Planning email software audience research around four distinct populations

Good email marketing software audience research starts by separating the people involved. The business may buy the service, a decision-maker may approve it, a campaign operator may use it, and recipients experience the messages. One convenient sample cannot represent all four populations.

This is a desk-based method guide with no participant accounts or survey results. It sets out a process for research with organisations operating in England, using UK-wide law and official methods where appropriate. It makes no claim about audience size, buyer preference, demand or product adoption.

The evidence review closed on 5 September 2026 and must be repeated before publication.

What to take away

Separate business, decision-maker, operator and recipient populations before choosing any research method.
Define the target population, sampling frame and achieved sample, and report all three.
Keep research participation records separate from any promotional marketing permission.
Use interviews to observe real tasks and artefacts, not to measure prevalence.
Surveys need a defined population, usable frame and tested questionnaire before use.
Use the ONS business activity dataset to set geography and industry, not to identify software users.
Recruit through more than one route when a single channel would omit part of the target population.
Test eligibility rules with a second researcher on ten records before starting recruitment.

Name the four populations

Write a separate definition for each population before choosing a method.

Population	Unit to recruit or observe	Question it can answer	Common category error
Business	legal enterprise, operational unit or buying account	whether the organisation has a relevant workflow and authority to change it	treating every registered business as a prospect
Decision-maker	person who approves budget, risk or contract	how a purchase is evaluated and governed	assuming the most senior title owns the decision
Campaign operator	person who prepares, checks, sends or reports email	where work, defects and handover points occur	asking only managers about day-to-day operation
Recipient	person or subscriber receiving a message	how the message, preference route or linked journey is experienced	importing an existing contact list into research or promotion without a separate assessment

These populations can overlap in a small organisation, but the roles still answer different questions. Record each participant's actual responsibility rather than assigning it from a job title.

Before recruiting, convert each population into an eligibility rule, an evidence source and an exclusion. This worksheet keeps operator comments from being read as purchase demand and recipient feedback from being read as permission to market.

Population	Eligibility rule	Evidence to inspect	Exclusion
Business	England operating context plus a named recurring email process	organisation record and current workflow	address with no relevant activity
Decision-maker	approved or rejected a related purchase in a stated period	evaluation or approval record	senior title without decision involvement
Operator	personally prepares, checks, sends or reports campaigns	walkthrough of a recent task	manager describing another person's work
Recipient	fits the intended recipient definition	message and preference journey	contact included only because an address exists

Turn assumptions into research questions

A research question should change a decision. Write the decision this research must support, such as choosing one workflow for a prototype or deciding whether an England retail segment has enough recurring operational work for further discovery.

Neither question requires an invented national audience size. "Do marketers like automation?" is too broad. "How does an operator in a ten-person England-based retailer approve a suppression change across its current systems?" identifies the actor, setting, process and evidence to inspect.

The GOV.UK Service Manual advises teams to agree what they need to learn, which assumptions they need to test and what decision follows. Its research-planning guidance also says participant numbers depend on the method. The figures given there for government service research are practice guidance, not a statistical rule for the email-software market.

Create a question register with an owner, population, method, decision date and stop condition. Remove questions that ask respondents to predict distant purchasing or that can be answered from an existing contract, workflow log or public record.

Build the sampling frame before recruiting

The target population is everyone to whom the conclusion is meant to apply. The sampling frame is the list or route from which people can actually be selected. The achieved sample consists of eligible people who take part. Report all three.

ONS explains that sample design and estimation covers the frame, selection method, sample size, weighting and uncertainty. Its account is UK official-statistics methodology. The method asks teams to report uncertainty rather than present one figure as a settled finding.

For business research, start with an explicit unit: the ONS 2025 business dataset counts VAT or PAYE registered enterprises and local units from a 14 March 2025 snapshot.

The quality notes explain that an enterprise and a local unit are different. Low-turnover non-employing businesses outside those registrations are omitted. This can support a frame description, not a claim that the listed organisations use email software.

A workable frame records geography, sector, size rule, operational evidence and source date. If it comes from existing customers, an event list or a professional network, say so.

For England-specific coverage, combine routes such as a Companies House-derived company list filtered by SIC code, a relevant trade body membership list and a sector event in England. Record which route reached each participant, because each route has its own coverage limits.

Common routes differ in coverage. A customer list reaches existing relationships but excludes non-customers. A trade body reaches members rather than the whole sector. A public link produces a self-selected sample. A recruitment agency is limited by its panel and screener.

Recruit for eligibility and variation

The GOV.UK guide to finding research participants recommends defining recruitment criteria, including actual or likely users and considering disabled people and people with limited digital skills. Apply that principle to each population. Use more than one route when a single channel would omit part of the target population.

For a decision-maker, screen for a purchase or renewal handled within a stated recent period and ask what part the person played. For an operator, require direct involvement in a named campaign task. For a recipient study, recruit people who fit the intended recipient definition without drawing addresses from a marketing database by default.

Keep an exclusion log. It should show duplicate accounts, agencies speaking for clients, respondents outside the geography, people without the claimed responsibility and anyone recruited by a route that conflicts with the study design.

Keep a recruitment log with frame source, invitation date, route, invited count, screened count, eligibility outcome, completion count and reason for exclusion. Report invited, screened, eligible, completed and analysed counts separately. A completion rate needs its numerator and denominator. Test the eligibility rules with a second researcher on ten records before starting recruitment.

Worked example: one completed round

This illustrative example uses practice guidance, not a real study or a statistical rule.

A product team needed to choose one approval workflow for a prototype. The population was campaign operators in England retail businesses with a recurring email task. The frame combined a Companies House-derived list of retail companies with a sector trade body list and an England trade event.

The team screened 40 contacts, found 18 eligible, completed 12 interviews and logged 22 exclusions. It then selected the workflow with the fewest approval steps that operators could evidence from a recent campaign.

Indicative practice ranges for method planning are 6 to 12 interviews for a single population and 5 to 8 participants for a round of moderated usability testing. Treat these as typical practice guidance, not sample-size formulas. A survey still needs its own sample-size calculation.

Keep participation separate from promotion

Genuine research and later marketing are different purposes. The ICO's guide to identifying direct marketing says genuine market research is not direct marketing. A message becomes direct marketing if it contains promotion or the exercise is conducted to enable later marketing to those people.

Participant agreement also needs careful language. GOV.UK's informed-consent guidance covers what participants should understand about the session and collected material. The ICO separately states that participation consent should not be confused with the UK GDPR lawful basis for processing personal information. A qualified data-protection reviewer must select and document the appropriate basis for the actual study.

Permission to attend an interview does not grant permission to add the person to a promotional mailing list. Nor does it settle whether an imported address may receive marketing email. The ICO's current electronic-mail guidance requires a separate assessment of the recipient, acquisition route, purpose and applicable PECR conditions.

Use different records for research invitation, participation, recording, incentive, retention, follow-up research and marketing. Do not hide a sales call inside an interview.

Use interviews to observe decisions and work

Interviews can reveal how a process operates, but they do not measure prevalence without a defensible sampling design. The Service Manual's in-depth interview method recommends speaking with current or likely users, planning accessible sessions and arranging recording and note-taking. For this category, ask for a walkthrough of a recent real task with confidential material removed.

An operator might reconstruct how a campaign request arrived, where recipient eligibility was checked, who approved claims and how an objection reached the suppression record. A decision-maker might show an evaluation sheet, approval chain or renewal timetable. Concrete artefacts are usually more dependable than a hypothetical statement about what someone would buy.

Record the interview date, length, location or remote mode, participant role, recruitment route, eligibility result, questions asked and people present. Separate direct observation from interpretation. Do not publish quotations unless the participant agreed to that use and the wording was checked against the recording.

Use surveys only for questions they can answer

A survey needs a defined population, usable frame and tested questionnaire. ONS guidance for publishing survey methods calls for the collection mode, sampling units, sample size, response rates, coverage and the relationship between frame and target population. Commercial researchers should preserve the same audit trail even when they cannot claim official-statistics quality.

Before launch, conduct cognitive interviews on the questions, test keyboard and screen-reader routes, confirm mobile behaviour and rehearse the export. Avoid leading labels such as "advanced" or "best". Ask about a bounded event and period. A question about the last campaign approval is easier to interpret than a rating of general efficiency.

After fieldwork, publish the number invited, number eligible, number completing each item and the denominator for every percentage. Describe recruitment, incentives, exclusions, geography and dates. If respondents volunteered through a public link, the result describes those respondents. The report should name the route and the dates so readers can judge what the result covers.

Build personas as traceable evidence summaries

Personas are useful only when readers can trace each element to research. GOV.UK says that user profiles or personas should capture what a team has learned and that needs should be based on research rather than assumptions.

Replace fictional biographies with role cards, which can state work context, decision authority, recurring task, evidence observed and one constraint. They can also record group variation, supporting studies and unanswered questions.

Suppress needless personal details, and if evidence comes from two interviews in one sector, say exactly that. Never give the card an invented name, photograph, quotation or market prevalence.

Review contradictory cases before merging participants. A lone operator in a microbusiness and a campaign specialist in a regulated enterprise may touch the same software but face different approval, access and retention constraints.

Research competitors without manufacturing market share

A competitor sheet should preserve the entity, page title, URL, capture date, geography and evidence type for each claim. Companies House explains its public records in the register search guide, but its disclaimer says filed information is not verified or validated by the registrar. A legal entity record does not prove current trading activity, product capability or market position.

Use current first-party product documentation for features, terms and packaging. Treat customer reviews as a separate evidence class. The CMA's online consumer reviews case page records current enforcement activity and guidance around fake and misleading reviews. A review count or average is not a customer total, a representative satisfaction measure or market share.

Likewise, a visible search result reflects a query, location, date and personalisation context. Save the method and page examined. Do not turn ranking visibility into revenue or adoption.

Make a decision log, not a findings theatre

At the end of a round, give each conclusion an evidence ID, population, geography, fieldwork dates, numerator and denominator where relevant, limitations, owner and review date. Mark it observed, reported, inferred or unresolved. Link decisions to disconfirming evidence as well as supportive examples.

First, write one decision question for each of the four populations, then choose the smallest method to answer it. Stop research if the sampling frame cannot reach the intended group, participation is tied to promotion, necessary privacy controls are unresolved, or the team cannot state what decision the evidence will change.

Publication remains on hold pending named statistical, user-research, privacy and electronic-marketing reviews. Named accessibility and commercial reviews are also pending.

Before you act

Define each of the four populations separately.
Write research questions that change a decision.
Describe the sampling frame and its source date.
Screen participants for actual responsibility, not job title.
Log exclusions and report counts separately.
Get data-protection review before processing participant data.

Common questions

How many people should I recruit for this research?

The article gives indicative practice ranges for some methods, labelled as practice guidance rather than statistical rules. It says participant numbers depend on the method, citing GOV.UK research-planning guidance. Those figures are practice guidance for government service research, not a statistical rule for the email-software market. A survey still needs its own sample-size calculation.

Can I add interview participants to my marketing list afterwards?

No. The article states permission to attend an interview does not grant permission to add someone to a promotional mailing list. Research and marketing are different purposes, and the ICO requires a separate assessment of recipient, acquisition route, purpose and applicable PECR conditions before marketing email.

Why can't I just use my existing customer list as the sample?

The article says frames drawn from existing customers, event lists or professional networks systematically miss other people. Recruitment through a supplier's own newsletter likely favours current followers and should never be described as the England market. Such a frame can support a description, not a market claim.