Audience research for productivity software, from recruitment frame to interviews without selling

Business productivity software audience research should reveal how work happens, who can authorise a change and what would make adoption fail. It is not a hunt for encouraging comments about a proposed feature. A useful study begins with a decision the team needs to make and recruits people who have direct experience of the relevant workflow.

This guide sets out a research process for products aimed at organisations in England. It draws on UK-wide business data and government user-research guidance where suitable, but it reports no interviews, survey responses or original market findings. The sources were reviewed on 5 September 2026.

What to take away

Write the decision before the discussion guide

Research becomes unfocused when every stakeholder adds a favourite question. Name the decision first. It might be whether to serve single-site employers, whether supervisors need a mobile approval route, or whether setup requires a paid migration service.

The GOV.UK Service Manual advises teams to turn unfounded assumptions into research questions and choose methods that answer those questions with reasonable effort. Although the manual is written for government services, the discipline transfers well to business software.

Use a simple planning sheet:

Avoid objectives such as understand small businesses. A better question asks how a defined group approves urgent work when the person with authority is away from a desk. It names a behaviour that can be observed and a design choice that could change.

Establish an England recruitment frame

Official statistics can describe the population from which participants may be drawn. They cannot select the audience on their own. For the start of 2025, the Department for Business and Trade put England's private-sector business population at 5.0 million. The release includes businesses without employees and assigns regional location by head office.

The Office for National Statistics provides VAT or PAYE registered enterprise and local-unit tables by industry, employment band and geography. That population excludes some very small and unregistered activity. Select the dataset that resembles the intended buyer base, then document what it leaves out.

A recruitment frame needs more than sector and staff count. Add criteria tied to the workflow: recent use, transaction frequency, decision authority, existing systems, number of sites, support needs and whether the person joined during or after the last software change. These details distinguish a likely user from a convenient contact.

Do not apply national percentages to a local list and call the result representative. An English research sample can be purposive when the aim is to understand contrasting situations. State how people were found and which relevant groups were missed.

Separate the people around a purchase

The person who complains about a process may not choose the software. Business tools can involve an operator, line manager, system administrator, budget holder, security reviewer, finance approver and people whose information enters the system. One person may hold several roles in a microbusiness. A larger employer may split them across departments.

Map each role against four questions: what do they do, what can they approve, what evidence do they need, and what consequence do they carry if the change goes badly? This prevents a common error in persona work, where one fictional profile is made to represent incompatible responsibilities.

The UK Business Data Survey 2026 reports data practices among businesses that handle digitised data and shows variation by business size and data sophistication. Its population and UK geography limit generalisation. It can still inform recruitment criteria, such as whether an organisation has an internal data function or relies on external support.

Include non-users and former users when their experience matters. Someone excluded by access requirements can expose a barrier that active customers have worked around. A former customer may explain why rollout or renewal failed. Neither account should be treated as representative without further evidence.

Recruit for contrast, not convenience

GOV.UK guidance on finding research participants recommends actual or likely users and warns that recruitment channels can introduce bias. Existing customers are easy to reach, but they already accepted the current proposition. Professional networks may over-represent confident, digitally active owners. A supplier referral can omit dissatisfied clients.

Build a recruitment matrix with the differences most likely to alter the product decision. For a scheduling tool, those might include desk-based and field work, one and several locations, direct employees and mixed workforces, or central and local approval. Set a minimum for each relevant contrast rather than filling a single total.

Ask screening questions about facts and recent actions. What happened during the last rota change is more useful than whether someone considers the business innovative. Do not reveal the preferred answer in the invitation. If an agency recruits participants, review the screener, source channels and exclusions rather than outsourcing responsibility for sample quality.

Plan access at recruitment, not when the meeting begins. The same participant guidance covers assistive technology, communication support, accessible formats and travel needs. Make reasonable arrangements without requiring a participant to prove more personal information than the study needs.

Observe the present workflow

An interview tells the researcher what a participant remembers and chooses to describe. Observation can expose hand-offs, unofficial records and interruptions that a polished account misses. Ask the person to walk through a recent completed case. Where confidentiality allows, follow the record from trigger to outcome and note every system, role and wait.

The Service Manual's discovery research guidance recommends examining the current experience, support routes, offline steps, analytics and back-office workflow. For commercial software, this helps separate the user's job from a proposed interface.

Capture the sequence without forcing it into the planned product. Record workarounds and exceptions in the participant's terms. If sensitive screens cannot be shown, ask for a blank example or a verbal reconstruction. Never encourage disclosure of customer or employee data merely to make the session more concrete.

After the session, distinguish direct observation, participant interpretation and researcher inference. A note such as approval took two days describes an observed case only if the dates were actually seen. Otherwise record that the participant recalled a two-day delay.

Conduct interviews without selling

The government's in-depth interview guide recommends open, neutral questions and stories about real examples. Start with role and context, then explore a recent event. Ask what triggered the work, who became involved, what information moved and how the person knew it was finished.

Do not show the concept too early. Once participants see a proposed solution, they may discuss its appearance instead of the underlying process. When a concept must be tested, separate discovery questions from evaluation and note the point at which exposure occurred.

A discussion guide should provide consistent coverage without becoming a script that blocks follow-up. Pilot it with someone outside the target sample to find confusing wording. During the real session, ask for clarification instead of supplying terminology. Avoid questions that combine two issues or invite a prediction about hypothetical buying.

No number of conversations proves an England-wide rate. The GOV.UK planning guidance says participant numbers depend on method and suggests iterative rounds for qualitative research. Treat that as method guidance, not a universal sample-size rule for private-sector market studies.

Use a survey for a narrow measurement task

A survey can estimate a defined measure only when the sample and questionnaire support that inference. It should not be used to rescue an audience that has not been defined. First use interviews or observation to understand vocabulary, plausible answer options and points of variation.

Write one analytical purpose for each question. Remove fields collected just in case. Test the questionnaire with people resembling the intended respondents and watch how they interpret terms such as user, licence, automation or integration. A technically valid form can still produce unusable data if respondents answer different versions of the question in their heads.

Decide whether the survey is descriptive, comparative or intended to estimate a population value. Record the invitation channel, number invited where known, eligible responses, incomplete responses, field dates and exclusions. A voluntary link shared on social media cannot establish a rate for all businesses in England.

Tool selection follows the research design. Compare respondent access, question logic, accessibility evidence, account controls, export formats, retention, supplier terms and the organisation's approved systems. A familiar tool is not automatically appropriate for confidential business research.

Protect participants and research records

Research invitations, screening answers, recordings and notes can contain personal information. GOV.UK's participant privacy guidance recommends explaining collection and use early, minimising the information held, controlling access, setting retention and covering third-party services in relevant arrangements.

That page describes government practice and is not tailored legal advice for a private company. The Information Commissioner's Office states that personal information should be adequate, relevant and limited to what is necessary. Its guidance also notes ongoing updates following the Data (Use and Access) Act 2025.

Before recruitment, document the purpose, lawful basis, notices, access, processors, retention and deletion process. Keep contact information apart from analysis where practical. Do not publish a quotation because a name was removed if the role, location and incident could still identify the speaker. Obtain review from a qualified UK data-protection professional for the actual study.

Analyse evidence without inventing a persona

Complete first-pass notes while the session is fresh. The Service Manual's analysis guidance recommends separating observations from interpretation and involving the team promptly. Preserve links back to session records so a summary can be checked.

Use a finding table with participant code, evidence extract, workflow stage, affected role, confidence and contrary examples. Look for patterns across independent sessions, but retain meaningful disagreement. A problem experienced by one administrator may still matter if its consequence is severe; describe its observed scope instead of inflating its frequency.

Personas or role profiles should be outputs of research, not creative writing. Omit decorative names, stock photographs and invented biographies. Include only characteristics that change product or purchasing decisions, and cite the sessions or dataset behind each point. Mark hypotheses that need another round.

Turn findings into a recorded choice

End a research round by answering the decision written at the start. Use one of four outcomes: evidence supports proceeding, evidence supports a narrower segment, the question remains unresolved, or the current proposition should stop. List what changed in the product, message, sales process or research plan.

Maintain a short limitations section. It should name missing roles, recruitment bias, geography, timing and any material change to the guide. Do not convert favourable comments into testimonials without separate permission and review. Do not report a qualitative pattern as a percentage.

The immediate next step is to choose one product decision and write a recruitment matrix for the people who can inform it. Only then prepare a screener or interview guide. This article remains on publication hold until a named editor, fact-checker and qualified methodology, accessibility and data-protection reviewer approve it.

Before you act

Common questions

Why should a team name the decision before writing a discussion guide?

Research becomes unfocused when every stakeholder adds a favourite question. Naming the decision first, such as whether supervisors need a mobile approval route, keeps the study tied to a choice the team must make. The GOV.UK Service Manual advises turning unfounded assumptions into research questions and choosing methods that answer them with reasonable effort.

Can official business statistics select a research sample on their own?

No. Official statistics can describe the population from which participants may be drawn, but they cannot select the audience. The Department for Business and Trade put England's private-sector business population at 5.0 million for the start of 2025, and the ONS provides registered enterprise tables. Select the dataset resembling the intended buyer base and document what it leaves out.

How should a researcher handle a concept test during interviews?

Do not show the concept too early. Once participants see a proposed solution, they may discuss its appearance instead of the underlying process. When a concept must be tested, separate discovery questions from evaluation and note the point at which exposure occurred. The government's in-depth interview guide recommends open, neutral questions and stories about real examples.