Audience research for landing page software starts with a decision, not a persona

Landing page software audience research should answer a named product decision, not manufacture a portrait of an "England audience". Write down which organisations could buy, which people take part in the decision, which people operate the software and which visitors encounter the published pages. Then choose participants and methods that can test a specific uncertainty.

No interviews, surveys or usability sessions were conducted for this guide. It is a desk-researched planning method, checked on 6 September 2026. Government user-research material is written for public services, so it supports research practice here rather than proving commercial demand. Data-protection questions require a qualified UK reviewer before any recruitment or processing begins.

What to take away

Start with the decision, not a persona

A useful plan begins with a decision such as whether to build for a marketing team that cannot safely release pages through its current workflow. "Understand small businesses" is too broad. It neither identifies a purchasing event nor says what evidence could change the proposed product.

The GOV.UK Service Manual recommends turning unfounded assumptions into research questions when teams plan user research. That advice translates into a short decision record:

For example, do not ask whether people like landing page builders. Ask whether eligible page operators can publish a reviewed campaign page under their organisation's approval, accessibility and data-handling constraints. The evidence could be a recent workflow account and a controlled task using a non-production prototype. A preference alone cannot answer it.

Keep six populations apart

Audience work becomes unreliable when different populations are merged. Use separate registers for:

One person may occupy more than one role, but the analyst must not assume that. A visitor is not necessarily a lead; a form submission does not prove buying authority; a software user does not automatically represent the buyer organisation.

Official business data can help describe a sampling frame, but not landing-page-software demand. The ONS UK business activity, size and location dataset covers registered enterprises and local units with regional tables.

The Department for Business and Trade's 2025 business population estimates use a broader private-sector population including estimates for unregistered businesses.

Neither source identifies landing-page-software buyers. State which population is being used and why.

Build an observable sampling frame

Write eligibility rules before contacting anyone. An organisation might qualify because it operates in England, commissioned or changed a campaign page within a declared recent period, and has a person who directly performed or approved the work. Record how England is established, perhaps through the relevant operating site or responsibility, rather than a participant's accent, name or postcode alone.

A sampling frame is the actual list or route for reaching participants. It might be current customer records, a professional association invitation, a recruitment supplier or an opt-in research panel.

Each route excludes somebody. Customer records omit non-buyers; professional networks favour members; volunteer panels overrepresent those willing to join research. The Service Manual's guide to finding participants notes that timing, venue, activity and recruitment route can all introduce bias.

Create a recruitment log with invite count, eligibility failures, acceptances, withdrawals, no-shows and completed sessions. Do not disguise a convenience sample as representative. If the study cannot reach a necessary group, record the gap and narrow the decision.

Recruit without turning research into promotion

An invitation should identify the researcher, purpose, requested activity, time commitment, incentive, selection process and contact route. Research participation is separate from permission to send future marketing. Do not add participants to a lead-nurture list because they answered a study.

Before collecting personal information, a controller needs a documented purpose and an appropriate lawful basis. The ICO's lawful-basis guide says this choice depends on the particular purpose and relationship, and must be made before processing.

This article does not select one for a reader. ICO guidance reflects Data (Use and Access) Act changes dated 2 April 2026, while other ICO pages may still be marked under review. The reviewer must recheck the current position.

The Service Manual's informed-consent guidance says participants should understand who is conducting the research, what will happen, what data is collected, recording, sharing, retention, voluntary participation and withdrawal. Its public-service wording is a practical consent model, not a substitute for a case-specific legal assessment.

Offer information and consent materials in a usable format. Ask about access needs directly and neutrally. Do not infer disability, language, ethnicity, health, religion or other characteristics from names, browsing behaviour or social profiles.

Match the instrument to the uncertainty

Choose the lightest method that can answer the question:

Write the instrument before recruiting and keep a version number. In interviews, start with a recent concrete event, then cover chronology, artefacts, exceptions and consequences, and avoid pitching features or asking participants to predict purchases.

For surveys, include eligibility checks, one idea per question, a suitable response option for uncertainty and completion timestamp, and pilot for comprehension, not favourable answers.

Analyse what the evidence can bear

Keep four layers separate: the participant's words, the researcher's observation, an interpretation and the product decision. The Service Manual's guide to analysing a research session separates observations, findings and resulting actions. A recurring theme in a purposive sample is still not a population estimate.

An evidence table should carry the participant code, eligibility evidence, event date, source artefact, observation, contradictory case and analyst. Count only when the numerator and eligible denominator are defined. Do not report percentages from a small qualitative set as though they describe England. Preserve dissenting cases, non-response and missing data.

Personas can be created only after analysis. Each persona should cite the sample, field dates, behaviour pattern, contrary evidence and expiry date. A polished fictional biography can be memorable, but it is not a finding.

Control data from collection to deletion

Research notes, recordings, screen captures and recruitment records may contain personal information. GOV.UK's guide to managing research data and participant privacy advises collecting the minimum required, restricting access, agreeing retention and keeping consent evidence linked to the relevant data.

Maintain a data map covering controller, processors, collection point, fields, access roles, storage location, transfer route, retention trigger and deletion evidence. Keep contact details apart from analysis where practical. Use participant codes in working outputs and check whether unusual combinations could still identify someone. A transcription or survey supplier does not remove the buyer's responsibility to examine roles and contracts.

Use uncertainty and stop rules honestly

Set limits before seeing results. Stop if consent records cannot be matched to data, a recruitment route systematically misses an essential group, recording fails, the instrument changes mid-round without versioning, or the team cannot act on the evidence. Pause product conclusions if participants describe a different problem from the one being tested.

At the decision meeting, state what the study supports, what it contradicts and what remains unknown. The owner can proceed, commission another round, narrow the intended segment or stop. None of those outcomes turns research participants into customers or validates an England-wide audience.

A blank plan for the next round

Complete these fields before an invitation leaves the organisation:

A named user researcher should approve the method, while qualified data-protection, accessibility, security and commercial reviewers clear their own areas. Publication review must reopen every current source and confirm that the page reports a plan, not research that never happened.

Before you act

Common questions

What should a landing page software research plan start with?

It should start with a named product decision, such as whether to build for a marketing team that cannot safely release pages through its current workflow. Broad aims like understanding small businesses identify no purchasing event and no evidence that could change the proposed product.

Why should buyer organisations and research participants be kept apart?

Merging populations makes audience work unreliable. A research participant is a sample, not the whole market. A visitor is not necessarily a lead, a form submission does not prove buying authority, and a software user does not automatically represent the buyer organisation.

When can personas be created for this kind of research?

Only after analysis. Each persona should cite the sample, field dates, behaviour pattern, contrary evidence and an expiry date. A polished fictional biography cannot substitute for evidence, and a recurring theme in a purposive sample is still not a population estimate.