How to write a press release about a survey or report

Put the methodology in the release itself. A competent reader expects the client who commissioned it, the dates of interviewing, the method by which interviews were obtained, the universe effectively represented, and the base sizes the percentages are calculated on. If your sample is an opt-in online panel rather than a probability sample, do not quote a margin of error — give the sample size, the method, the field dates and the weighting instead.

Of all the announcement types, this is the one where a journalist has an actual checklist and will work through it. A survey release either supplies the methodology or it does not get used, and the release that omits it is not saving words — it is failing the only test that applies.

The checklist, published, by people who do this for a living

The British Polling Council's rules of disclosure require that findings entering the public domain carry:

Client commissioning the survey; Dates of interviewing; Method of obtaining the interviews (e.g. in-person, telephone, internet); The universe effectively represented (all adults, voters etc); The percentages upon which conclusions are based; Size of the sample and geographic coverage

Those rules bind BPC members conducting political and social polling in the UK, not every company that runs a customer survey. But they are the clearest published statement of what a competent reader expects, and adopting them voluntarily costs about forty words. A release carrying all six is immediately in a different class from one carrying "a recent survey found".

The BPC rules also contain a line that speaks directly to this category. Alongside media polls, they contemplate "those conducted to support the PR activities of a commercial client" — the industry's own standards documents recognise the PR survey as a distinct thing. Being recognised is not the same as being trusted, which is the gap the disclosure closes.

The UK's Market Research Society Code of Conduct added a related obligation in its 2023 edition, requiring the reporting of "sampling characteristics and parameters used when defining samples as representative of segments of the population" — that is, if you are going to call a sample nationally representative, you have to say what you mean by it.

Say "sample of 2,000 UK adults", not "margin of error"

Most commercial research is run on an online access panel. Panel respondents opt in; they are not drawn at random from a known population. A margin of sampling error is a statement about a probability sample, and quoting one for a panel survey imports a precision the design cannot support. Editors who cover data notice this, and it is the fastest way to lose the story.

What to write instead: the sample size, the method ("online panel"), the field dates, the universe and how the data were weighted. "2,014 UK adults surveyed online between 3 and 6 September 2026, weighted to national profiles for age, gender and region" is more informative than any error bar, and it is true.

Who paid for it has to be visible

If you commissioned the research and the research supports your product, that is a material connection. The FTC's endorsement rules address the case directly. 16 CFR 255.5 works through a worked example of a company that commissions an outside research organisation, sets the subject and pays the expenses while the researcher controls the protocol, then cites the results as that organisation's findings:

the weight consumers place on the reported results could be materially affected by knowing that the advertiser had funded the project. Therefore, the advertiser's payment of expenses to the research organization should be disclosed in the advertisement.

The practical version is one clause: "research commissioned by [company] and conducted by [agency]". That clause is also what makes the finding quotable, because it tells a journalist who to ring.

What counts as news in a research release

The finding, not the report. Three tests:

• Is there a number that surprises someone? A survey confirming what everyone assumed is a marketing asset, not a story.
• Is it about the world rather than about you? "58 per cent of finance teams still close the month in spreadsheets" travels. "72 per cent of our customers are satisfied" does not.
• Can someone else use it? The releases that get picked up are the ones that hand a journalist a statistic for a piece they were already going to write.

A tip that is only true here: lead with the finding, not the report. The headline "New research finds X" beats "Company publishes 2026 State of Y report" every time, because the second one is about a document and the first is about the world.

The shape that clears editorial

Gating the report behind a form is a legitimate commercial choice and a genuine cost: a journalist on deadline will not fill in a form, and neither will anything summarising your research later. If the point of the exercise is citation, publish the tables.

A headline and lede you could file

Six in ten UK finance teams still close the month in spreadsheets, survey finds

>

LONDON, United Kingdom, 14 September 2026 — Fifty-eight per cent of UK finance teams still perform their month-end close primarily in spreadsheets, according to research commissioned by Ledgerpoint and conducted by Ashcombe Research among 512 finance professionals at UK companies with 50 or more employees, surveyed online between 3 and 6 September 2026.

The base, the method, the field dates, the universe, the commissioner and the agency are all in two sentences. Nothing has been claimed that the design cannot carry.

What the announcement can and cannot do

It can make a number attributable and citable, which is the most durable thing this format produces — a statistic with a named source and a date attached gets reused for years, including by systems summarising a topic. What can and cannot be shown about press releases in AI answers is a bounded question, but a page containing a sourced statistic is the kind of page that gets quoted at all.

It cannot make weak research strong. It cannot survive a methodology question you have not prepared for, and the question will come from the one journalist you most wanted. It will not generate the coverage on its own: the pickup comes from the report plus a direct pitch, and the difference between a paid page and an earned one is stark in this category, because a genuinely interesting statistic is one of the few things that still earns real coverage. And it will not pass link credit to a gated report — press links are overwhelmingly nofollow, which is normal rather than a defect and published outlet by outlet where we have read it, and one more reason to make the report itself easy to reach.

The check before you send

• Sample size, method, field dates, universe, weighting and commissioner all stated.
• No margin of error quoted for a non-probability panel.
• Percentages carry their bases, including for subgroups.
• The agency that fielded it named, and briefed to take a call.
• The full tables published at a URL that resolves without a form.
• The headline is a finding about the world, not an announcement about a document.