Research · May 18, 2026 · 6 min read · by the Pressfold team

Survey design for PR without breaking the data

The fastest way to get a quotable statistic is to commission a survey. It's also the fastest way to produce a number that falls apart the moment a reporter looks twice. The tension is real: a survey designed purely to manufacture a headline tends to break, and a survey designed for academic rigour tends to produce nothing punchy. Good PR research lives in the overlap.

Write questions you'd be comfortable defending

The most common failure is the leading question — phrasing that all but dictates the answer, then reporting the answer as a discovery. "Are you worried about rising costs?" will always return high worry. A defensible version asks people to rank concerns or pick from a balanced list, so the finding reflects what they think rather than what you nudged them toward.

Decide the headline you're testing, not the headline you want

It's fine — necessary, even — to design a survey around a hypothesis. "We think remote workers underestimate how often they're online after hours." What's not fine is deciding the result before the data arrives. Frame it as a question the survey can answer either way, and be ready to pitch the surprising version even if it isn't the flattering one.

Mind the sample

• Size: a few hundred respondents is usually enough for a directional finding; below that, margins of error swallow the story.
• Source: a representative panel beats your own newsletter list, which is biased toward people who already like you.
• Segments: the headline often lives in a subgroup — but only quote a subgroup if it's large enough to mean something.

Report the method, not just the result

Always publish the sample size, the field dates, the provider and the exact question wording. This feels like handing critics a weapon; it's the opposite. A visible method tells a reporter the work is honest, and it's the single fastest way to move from "interesting claim" to "citable source." The campaigns that hide their method are the ones that get retracted.

When the data says nothing

Sometimes the survey comes back flat — no surprise, no angle. The professional move is to say so and look again, not to torture a sub-segment until it confesses. We'd rather refield a question than pitch a finding we can't stand behind. A null result is cheaper than a debunking.