Audience research for a lead network that separates both sides of the marketplace

Lead generation networks audience research should answer one operational question: which people experience the matching problem, under what conditions, and what evidence would change the product decision? A demographic label or fictional persona cannot do that work. Begin with a defined population, recruit transparently, observe how decisions are made, then preserve contrary evidence as carefully as favourable comments.

No interviews or surveys were conducted for this guide. It does not report demand, customer preferences or competitor behaviour. The method was assembled from records retrieved on 5 September 2026 and is designed for an England-facing service. UK data-protection and electronic-marketing guidance applies in England, while statistical sources must retain their published geography.

What to take away

Frame the decision before finding people

Write the decision in a form that can fail. "Should we release a request to several suppliers only after the customer sees their names?" is researchable. "Do people like convenient marketplaces?" invites agreeable but unusable answers.

Record four fields before recruitment:

The population is not automatically every business or adult in England. ONS publishes UK business activity, size and location data for 2025, including enterprises and local units by region and industry. Those counts can help describe a sampling frame. They cannot show who currently buys introductions, which firm has a particular acquisition problem or who will participate in research.

Separate the sides of the network

A matching service normally has at least two populations: people requesting help and businesses considering an introduction. Research them separately. A requester may care about how many firms receive their details, while a buyer may care about when a charge becomes final. One blended "user" persona hides that difference.

There may also be people who never complete the journey, employees who screen requests, and staff who handle objections or disputes. Include them only when the research decision concerns their task. Do not collect background characteristics merely because a survey tool provides the field.

For business participants, classify the sampling unit precisely. An enterprise, local branch, owner and employee are not interchangeable. ONS methodology describes sample frames, stratification, selection and weighting in its sample design overview. A convenience sample from a founder's contacts can support discovery, but it cannot support a claim about all businesses in England.

Recruit without disguising the purpose

An invitation should name the organisation, research topic, session type, time commitment, recording plan, incentive, eligibility screen and contact route. It should say that taking part is voluntary. If a recruitment partner is involved, explain its role and which details each party receives.

The GOV.UK Service Manual recommends recruiting actual or likely users, supplying an information sheet, collecting only the details needed for participation and considering disabled people during recruitment. Its participant recruitment guidance was written for government services, so it is a practical reference rather than a universal legal test for a commercial network.

Keep recruitment and promotion apart. ICO guidance says genuine market research is not direct marketing, but a survey containing promotional material or collecting details for later marketing is treated as direct marketing. Calling an acquisition form "research" does not change its purpose.

The participation choice is equally separate from permission to send offers or disclose an enquiry to suppliers. If marketing is proposed, provide its own explanation and assess the channel under PECR and the processing under the UK GDPR. Refusing marketing must not invalidate an otherwise eligible research session.

Make participation understandable

Before a session, tell the participant who conducts it and why. Explain what happens, what information you collect, whether anyone observes and whether you record it.

Explain how you will share findings, how long you will keep material and how to withdraw. The GOV.UK guide to informed consent for user research also recommends keeping evidence matched to the research record.

That participation record is an ethical and operational safeguard. It does not by itself decide the lawful basis for every use of personal data. A qualified reviewer should confirm the lawful basis, privacy information, controller and processor roles, special-category conditions where relevant, retention and withdrawal procedure.

The ICO's data-minimisation guidance requires personal data to be adequate, relevant and limited to what is necessary for the stated purpose. Screen with the smallest set of attributes that establishes eligibility. Store scheduling contacts separately from notes where practical, and remove direct identifiers from the analysis file unless the research question genuinely needs them.

Build a sample that exposes disagreement

For early interviews, use a purposive matrix rather than pretending to estimate prevalence. Rows might represent requester, buyer and operator roles. Columns could cover recent success, recent failure, abandonment and no prior use. Add English regions or business sizes only when the product decision could differ across them.

State how each participant was found. Existing customers, professional bodies, social posts, paid panels and direct invitations create different selection pressures. Record refusals and unreachable groups as limitations. A high number of completed interviews does not cure a biased frame.

For a survey intended to estimate a proportion, define:

reported proportion = eligible respondents giving the specified answer / all eligible respondents answering that item

Publish the numerator, denominator, fieldwork dates, recruitment source, exclusions, missing responses, weighting method and uncertainty. Do not describe an opt-in web poll as representative of England. ONS notes in its survey sample and response tracker that achieved interviews and response rates do not by themselves reveal bias or representativeness.

Use interviews for mechanisms, not percentages

Ask for a recent event. A buyer can walk through the last time it paid for an introduction: what arrived, what it checked, when it acted and what ended the process. A requester can reconstruct where it expected details to go and what contact followed. Concrete sequences reveal hand-off assumptions that general opinions miss.

Avoid questions that teach the proposition or request praise. "What happened after you submitted the form?" leaves room for surprise. "Would transparent matching make you trust us more?" names the desired answer. Ask the same core questions in the same order, but follow unexpected details when they bear on the decision.

Interview notes should distinguish what the participant said, what the researcher observed and what the team inferred. A single vivid comment can identify a risk; it cannot establish frequency. Do not turn an interviewee into a named persona unless the person agreed to that use and publication is necessary.

Design surveys around analysis

Draft the output table before the questionnaire. Each item needs a defined variable, response options, missing-value treatment and intended comparison. If no decision uses an answer, remove the question.

Use mutually intelligible choices and offer "not applicable" or "prefer not to say" where appropriate. Pilot the wording and routes with people who resemble the target population. The W3C accessible forms tutorial explains the importance of labels, grouped controls, instructions and feedback. It also advises asking only for information required by the process.

Accessibility is not settled by a vendor badge. Test the actual questionnaire with keyboard navigation, magnification and relevant assistive technology. Offer an alternative mode when an online form creates a barrier. Record which devices and routes were tested, by whom and on what date.

Select a tool after the method

Branching and exports are useful only when the research plan needs them. Microsoft Forms documents forward branching.

A separate Microsoft page documents opening responses in Excel.

Google Forms documents section routing by an answer.

Its response guide covers CSV download and Sheets linkage.

Jisc Online Surveys explains its question and page logic.

Jisc also lists CSV, XLSX and SPSS exports.

Those first-party pages describe functions, not suitability, accessibility conformance, data-protection compliance or research quality. Before procurement, test the planned questionnaire, access controls, deletion, export, collaborator permissions, contractual terms, sub-processors and exit route. No tool was accessed for this guide.

Analyse claims, not personalities

Create an evidence grid with one row per research question. Attach source quotations or observation references, contradictory examples, participant eligibility and researcher interpretation. For qualitative work, code independently where the decision is consequential, compare disagreements, then retain the coding notes.

For quantitative work, lock the analysis plan before reading results. Do not try several segment definitions and publish only the most attractive difference. Report small bases and missing answers beside every subgroup. If the sample was not designed to represent a wider population, keep the finding inside the recruited group.

Personas should emerge as provisional decision aids. Each statement needs a source tag such as interview, survey, operational record or hypothesis. The MRS Code of Conduct page identifies the 2023 edition as the current code. Its scope and reporting principles can inform professional practice, but MRS membership does not replace statutory duties or qualified advice.

Set a decision gate

End the round with one of four outcomes: proceed to a bounded test, revise the proposition, collect a named missing evidence type, or stop. A proceed decision should specify the user, problem, hand-off, data boundary, unresolved risks and next measurement.

Stop if recruitment excludes people most affected. Stop if participation and marketing are bundled, notes cannot be traced to valid information, a survey route is inaccessible, or a claimed segment rests on two anecdotes.

Recheck ICO guidance, current MRS code, tool records and accessibility evidence before publication or live fieldwork. Named research, privacy, PECR and accessibility reviewers should sign the method, not just the article copy.

Before you act

Common questions

What four fields should I record before recruiting participants?

Record the person or business expected to make the decision, the event being studied, the evidence that would support, change or stop the proposition, and the period and English geography the decision applies to. These fields keep the research tied to a decision that can fail.

When does market research become direct marketing?

ICO guidance says genuine market research is not direct marketing. But a survey containing promotional material, or one collecting details for later marketing, is treated as direct marketing. Calling an acquisition form research does not change its purpose, so keep recruitment and promotion apart.

How should I treat interview findings in the analysis?

Interview notes should separate what the participant said, what the researcher observed and what the team inferred. A single vivid comment can identify a risk, but it cannot establish frequency. Do not turn an interviewee into a named persona unless that person agreed to that use.