vertical video cannot show the clause that matters; a testimony story told as text discards the thing that made it worth reporting. Format-first commissioning — deciding to make a video and then finding a story for it — produces this failure reliably, and it is visible to readers even when they cannot name what is wrong.

3. Budgeting reading and listening time

Length can be estimated before a word is written, and doing so improves commissioning conversations. Silent adult reading of general prose is commonly referenced in the 200–250 words-per-minute range, with denser or technical material slower; broadcast and podcast delivery is usually planned around roughly 140–160 words per minute. These are planning conventions rather than laws, and they vary by reader, subject and language.

Publishing an estimated reading time is a small courtesy with a real effect: it lets a reader decide whether to start now or save the piece, which is a better outcome than an abandonment three paragraphs in.

4. Architecture for text

Long text is read selectively. Assume the reader will skim, scan the headings, stop at anything visually distinct, and read continuously only where the material earns it. Structure accordingly: an opening that states the finding, numbered or clearly titled sections that are each self-contained, one idea per paragraph, and tables or boxes for anything comparative.

Two conventions do most of the work. First, put the qualification with the claim rather than in a later paragraph, because a selective reader may never reach the later paragraph. Second, make headings informative rather than teasing: “Only two of the nine sites publish a correction log” tells a scanner something; “A surprising finding” does not. Informative headings also make the piece far more usable to anyone reading with a screen reader.

5. Audio structure

The ear cannot re-read. Attribution has to precede the claim — “according to the commission’s report, the figure rose” rather than the reverse — because a listener who hears the claim first has already accepted it. Numbers need rounding to something holdable, comparisons need to be concrete, and subordinate clauses need unpicking into separate sentences.

Interview audio adds duties. Consent to record and clarity about intended use come first. Editing must preserve the meaning of an answer, not merely its words: removing a hesitation is ordinary, removing a qualification is not. Publish a transcript with every episode. It serves listeners who cannot use audio, it makes the material searchable and quotable, and it gives the desk a correctable text record of what was actually broadcast.

6. Vertical video and its hard limits

A vertical clip is a demanding format precisely because it looks easy. It is watched once, on a small screen, frequently without sound, against infinite alternatives. The production rules follow: establish the subject in the first seconds, burn in legible captions, keep essential content away from the edges where interface elements sit, and treat the opening frame as the headline.

The editorial limit is qualification. Many important claims are only accurate with a condition attached, and a clip short enough to hold attention may not be able to carry the condition. When that happens the correct decision is to change the claim, not to drop the condition. If neither is possible, the story does not belong in the format, and saying so at commissioning is cheaper than discovering it in the edit.

7. Newsletters and live formats

A newsletter is the format with the most editorial freedom and the least correction latitude: sequence and selection are entirely the editor’s, but once sent it cannot be amended. Verification must therefore be complete before dispatch, and the correction route — a note in the next edition plus a corrected item on the site — should be stated in the newsletter itself so readers know it exists.

Live coverage inverts both properties. It is published continuously and amended constantly, which makes it fast and makes every intermediate error permanently quotable. Timestamp every entry, mark superseded entries rather than deleting them, maintain a current-state summary at the top, label unconfirmed material explicitly, and close the live blog with a settled article. An abandoned live blog left as the permanent record preserves every error it ever contained.

8. Accessibility as a format requirement

Each format carries a specific duty, and each duty is cheapest when planned at commissioning. Audio needs a transcript. Video needs accurate captions and, where meaning is carried visually, a description. Charts need a text alternative that states the finding rather than describing the picture. Interactive features need keyboard operation and a sensible experience without scripting. Live blogs need a heading structure that lets a screen reader move between entries.

Automatic captioning and machine transcription are starting points only. Both mishandle proper nouns, place names and technical vocabulary — exactly the words a news item cannot afford to get wrong — so a person must correct them before publication. Budget that time when the item is commissioned, not when it is due.

9. Measuring format performance without distortion

Different formats need different success definitions, set before publication. A briefing succeeds when the reader leaves quickly with the answer; a feature succeeds when a meaningful proportion reach the end; an interactive succeeds when readers use the interaction rather than scrolling past it; a newsletter succeeds when subscribers return next week. Applying dwell time to all four guarantees the wrong conclusion three times.

Record the expectation at commissioning and compare it afterwards. Over a few months this produces something more valuable than any dashboard: a written record of what the desk predicted, what happened, and how the prediction improved. That record is also the honest basis for retiring a format that is not working, which is a decision newsrooms find unusually hard to make.

10. A format review cycle

Formats accumulate. Most desks are running at least one that nobody has evaluated since it launched and nobody wants to be the person who ends it. A fixed review makes that decision routine rather than personal.

1. Step 1 — inventory List every format currently in production, with its publication rhythm and the person responsible.
2. Step 2 — restate the purpose Write one sentence per format saying what it is for and who it serves.
3. Step 3 — check the success definition Confirm each format is measured against its own stated definition, not a shared dwell-time figure.
4. Step 4 — cost the maintenance Estimate the recurring hours each format consumes, including captioning, transcripts and upkeep.
5. Step 5 — test accessibility Sample recent items and verify transcripts, captions, alternatives and keyboard operation.
6. Step 6 — decide Continue, change the rhythm, or retire — and write down the reason so the next review has a baseline.

Format commissioning checklist

• The central evidence of the story is named before the format is chosen.
• The format can carry the qualification the claim requires.
• Estimated reading or listening time has been calculated and, where useful, published.
• Headings are informative rather than teasing.
• Qualifications sit with their claims, not in a later paragraph.
• Audio scripts place attribution before the claim.
• Transcripts, captions and text alternatives are budgeted at commissioning.
• Interactive elements work by keyboard and degrade gracefully without scripting.
• Live coverage has a planned end point and a settled article to replace it.
• A success definition specific to the format is recorded before publication.
• Recurring maintenance hours have been estimated for at least ten editions.

The list is a commissioning tool. Applied after production it merely documents decisions that can no longer be changed cheaply.

Questions readers ask