Researching marketplace sellers around one decision without turning research into sales

Marketplace seller software audience research should answer one buyer decision: can an England merchant improve the way one authorised product record, stock position and later order states move between its source system and a named marketplace account? It should not begin by collecting opinions about generic ecommerce tools.

Define the marketplace, account owner, catalogue in scope and current process before recruiting anyone. Then recruit people because they performed an observable part of that process or will make the purchase decision. A public profile, postcode, company filing or job title cannot prove that authority.

Name the decision and the boundaries

A useful research statement names the current failure, choice and deadline for evidence. For example: decide whether to retain a controlled manual reconciliation or investigate a new operating route for listing changes, stock conflicts and refunds. Do not promise automation, savings or fewer errors.

Map the actors before describing an audience. The legal seller contracts with the customer. The marketplace operator supplies the venue and may carry out other activities under its terms. The software buyer controls procurement. Catalogue staff own approved product content; operations staff deal with stock and orders; customer-service staff handle queries and later cases. Product-safety, privacy, security, accessibility, finance and VAT specialists each hold a separate review gate.

Consumers and customers are another population. They may experience seller information, price presentation, delivery, cancellation, returns and support, but they do not automatically know how the merchant's integration works. A managed provider or software supplier can explain its documented service, not speak for seller staff or customers.

GOV.UK guidance for online sellers provides a current list of seller and ordering information. Use that record to define journey tasks, not to conclude that a research participant understands consumer law. OPSS product-safety advice is a reminder that product role and category matter. Do not expose participants to a live item with unresolved evidence or ask them to judge safety outside their competence.

Turn the audience into a sampling frame

State the England rule precisely. The decision owner might work for an organisation with an operating establishment in England and authority over the specified marketplace account. Record the evidence used and note groups operating elsewhere. Residence is not a safe substitute for business responsibility.

HMRC's page on selling through digital platforms defines a seller for its reporting context. It also distinguishes individuals and entities. That can inform a screener, but it does not establish software use, product category, trading status, turnover or budget.

ONS business activity and location data classifies enterprises and local units from the IDBR. It cannot supply a marketplace-software audience. Use an official register fact only for the fact recorded. Build the actual frame from authorised merchant records, documented roles and recent workflow evidence.

For each population, define inclusion and exclusion before contact. A catalogue participant might need to have approved or corrected an in-scope listing within a stated recent period. An order-operations participant might need to have investigated a stock mismatch or refund. A buyer must have formal input to the decision. Record every person screened, eligible, invited, booked, attending and completing each item. Unknown eligibility stays unknown.

Recruit without turning research into sales

Record where each contact detail came from, who may use it and why the invitation is appropriate. Research participation is not consent to marketing, permission to message through a marketplace or authority to enter an account. Keep sales follow-up and advertising lists outside the research file.

The current MRS Code of Conduct applies to MRS members and Company Partners. It defines professional duties around participant interests, transparency and reporting. It is not a substitute for legal review. The government user-research recruitment guide advises defining target groups, protecting privacy and avoiding recruitment bias. Its examples concern public services, so adopt only the general method.

Give invitees a plain participant sheet before they agree. It should name the controller, purpose, activities, observers, recording and transcription, intended uses, recipients, retention, incentive, withdrawal route and contact for questions or complaints. Make materials and sessions available in formats participants can use. Do not infer disability from a device or ask for a diagnosis when a practical participation adjustment is enough.

Separate ethical permission from data protection

The GOV.UK guide to informed consent for user research says participants should understand the purpose, collected data, recording, sharing, retention and withdrawal. This is ethical permission to take part.

The ICO explains that research participation and UK GDPR lawful basis are distinct questions. A privacy specialist must identify the controller, purpose and appropriate lawful basis for each processing operation. Do not label ordinary commercial discovery as protected research without qualified analysis.

The ICO's research-safeguards page is under review following DUAA changes. It still provides a useful current boundary around minimisation, anonymisation and pseudonymisation, but publication-day review is required. Pseudonymous data remains personal data.

Collect only what the study needs. A seller interview rarely requires a customer's full name, address, payment details or complete order history. Demonstrate the instrument and workflow with synthetic records. If a real example is essential, obtain specific authority, reduce the fields, control access and document deletion.

Build instruments around recent events

An interview guide should move from role confirmation to a recent, concrete episode. Ask the participant to describe the last approved listing change or order exception they handled. Establish what triggered it, which records they opened, decisions they made, hand-offs, delays, workarounds and how they knew the case had finished.

Avoid asking whether a proposed tool sounds useful. Government Analysis Function questionnaire guidance advises short, clear wording and warns against vague, assumed or leading questions. Test the screener, discussion guide and survey with synthetic cases before fieldwork. Keep the version used for each response.

If the session covers customer experience, test a safe prototype or authorised journey. Do not alter live prices, stock, payments, returns or tracking for the sake of research. Stop a session if a participant reveals personal data that the study did not request, encounters an unresolved safety issue or cannot understand the permission materials.

Control recordings and suppliers

Decide whether notes are enough. If audio, video or transcription is proposed, explain the exact purpose, access and deletion rule before seeking permission. Record continued permission when the scope changes. A quote must be checked against the source and approved for its intended level of identification; otherwise paraphrase or omit it.

Recruiters, research platforms and transcription services may process personal data. The ICO's page on controller-processor contract terms covers instructions, confidentiality, security, subprocessors, rights assistance, deletion or return, and audits. The page is marked under review after DUAA changes. Verify the current position and the exact supplier chain.

NCSC supplier-assurance questions can help examine access, cloud dependencies and data protection. Answers are claims until backed by evidence for the contracted service. Test account removal and export with synthetic material. A certificate cannot decide the merchant's risk acceptance.

Analyse without inventing prevalence

Keep five columns in the analysis ledger: observed action, participant statement, documentary evidence, researcher interpretation and unresolved question. Preserve contrary cases and missing items. A small qualitative sample can reveal how a process fails, but it cannot establish how common that failure is across England.

For a survey, publish the frame, field dates, question version, invitations, eligible responses and item denominators. Explain non-response, partial answers, duplicate handling and any weighting. Do not report a percentage without its numerator, denominator and population.

Personas, if used, should be evidence cards rather than fictional characters. Each card needs a source sample, date, observed job, variation, counterexample and expiry. Never add a name, portrait, age, disability, turnover or motive simply to make the card feel real.

Decide, stop or research again

Before using findings, ask whether the relevant decision population was reached, whether permission and data controls held, whether evidence is traceable and whether opposing cases were retained. Consumer, product-safety, privacy, security, accessibility, finance and VAT reviewers must approve only their own area.

Proceed to a bounded synthetic test only when the observed seller problem survives those checks. Repeat recruitment if a critical role or workflow state is absent. Stop if participants were misled, consent records cannot be matched, live customer data was exposed, supplier deletion failed or the sample is being presented as England-wide demand.

Publication-day review

Re-open MRS, ICO, GOV.UK, OPSS and NCSC pages before publication. Assign named research, privacy, accessibility, consumer, safety, security and accounting reviewers. Keep the article on hold until the research date, sample, limitations, corrections route and data-deletion evidence can be published honestly.