Research prospective members properly before deciding what to build

Audience research should tell a community team who is trying to do what, where the present route fails and whether participants want help from peers. It should happen before a platform decision. Otherwise, the research is easily reduced to asking people which features they like in a product somebody has already chosen.

This guide is for organisations investigating a customer, professional, membership or peer-support community in England. It draws on UK public guidance and identifies where a commercial team must create its own evidence. No official dataset shows how many English people want to join a particular community.

What to take away

Frame one decision

Begin with the choice the research must inform. Examples include whether to open a member space, which group to serve first, whether discussion should be public, or which activity deserves facilitation. A broad aim such as "understand our audience" does not tell the researcher when enough has been learned.

Write the current belief separately from the question. For example:

This wording leaves room for a non-community answer. Better onboarding, clearer documentation or a direct service change may solve the problem more safely.

Set the geographic meaning of England. The relevant condition may be residence, workplace, service eligibility or the buyer's operating area. A UK-wide respondent group cannot quietly become England-specific evidence. If the sample covers several nations, preserve that fact in every conclusion.

Map the people around participation

Prospective members are only one research group. Moderators, subject experts, customer-support staff, safeguarding leads, data owners and budget holders can determine whether the proposed experience is viable. Their needs should not replace the member view, but they expose constraints a participant may never see.

Create a recruitment grid from real differences that could change the design. For a customer forum, those might include experience with the product, accessibility needs, organisation size and past use of support. For a professional network, they could include career stage, region, employment status and the ability to discuss work publicly.

Do not add attributes simply because they appear in a persona template. Collect sensitive information only when it has a defined research purpose and appropriate handling. The Information Commissioner's Office explains that personal data processing must be lawful, fair and transparent. A research incentive or consent form does not make unnecessary collection proportionate.

Recruit beyond the easiest participants

Existing superfans and vocal contributors are easy to find. They are also unusually familiar with the organisation. Include quieter users, recent joiners, people who left, people who chose another route and eligible people who never joined.

The Government Digital Service advises teams to recruit actual or likely users of a service and warns against relying on colleagues as stand-ins. That guidance was written for government services, not commercial communities, but the sampling problem travels well.

Use an eligibility screener with observable conditions. Ask what the person has done recently rather than whether they consider themselves "engaged". Keep a record of recruitment channels because a panel, customer email and public social post may reach very different people.

Sample size follows the decision and method. Qualitative interviews can reveal distinct needs and language; they cannot measure prevalence in England. A survey can count answers among its respondents, but a convenience sample does not automatically represent the wider population. State the sample frame and response pattern beside any percentage.

Pay and support participants fairly

Decide incentives before recruitment and apply the policy consistently. Payment recognises time; it does not buy a favourable view or continuing membership. Explain any effect on benefits, tax or employment only when qualified evidence supports that advice. Otherwise, tell participants what will be paid and invite them to check their own circumstances.

Offer practical adjustments without requiring people to disclose more than necessary. Alternatives may include a telephone session, captioning, extra time, a support person or materials in advance. Record which formats were available because a single research channel can determine who appears in the evidence.

Researchers should also plan for distress or accidental disclosure. Give participants a genuine way to pause or leave, and make clear which support the research team can and cannot provide. Incentive records, session notes and recruitment details do not need identical access or retention periods.

Combine methods for different questions

Observe existing behaviour

Start with support tickets, search terms, event questions, complaints and abandoned journeys where access is lawful and appropriate. These records show what people did in a particular setting. They do not reveal every motive, and the loudest problem may be overrepresented.

Create a small coding scheme before reviewing a larger sample. Record the task, obstacle, current workaround, consequence and whether another person could plausibly help. Retain examples that contradict the emerging pattern.

Interview recent experiences

Ask participants to describe a specific event: the last time they sought advice, tried to meet a peer or decided not to post. Follow the sequence and materials they used. Questions about a hypothetical future community invite polite speculation.

Government guidance on planning user research recommends selecting methods according to what the team needs to learn. In a commercial setting, document the same connection between the decision, participant group, method and analysis.

Avoid leading with a platform demo. Once a participant sees the proposed solution, they may comment on its interface instead of explaining the original problem. Show concepts later, after the present behaviour is understood.

Use a survey for bounded measurement

A survey works when the team knows which answer categories matter and needs to estimate their distribution within a defined sample. It is weaker for discovering unfamiliar problems. Pilot questions with people from the intended audience, test completion on relevant devices and include a route that does not force a misleading answer.

Do not report precision the design cannot support. An online customer survey excludes non-customers by definition and may underrepresent people who no longer open company emails. Weighting does not cure an unknown or badly biased sample frame.

Test a small community activity

Research can include a time-limited discussion, facilitated session or peer-matching experiment. The test should answer a question about behaviour, such as whether participants return to help one another without repeated staff prompting.

Tell participants what is experimental, who can see their contributions and what happens afterwards. The Government Digital Service has specific guidance on seeking consent for research. Commercial organisations must adapt the process to their circumstances and obtain specialist advice where risks warrant it.

Make privacy part of the method

Separate agreement to participate in a session from any later use of contact details, recordings or quotations. Explain the research purpose, access, retention and withdrawal route in plain language. If a supplier recruits or records participants, establish each party's role and contractual responsibilities.

Minimise identifiable material in notes. A verbatim quotation can identify a person through context even when the name is removed. Government researchers are advised to plan notes and recordings so the team captures what it needs while protecting participants.

Community research can expose health, employment, financial or safeguarding information unexpectedly. Define how a researcher should respond, stop recording and escalate before sessions begin. A generic privacy notice is not an incident plan.

Analyse without creating a fictional average member

Bring evidence together by need and behaviour, not by the popularity of individual suggestions. A useful finding describes:

Count coded material only after the categories are stable enough to apply consistently. Keep quotations tied to their context and permission. Do not turn one memorable comment into a representative claim.

Personas can help teams remember different decision patterns, but every field should trace back to research. Composite personas need an explicit note that they are analytical tools, not profiles of real individuals.

Decide what the evidence permits

At the end, answer the original question. The result may support a pilot for one segment, reject a public forum, recommend a non-community intervention or require further research. Name the evidence that would change the decision.

A recommendation to proceed should specify the initial member group, job, activity, owner, safety boundary and measure. Registrations alone are insufficient. For example, a peer-support pilot might examine whether members receive an accurate answer within an agreed period and whether unanswered cases reach staff.

Keep the research record usable. Store the brief, screener, participant mix, method, dates, consent records, notes, analysis and decision according to the organisation's retention policy. Mark England-only evidence separately from UK-wide or international material.

Review before implementation

Ask someone outside the project to challenge recruitment bias, unsupported generalisation and the proposed interpretation. Higher-risk subjects may need privacy, legal, accessibility or safeguarding review. Research quality is not established by the number of interviews alone.

Revisit the findings after launch. Member behaviour changes as the group grows, rules become visible and incentives shift. Continued learning does not mean reopening every decision weekly; it means recording which assumptions remain uncertain and checking them when relevant evidence appears.

A sound audience study gives the organisation permission to narrow its plan. It replaces a vague promise of engagement with a defined group, a real unmet need and a decision another reviewer can follow from evidence to action.

Before you act

Common questions

When should audience research happen?

It should happen before a platform decision. Otherwise the research is easily reduced to asking people which features they like in a product somebody has already chosen. Begin with the choice the research must inform, such as whether to open a member space or which group to serve first.

Why is it a problem to rely on existing superfans?

Existing superfans and vocal contributors are easy to find, but they are unusually familiar with the organisation. Include quieter users, recent joiners, people who left, people who chose another route and eligible people who never joined. Government guidance warns against relying on colleagues as stand-ins.

Can a survey measure how many English people want a community?

No. A survey can count answers among its respondents, but a convenience sample does not automatically represent the wider population. An online customer survey excludes non-customers by definition, and weighting does not cure an unknown or badly biased sample frame. State the sample frame beside any percentage.