How to plan branding audience research that can still disagree with you, before you hire an agency

Decision test: The choice that research could change.

Target population: The full group the conclusions are meant to describe.

Sampling frame: The working list from which potential participants can be approached or selected.

Achieved sample: The people who participate after exclusions and withdrawals.

Decision role: A person's part in approving, commissioning or using the work.

Theme saturation: The point at which two consecutive interviews add no themes relevant to the decision.

Inference limit: The boundary on the people, places and period the evidence can represent.

Protect the route before collecting evidence

A sound research plan establishes how participant information will be handled before anyone is approached. Name the controller, any suppliers processing data, the lawful route, retention period, permitted access and deletion process. Participation in research and permission for later sales contact are separate matters, not one blanket agreement.

Limit screening fields to what recruitment needs; a job title and work email can still be personal data. Record the approach alongside data protection requirements and the applicable UK rules and authorities. Agreement to participate does not resolve every question about processing personal data or later uses.

GOV.UK informed-consent guidance says participants should understand who is conducting the work, its purpose, the activities and information collected, recording, sharing, retention, withdrawal and complaints. Supply information in a usable format and keep evidence of agreed participation. Consent to take part should not be treated as an answer to every data-processing issue.

The ICO data protection principles require lawful, fair and transparent processing, purpose limitation, minimisation, accuracy and storage limitation, alongside security and accountability. The regulator marks its guidance as under review following the Data (Use and Access) Act. A privacy reviewer should check current advice against the actual information flow.

Research participation does not itself permit later marketing. The ICO direct-marketing guidance was updated on 28 April 2026. Consider UK GDPR and PECR requirements for recruitment and subsequent marketing as distinct purposes; do not add a participant to a mailing list without a separate basis.

Set the inference boundary

Use target population, sampling frame and achieved sample as three separate descriptions. The first defines the people conclusions concern; the second shows who could be reached; the third records who took part. Differences between these groups narrow what can be claimed, however detailed the participating people's accounts may be.

For an England study, connect the geographic boundary to the commissioning decision, rather than assuming that agency registration decides the matter. A Manchester-based UK marketing director might commission a London agency for a regulated product sold across England, Scotland and Wales. A UK statistic does not become an England result merely because the research intends to describe England.

The ONS UK Business: activity, size and location release covers enterprises and local units registered for VAT or PAYE, with regional and industrial tables. It may help form business-study strata, but does not identify branding purchasers, decision controllers or unregistered businesses. ONS sample-design guidance explains frames, stratification, clustering, weighting and precision.

A convenience list of founders or previous clients can support discovery, not an England-wide percentage. State the intended group, frame coverage and selection rule beside the intended audience. A buyer-persona method should not conceal those boundaries or present them as population facts.

Name the choice that research could alter

The decision test should identify a real choice that evidence might change, rather than a general wish to understand an audience.

Set out the available options, evidence that could rule each out, the person responsible and the date for review.

Evidence might affect service scope, channel choice, delivery design or a later commercial forecast. A budget template can make the options explicit; the planning guide can help frame whether commissioning is appropriate. Do not ask people to endorse a price, campaign or proposition already settled, since that tests reaction rather than an open decision.

A qualitative round may take four to six weeks from brief to report. Recruitment route, incentives, transcription, analysis and approvals in regulated sectors affect cost, so agree a not-to-exceed amount before recruitment starts. GOV.UK guidance on planning user research begins with the problem, assumptions, user groups and what a team needs to learn.

That government material can inform commercial research, but its examples concern government services rather than buyers of branding agencies. Keep the distinction visible when using its planning principles. The decision test should arise from the commissioning choice; check that it fits instead of transferring a service-research example without adjustment.

Worked decision test: hiring a branding agency

For example, a company asks whether to commission a full rebrand from a branding agency, a refresh from a freelance designer, or no change. The decision owner is the marketing director and the review date is set before fieldwork. This is an illustration, not a finding.

Evidence that would rule out the full rebrand: customers can already tell the company apart from competitors and describe it consistently. Evidence that would rule out no change: buyers misread what the company does. Evidence that favours the freelancer: the problem is limited to inconsistent assets, not positioning.

Recruit recent buyers, people who chose a freelancer, and people who delayed or declined a rebrand. Then use the results to build the agency shortlist. The shortlisting method and the freelancer comparison show how to move from the research answer to named options.

Recruit for disagreement, not endorsement

Recruitment should make room for people whose experience may challenge the expected answer. Set eligibility around a recent event and each person's decision role before contact begins. Include people who selected an internal team, delayed work or rejected an agency, as well as satisfied former clients.

For each role, plan 8 to 12 completed qualitative interviews, then stop when two successive sessions add no decision-relevant themes. Limit participation to two people from one organisation unless a broader internal view is necessary. Fewer than six eligible completions is scoping evidence, not a basis for percentages.

These limits broaden the alternatives heard, but cannot turn a non-probability sample into a representative one. GOV.UK guidance on recruiting participants for user research recommends reaching actual or likely users, varied recruitment routes, access support and attention to bias. Log eligible invitations, completions, refusals, unreachable candidates, withdrawals and exclusions.

Record whether participants came through client referral, trade group, public notice or paid recruiter; these routes cover different groups. Decide the incentive form, amount, funder and payment condition before contact, and disclose agency, client, supplier and recruiter relationships. The 2023 MRS Code of Conduct is a professional standard for members and company partners, not legislation.

Distinguish an account from an estimate

An interview should move from context to a recent event, then to decisions, artefacts, alternatives and consequences. Begin openly, ask what happened next, and use a consistent topic order while allowing unexpected routes through the experience. The interview guide gives related approaches for making interview material evidential.

GOV.UK in-depth interview guidance recommends neutral, open questions and recent examples. Assign an interviewer and note taker in advance, and record whether an observer or recording is present. Avoid prompts that presume a difficulty or emotion, such as asking how frustrating inconsistent brand assets were.

GOV.UK advice on notes and recordings addresses consent, secure handling and removing unnecessary identifiers. Do not reconstruct a polished quotation from memory. If exact wording was not captured and cleared for the stated use, describe the finding as a paraphrase without quotation marks, or leave it out.

Qualitative work can inform a questionnaire, but does not replace a survey method. Before a form opens, specify the population, frame, selection method, contact mode, collection period, response calculation, cleaning rules, weighting, uncertainty and publication plan. An online instrument alone does not define how a survey was conducted.

The Government Analysis Function questionnaire design guidance separates questionnaire design from the wider survey method. It addresses burden, inclusive design, neutral wording, cognitive testing and usability testing. Its page is under review, so treat the principles as a dated design input rather than current certification.

Arrange questions around the decision: check eligibility, ask about a recent event before attitudes, and explain why sensitive information is needed. Randomisation, branching and mandatory responses affect who sees each question and which records go missing. Retain the tested questionnaire version and a map of its logic.

ONS notes that sampling and non-sampling error both matter. Unreachable people, refusals, inaccurate answers and processing mistakes may affect findings even when many responses are collected. Report response numerators and denominators, frame coverage and known data loss. An opt-in poll does not become a probability sample through response volume alone.

Treat tools and competitor records as evidence

Start with a manual baseline: an approved discussion guide or questionnaire, a controlled contact log, and a response store with evidenced access and deletion.

Consider a platform only when it improves a necessary part of the protocol without failing another requirement; assess the specific product for access, routing, export and accessibility.

Supplier descriptions do not prove that a configured study works. Test with synthetic records before using live data. The W3C forms tutorial covers labels, grouped controls, instructions and feedback; the NCSC small organisations guide offers a starting point for accounts, devices, backups and scam risks. Keep accessibility, equality, security, privacy and procurement as separate gates.

For competitor research, define the eligible service and evidence period before gathering agency names. Keep register facts, supplier statements, independent evidence and editorial interpretation distinct. A filing may identify a legal entity without verifying service quality; a case study remains the supplier's account unless the buyer and outcome are independently confirmed.

Use bounded competitor research to keep those evidence categories apart. Do not scrape a directory merely because it is public. Check access terms, licences and technical controls, and collect only fields needed for the documented purpose. Record inaccessible evidence as unknown, rather than treating it as a negative result.

IPO database-rights guidance explains that copyright and database rights can protect different parts of a database. Seek qualified intellectual-property and privacy review before automated or substantial extraction. Preserve sources and limits with resulting material, rather than presenting it as a complete account of competitors or their performance.

Release a claim only within its boundary

Release a decision with a bounded statement: identify the choice the evidence can support, contrary cases still present and what remains unproven.

An audience description is useful when its scope, period and uncertainty are clear enough to be questioned by later work, rather than made more certain than its sample permits.

Retain the decision owner and review date; England boundary and reference period; recruitment routes and inclusion rules; question order, interviewer and tested version; field dates, invitations and responses; incentive, funding and conflicts; privacy and deletion arrangements; evidence types; contrary cases; uncertainty; threshold; stop condition; and next evidence need.

Hold the choice if a material buyer role is absent, the geographic rule shifted during the study, participant information lacks an approved route, or a qualitative theme is presented as prevalence. Use the selection checklist as a gate, and proceed only within the population and period the evidence can support.