Make a data-led media angle worth questioning

A useful statistic needs a relevant question, a fair comparison and a method readers can inspect. Build the story around what the evidence actually changes for its audience.

A company has data, so someone suggests turning it into a report. A headline is written, a chart is commissioned and the team begins searching for a number large enough to attract attention.

Reverse that sequence. Start with a question a particular audience would care about, then establish whether the available data can answer it. A narrower finding with a defensible explanation is more useful than a dramatic claim assembled after the analysis.

The editorial test is simple: what would a reader understand or decide differently after learning this? Name that reader before choosing the format, spokesperson or publication list.

Find the consequence behind the number

Imagine a business analysing appointment bookings. “Bookings changed” is a measurement. “Customers increasingly request evening appointments, while available capacity remains concentrated in daytime hours” suggests a practical tension worth investigating.

That second statement still needs evidence for both demand and capacity. If the dataset contains only completed appointments, it may not reveal unmet requests at all. The most interesting angle can expose a gap in the data rather than justify immediate publication.

Write the proposed finding in a sentence, followed by the evidence each part requires. This makes unsupported leaps visible before they become a headline that colleagues feel committed to defending.

Choose a fair comparison

A change between two periods can reflect a changing customer mix, a new product, different coverage or a revised definition. Establish whether the same thing was measured in a comparable way.

In a hypothetical dataset, average order value rises after a premium product launches. That does not by itself show that existing customers became more willing to spend. The product mix offers another explanation.

Show the denominator and the relevant period. “Twice as many” has little meaning without the starting value. A percentage calculated from a very small group can be arithmetically correct and still support a fragile story.

If the comparison cannot be made fairly, change the angle. Explain the new dataset or the pattern within a defined group instead of implying a broad trend that the evidence cannot establish.

Give uncertainty a place in the story

The ONS discussion of uncertainty explains why an estimate should be understood alongside its uncertainty. For a commercial report, the practical question is which limitations could change the reader’s conclusion.

Sampling uncertainty is only one issue. Missing records, self-selection, incomplete coverage and ambiguous questions can matter even in a large dataset. More rows do not automatically make the population representative.

State what the evidence covers and what it excludes. If respondents are customers of one platform, describe them that way. Do not silently promote their answers into a claim about every business or consumer in the country.

A limitation should be near the claim it qualifies. Hiding a decisive restriction in a distant technical appendix leaves the headline doing work the evidence cannot support.

Build the journalist’s working pack

Provide the finding, the relevant table, a plain-language method and contact details for someone who understands the analysis. Include definitions, collection dates and the rules used to exclude records.

Offer a chart that can be interpreted without a presenter explaining it. Label axes and units, identify the source and avoid visual scales that exaggerate a modest difference. Supply the underlying shareable figures when permission and confidentiality allow.

The Office for Statistics Regulation’s transparency principles concern public statistics and analysis. Their emphasis on accessible evidence offers a useful editorial standard for a company choosing how to explain its own research.

This is not a requirement to expose customer records or commercially sensitive information. Publish enough method and aggregated evidence to support the claim, while handling privacy and permissions through the appropriate internal process.

Add interpretation without claiming causation

A spokesperson can explain why a pattern may matter, what other explanations are plausible and which practical questions it raises. They should not convert a correlation into proof that one factor caused another.

Prepare for the strongest reasonable challenge. Why this sample? Why this period? What happened when outliers were removed? Was the question chosen before or after the results were visible?

If the analysis changes after challenge, correct the pack consistently. Keep a version record so the chart, release and spokesperson notes do not circulate with different numbers.

Pitch the right audience

Match the angle to the journalist’s beat and the reader’s concern. A local operating pattern may be valuable to a specialist publication even when it cannot support a national headline.

Lead with the finding and consequence, then make the evidence easy to inspect. Keep the company’s promotional message secondary to the value of the information.

The commercial benefit comes from becoming a credible source on a relevant question. That requires being willing to publish a modest finding, acknowledge an inconvenient limitation or decide that the current dataset is not ready to carry the story.