OUTREACH & PR

Find data story angles in your own numbers

Use when you have internal data and suspect there is a press story in it but cannot see which one.

data-story-angles-from-own-data.md

Download .md

You are a data journalist assessing whether a dataset is worth a story.

Dataset description: {{WHAT_THE_DATA_MEASURES}}
Sample rows or summary statistics: {{DATA_SAMPLE}}
Time period covered: {{TIME_PERIOD}}
Geography or segment breakdown available: {{BREAKDOWNS}}
Target publications: {{TARGET_PUBLICATIONS}}

Produce a table with these columns:
Angle | The claim in one sentence | Which cut of the data proves it | Why a reader outside my industry cares | Publication most likely to run it | Strength (1-5)

Rank rows by Strength descending. Strength is judged on: is the finding counter-intuitive, is it local or personal to a reader, does it connect to something already in the news, and is the sample large enough to defend.

Then add:
- Two angles you considered and rejected, with the reason.
- The three questions a skeptical editor will ask about methodology.
- Any claim in the table that the data as described cannot actually support. Say so explicitly rather than softening it.

Do not invent figures. If a number is needed and not supplied, write NEEDS FIGURE.

Fill in before running

Replace each placeholder with your own detail. The more specific you are, the less the model invents.

{{WHAT_THE_DATA_MEASURES}}
{{DATA_SAMPLE}}
{{TIME_PERIOD}}
{{BREAKDOWNS}}
{{TARGET_PUBLICATIONS}}

Getting a better result

Give it the sample size and date range up front - most weak angles die on sample size and it will tell you which ones.
The rejected angles column is often more useful than the accepted ones, because it shows what the data cannot do.
Run it again with a different target publication list; the same data ranks completely differently for trade press.

Questions about this prompt

When should I run this rather than pitching the dataset as it stands?

When you can see the numbers are interesting but cannot say which finding a reader outside your industry would care about. The prompt sorts angles on counter-intuitiveness, personal relevance, connection to current news and defensible sample, which is a different question from what your own team finds interesting.

What does it need about the dataset?

Summary statistics or sample rows, the TIME_PERIOD, the sample size, and whatever BREAKDOWNS by geography or segment exist. Also a real TARGET_PUBLICATIONS list, because the same data ranks completely differently for a national title and for trade press. Weak angles mostly die on sample size, so give that up front.

What does the table give me, and what should I read first?

A ranked table with the claim, the cut of data that proves it and the likeliest publication, then rejected angles, the questions a sceptical editor will ask, and any claim the data cannot support. The rejected angles and the unsupported claims are the useful bit, because together they mark the edge of what you can safely say.

What is the mistake that kills an angle?

Treating a strength five row as ready to pitch while its NEEDS FIGURE markers are unresolved. The prompt refuses to invent numbers, so a strong angle can still be missing the figure that carries it. Pitching with an approximation is what gets picked apart by the first editor who asks about methodology.