The method follows the question.

Understand what the research needs to answer

A request for “a survey” isn’t really a research brief yet. Before designing anything, I want to know what is already understood, where the uncertainty sits and what somebody might do differently once the research is complete. That affects almost everything that follows, from who should be included in the study to the questions worth spending respondents’ time on.

Sometimes that process reveals the research is being asked to validate an existing assumption, not answer an open question, and it’s better to know that before the work starts than after.

Design with the analysis in mind

The design starts with the evidence needed to answer the question. Sometimes that means primary quantitative research, sometimes interviews or other qualitative work, and sometimes the strongest answer comes from existing data, published research, company material or other secondary sources.

For quantitative work, I think about the comparisons we’ll need to make, whether the sample can support them, and whether the questions will produce distinctions that mean anything. A clever story can’t rescue a weak sample. I pay close attention to who is responding, how the questionnaire behaves in field, and whether apparent differences survive basic scrutiny.

Qualitative work gets at things a questionnaire often can’t: why people behave as they do, how they describe a problem in their own words, and which assumptions in the original brief don’t survive contact with reality.

Secondary research needs the same discipline. The real difficulty is rarely finding information. It’s deciding which sources are credible, comparable and relevant, what can safely be inferred from them, and where the evidence stops.

Be suspicious of the easy story

Analysis usually produces several possible narratives. Some are statistically stronger and more commercially interesting. Some simply make a better headline. They aren’t necessarily the same thing.

I look at how findings behave across different groups, whether apparently large gaps are based on enough evidence, and whether another part of the study contradicts the interpretation. I also look for findings that don’t fit the original hypothesis. They’re often more useful than the ones everybody expected.

You can’t remove judgement from analysis entirely. It has to stay anchored to what the research can actually support.

Not every finding belongs in the story

A large research programme can generate hundreds of potentially interesting numbers, but most audiences will remember very few of them. The job is to work out which findings change the answer, and which are supporting evidence.

The final form depends on what the research needs to achieve. Sometimes that’s a detailed report, sometimes a presentation or recommendation, sometimes a series of shorter pieces or an evidence base people can come back to.