How to research customer data platform buyers in England without manufacturing a persona

Audience research for a customer data platform should begin with a decision and several distinct populations. The organisation buying a service, the person approving it, the operator configuring it and the customer whose data may enter it are not one audience. Blending them produces unreliable findings and can hide conflicts between commercial goals and people's interests.

The research plan must also define what customer data platform means for this study. It might refer to identity reconciliation, profile governance, preference propagation or audience activation. CRM, warehouses, consent tools, tag managers, analytics and clean rooms should remain outside the frame unless the declared function requires them. A product label is not evidence of function or lawful use.

The evidence cut-off for this desk guide is 6 September 2026. The buyer geography is England; cited privacy and electronic-marketing guidance applies across the UK. No interviews, surveys or product tests were conducted. The work remains on publication hold until the named qualified reviewers assess the proposed research and data handling.

What to take away

Start with the decision record

Write the decision before choosing a method. For example: should an England organisation commission a controlled test of a system that propagates a correction and suppression instruction across three authorised data stores? That wording identifies a buyer, a function, an observable task and a bounded next step. It does not assume demand or a preferred supplier.

The record should name the accountable owner, options under consideration, current manual route, evidence needed, decision date and conditions that would stop the study. Add the uncertainty that research cannot remove. A small interview study may explain how teams describe recent work, but it cannot establish prevalence across England.

Separate the people involved

At least nine populations may matter:

These are research roles, not demographic profiles. Do not infer authority, residence, disability, consent, budget or software use from a name, job title, public profile or postcode. Ask and verify only what the eligibility rule requires.

The ONS UK business activity, size and location methodology can inform a frame of registered enterprises or local units. It cannot identify CDP buyers or the right respondent inside an organisation. The geography rule should state whether eligibility follows the contracting address, operational establishment or another declared location, and it must stay fixed.

Design recruitment without disguising promotion

Record the source and authority for every contact. A business email address on a website proves neither buying authority nor agreement to research, and shows no wish for later marketing. The ICO's direct marketing guidance was updated on 28 April 2026 to reflect Data (Use and Access) Act changes.

Researchers must classify the actual communication. A request for research participation and a promotional follow-up need separate treatment.

Recruitment materials should state who conducts and who commissioned the work, the topic, why selected, what participation involves.

They must also cover intended data uses, recording or transcription, incentives and retention.

Name recipients and withdrawal route; the ICO page on privacy information is under review because of the Act, so the final notice needs publication-day checking and qualified privacy advice.

The MRS Code of Conduct is a professional standard for its members and Company Partners. It is not legislation or evidence that a supplier followed the code. Its participant-centred duties are a useful additional control where the project falls within market, opinion or social research.

Build a sampling and fieldwork log

Define the sampling frame, selection rule, contact channel, screening questions and replacement policy before recruitment. Keep a disposition for each selected unit: ineligible, unreachable, declined, scheduled, completed, withdrawn or unusable. Report the eligible and responding denominators separately. Never replace a difficult respondent silently with a convenient one.

Screen for recent observable work. An operations participant might need to have resolved a duplicate identity within a declared period. A procurement participant might need to have reviewed a data-service contract. This is stronger than asking whether someone is interested in customer data.

Accessibility and safeguarding need planning before contact. The Government Social Research profession's inclusive social research guide discusses barriers, recruitment and adjustments in a government-research context. It does not prescribe the answer for every commercial study. Offer usable formats and channels, minimise disclosure, and let the participant choose whether to discuss an access need.

Write neutral instruments and test them

An interview guide should move from a recent event to evidence and then interpretation. Ask what happened, which system or record was used, who decided, what was unavailable and what followed. Do not begin with a proposed CDP feature or ask whether a single view would help.

For surveys, define every term and response option. The Government Analysis Function's questionnaire design guidance is marked under review. It warns against leading questions and covers respondent burden, accessible design and testing.

Use cognitive or pilot testing with synthetic or specifically authorised material. A pilot checks comprehension and routing, not market findings, unless its frame and method support them.

Keep question version, presentation order, channel, interviewer, date and any prompt visible in the research record. If a recording or transcript is proposed, document its purpose, access, correction route, retention and deletion. Do not upload participant material to an undeclared transcription or analysis service.

Analyse evidence without manufacturing a persona

Separate five layers:

A proto-persona is an evidence card for a role and task, not a fictional biography. It should cite the sample, field dates and source records, show contradictory observations and expire when the underlying process changes. Protected traits belong only where they are necessary, ethically collected and analysed under qualified review. They should not be guessed from behaviour or names.

Quantitative reporting needs the eligible population, sampling and responding denominators, question version, missing answers, weighting if any and uncertainty. A percentage from a convenience survey does not become an England buyer estimate. Qualitative themes should retain the number and type of contributing cases without implying statistical prevalence.

Control research data and hand-off

The research map should list personal data, sources, purposes, recipients, storage, access and deletion. The ICO's data-protection principles guidance provides the UK framework for minimisation, accuracy and storage limitation. Its controllers and processors guidance makes clear that roles depend on actual purposes and means, not a supplier label.

If an external survey, recording or transcription service is considered, identify the legal entity, exact service, data role, access path, subprocessors, transfers, security, accessibility, export and deletion. The NCSC's supplier assurance questions can structure evidence requests about incidents, data return and termination. It does not approve a service.

Hand-off only the evidence needed for the stated decision. A research contact list must not become a marketing audience by default. Remove direct identifiers where they are unnecessary, preserve limits beside every finding, and name who may reopen the data.

Decide, stop or research again

Before fieldwork, set stop rules for unsafe recruitment, ambiguous authority, inaccessible participation, undisclosed recording, unreliable screening, material non-response or an instrument that pushes respondents towards the preferred answer. Pause when the project begins collecting evidence unrelated to its decision.

The final decision log should state what the research supports, what it contradicts and what it cannot answer. A finding may justify a synthetic workflow test, further sampling or no procurement. It cannot by itself establish that a platform is lawful, secure, accessible or commercially suitable.

Recheck ICO guidance marked under review, the MRS code version and every supplier record immediately before fieldwork or publication. If the purpose, participant population, recruitment route, processing or hand-off changes, reopen the relevant approval instead of relying on the original plan.

Before you act

Common questions

Why should audience research start with a decision record?

A decision record names the accountable owner, options, current manual route, evidence needed and decision date. It identifies a buyer, function, observable task and bounded next step. It does not assume demand or a preferred supplier, and it adds uncertainty that research cannot remove.

How should researchers handle recruitment without disguising promotion?

Record the source and authority for every contact. A business email address does not prove buying authority or consent to research. The ICO direct marketing guidance updated on 28 April 2026 means a research request and a promotional follow-up need separate treatment under PECR.

What makes a proto-persona different from a fictional biography?

A proto-persona is an evidence card for a role and task. It should cite the sample, field dates and source records, show contradictory observations and expire when the underlying process changes. Protected traits belong only where necessary, ethically collected and analysed under qualified review.