Platform audience research that keeps seven buyer populations separate

What to take away

Write the buyer decision before recruiting

State the decision in one sentence. For example: should a named England-established merchant retain, repair or replace the route used for a defined adult-consumer catalogue during a specified contract window? Record the decision owner, options, deadline, systems in scope and evidence threshold.

Platform options may include Shopify, WooCommerce, BigCommerce, Wix and Squarespace. Enterprise options include Adobe Commerce, Salesforce Commerce Cloud and Shopware.

This prevents research about one population being used to answer a question about another. GOV.UK's online-selling guidance describes information and order steps for online sellers. It can help define the journey to study, but it does not identify platform buyers or decide which system fits them.

Build a population register

Use a population register before writing recruitment criteria.

A completed register row can look like this:

Do not infer protected characteristics, accessibility requirements, sector, technical skill or authority from a name, profile, postcode, device or transaction. Ask only eligibility questions that are necessary for the research decision. The Government Service Manual says participant criteria should reflect actual or likely users and warns teams to manage recruitment data carefully in its participant-recruitment guidance.

Compare platform options by evidence population

Name the platforms under consideration before recruitment. The comparison below uses public plan pages and typical ranges as of 2026. Check current quotes, tax, payment fees and contract terms for your England store.

Use the table to decide which populations must supply evidence. A Shopify or BigCommerce decision needs merchant, staff and consumer evidence. An Adobe Commerce or Salesforce Commerce Cloud decision usually needs procurement, IT and B2B buyer evidence. A marketplace seller study should not stand in for direct-store needs.

Define the target population and sampling frame

Write an inclusion rule using observable evidence. A merchant-buyer frame could require a verified England establishment, one stated selling model, an active decision event, responsibility for the relevant operation and permission to participate. Exclude suppliers being assessed, speculative leads, people without decision knowledge and records whose geography cannot be verified.

Choose a sampling frame that can reach eligible people: an approved customer register, contracted merchant panel, staff directory, or openly advertised invitation with screening. For an England merchant frame, a named route such as the Greater Manchester Chamber of Commerce member base can reach eligible businesses, subject to consent and permission.

The ONS UK business activity dataset has regional business tables. Its VAT and PAYE population does not reveal online selling, platform use, or buying authority. Do not use it as a participant list or infer missing characteristics from its classifications.

Record invited, eligible, consented and completed cases separately. Record withdrawn and unusable cases separately. Report who the frame misses, how non-response may change the result and whether recruitment incentives or supplier relationships could influence participation.

Separate participation from data-law decisions

Give each participant an accessible information sheet before the session. It should identify the controller, purpose, activity, observers and recording. It should also cover recipients, retention, withdrawal route, complaints contact and intended reporting. The GOV.UK guide to informed consent for user research says participation must be voluntary and explains how to record consent and handle withdrawal.

Ethical agreement to participate is not automatically the UK GDPR lawful basis for processing personal data. The ICO's research principles and grounds makes that distinction and requires an appropriate Article 6 basis, plus an Article 9 condition if special-category data is processed.

In England, the Data Protection Act 2018 and the UK GDPR set the processing rules, and the Privacy and Electronic Communications Regulations 2003 govern electronic marketing.

Qualified privacy counsel must assess the real study. Avoid collecting such data unless it is necessary and approved.

Research contact must also stay separate from sales contact. The ICO's electronic and telephone marketing guidance says genuine market research is not direct marketing, but a survey containing promotion or collecting details for later marketing is treated differently. Never move a participant address into a sales list without a separately documented route.

Build an instrument around recent events

Use an interview, task observation or survey according to the question. Start with eligibility and participant control. Ask for a recent, bounded event: what task occurred, what the person did, what evidence exists and what remained unresolved. Ask about desired features after current work and constraints are understood, so a supplier vocabulary does not frame every answer.

Keep question wording neutral. Do not ask whether a participant agrees that a platform is difficult or needs replacement. Randomise or rotate comparable options where order may matter, and preserve the instrument version. Test it with synthetic data and people who are not in the analysis sample.

If a session is recorded, obtain specific permission, name every observer and define access. GOV.UK's recording and note-taking guidance advises teams to use recordings only within the agreed purpose and remove personal or confidential details from outputs.

Analyse observations without manufacturing certainty

Maintain three fields: what was observed or said, the researcher's interpretation and the proposed decision. GOV.UK's session-analysis guidance recommends capturing observations before interpreting them and involving multiple observers to reduce individual bias.

For every finding, state the eligible population, sample source, field dates and instrument version. Add completed and missing cases, contradictory evidence and confidence. Do not turn a small purposive interview set into a percentage. Do not weight a convenience survey unless a statistician has a defensible frame, selection probability and adjustment method.

Verify any quotation against the authorised recording or notes and the participant's reporting permission. Paraphrase or omit it if identification risk cannot be controlled. GOV.UK's participant-data guidance recommends anonymising extracts where possible and restricting material that can identify someone.

Use personas as evidence indexes

A persona should summarise a bounded evidence pattern, not imitate a real individual. Give it a functional label such as "marketplace seller handling refund exceptions".

Attach the sample, field dates, observed task and frequency within that sample. Add counterexamples and unanswered questions. Do not add a fictional name, portrait, quotation or age. Do not add a disability, turnover band or skill level.

Retire the persona when its source evidence expires, the operating model changes or later research contradicts it. A persona that cannot be traced back to authorised evidence belongs in an assumptions log, not a buyer recommendation.

Choose collection tools after the protocol

Tool selection follows the data map. Compare the participant access route, question types, identity settings, consent presentation and accessibility evidence. Compare storage, processor terms, security, export and deletion. Keep a no-new-tool route and account exit in the comparison.

For example, Jisc publishes separate current records for Online Surveys, including its terms effective 23 July 2026, accessibility statement dated 6 June 2026 and response-export formats. Those first-party statements need contract and hands-on verification for the buyer's exact account. They do not prove that the service is suitable or compliant.

Keep competitor research outside buyer evidence

A competitor's public product page can establish what it claims on a date. A Companies House filing can help verify a legal identity, subject to the register's limits. The Companies House search guidance says the register is not comprehensive and may contain inaccuracies. Neither source reveals customer demand, platform performance or market share.

Use authorised synthetic journeys to compare observable seller information, catalogue steps, checkout wording and support routes. Do not copy databases, personal data, photographs or customer reviews into the research set without a reviewed basis. Public material remains subject to privacy, rights and site-term questions.

Set decision and stop rules before fieldwork

A research finding should change a named requirement, test, shortlist or hold decision. Define the minimum evidence for that change and the reviewer who accepts it. For a qualitative merchant study, an example threshold is 12 completed interviews per operating model, with recruitment stopping when three consecutive sessions add no new decision-relevant theme.

Stop recruitment if eligibility cannot be checked without disproportionate personal data, participants cannot understand the information, pressure or conflict undermines voluntary participation, or the team cannot secure recordings and notes.

Hold the platform decision when the sample excludes a necessary operating model, missingness is unexplained, supplier participants dominate, or a finding depends on an unverified quote. The final report should preserve dissent and unknowns. Research reduces uncertainty for a bounded buyer decision; it does not certify a platform or represent every merchant, worker or consumer in England.

Before you act

Common questions

What should the buyer decision statement record?

State the decision in one sentence, then record the decision owner, options, deadline, systems in scope and evidence threshold. This prevents research about one population being used to answer a question about another. A named researcher, statistician and qualified privacy, consumer and accessibility reviewers must approve the relevant parts before fieldwork or publication.

Why is the ONS business activity dataset not a participant list?

Its VAT and PAYE population does not reveal online selling, platform use or buying authority. It has regional business tables. Do not use it as a participant list or infer missing characteristics from its classifications.

Instead choose a sampling frame that can actually reach eligible people, such as an approved customer register or contracted merchant panel.

How should findings and quotations be reported?

For every finding, state the eligible population, sample source, field dates and instrument version. Add completed and missing cases, contradictory evidence and confidence. Do not turn a small purposive interview set into a percentage.

Verify any quotation against the authorised recording or notes and the participant's reporting permission. Paraphrase or omit it if identification risk cannot be controlled.