Accounting services audience research: a practical method

Accounting services audience research should reveal who has a defined task, what records and authority shape it, and where the present route fails. It should not decorate a broad SME segment with invented personas or treat a convenient client list as evidence about the wider market.

This guide uses UK government and regulator sources. Government user-research material is a useful methodological reference, not a legal standard for private accountancy practices. Tax and company rules must be checked in their own scope.

What to take away

Frame the decision first, then choose methods that can answer the priority questions.
Define audiences by delivery variables such as entity, trigger, records and authority, not broad labels.
Map all roles in the purchase, including record holders and approvers, not just the public decision-maker.
Recruit actual or likely users inclusively and track recruitment sources to avoid bias.
Build personas from observed patterns and retire attributes that never change a service decision.

Frame a decision before collecting data

Write the action that the research will inform. Examples include whether to launch a service for the April 2027 MTD cohort, which filters an accountant directory needs, or how to improve handover from qualification to bookkeeping.

Turn beliefs into questions:

Frame a decision

Which actual event causes the customer to seek help?
What must be completed, and by when?
Which person controls the records and approval?
Where does the customer currently go?
What prevents accurate or timely completion?
Which cases can the proposed team serve competently?

State the evidence threshold for the decision. Interviews may explain an unfamiliar journey, but they do not estimate national incidence. A survey can measure a defined population only if the sample, response and questionnaire support the inference. Operational records show behaviour inside an existing service, not people who never entered it.

The GOV.UK guide to planning user research recommends setting objectives, identifying relevant groups and choosing methods that can answer the priority questions. Adapt this discipline to the accounting decision.

Define the audience through delivery variables

"Small business owner" hides too much. A research segment should contain factors that materially alter the service:

Audience delivery variables

Dimension

Entity: Individual, partnership, company, charity
Activity: Trade, property, profession, retail, mixed
Work: Records, payroll, accounts, tax, advice
Trigger: Incorporation, first worker, notice, growth
Record condition: Reconciled, fragmented, incomplete, unavailable
Authority: Owner, director, finance employee, trustee
Dependency: Bank, software, previous adviser, payroll
Access: Online, telephone, office, assisted formats

Examples

Entity
Activity
Work
Trigger
Record condition
Authority
Dependency
Access

Define the audience

Dimension	Examples
Entity	Individual, partnership, company, charity or group
Activity	Trade, property, profession, retail or mixed income
Work	Records, payroll, accounts, tax, reporting or advice
Trigger	Incorporation, first worker, notice, error, growth or switch
Record condition	Reconciled, fragmented, incomplete or unavailable
Authority	Owner, director, finance employee, trustee or delegate
Dependency	Bank, software, previous adviser, payroll or customer system
Access	Online, telephone, office, assisted and accessible formats

Use geography only where it affects the question. England may matter for local discovery or face-to-face service, while many tax and company processes are UK-wide. Record where the customer is based, where the provider operates and where the service is delivered instead of assuming they are identical.

The ONS UK business activity, size and location data can help researchers understand registered enterprise and local-unit structure. It cannot identify the needs, buying role or record quality of individual firms.

Map roles around the purchase

One person may search while another approves. The employee who uses bookkeeping software may not sign the engagement. A previous accountant or bookkeeper can control the practical handover, and a director still needs to approve important figures.

Map six roles where relevant:

Problem experiencer:: the person who encounters the accounting task or problem.
Initiator or searcher:: the person who starts looking for a service.
User or operator:: the person who does the bookkeeping or uses the service.
Records holder:: the person who controls or supplies the records.
Recommender or influencer:: the person whose advice affects the choice.
Approver or budget holder:: the person who authorises the appointment or spend.

One person may hold several roles, and not every purchase involves all six.

Six roles around a purchase

Experiences the problem
Gathers provider information
Holds source records
Approves budget or engagement
Authorises submissions
Influences the choice

Recruit more than the public-facing decision-maker. Observation of the staff member correcting transactions may reveal needs that an owner does not see. Interviews with the owner may explain risk and authority that an operator cannot decide.

Recruit actual or likely users inclusively

GOV.UK advises researchers to recruit actual or likely users and to include disabled participants and people who may need support. Write a screening brief from the segment definition, including recent task, role, system, record state and communication needs.

Use several sources. Existing customers bring detailed experience but have already accepted the practice's route. Enquiries reveal earlier-stage questions. Former clients, switchers and people who declined can expose unmet expectations. Self-servers show why a paid service did not win. Referral partners should not supply every participant.

Track recruitment source and exclusions. Do not let a marketer replace a hard-to-reach group with easier respondents who look similar on age or turnover but use a different accounting process.

Provide information about purpose, participation, recording, data use, withdrawal, incentives and contact routes. The GDS informed-consent guide treats consent as continuing through the session rather than a one-off signature.

Research participation and direct-marketing permission are separate. Do not add participants to a sales list merely because they agreed to an interview.

Protect financial and personal information

Accounting research can accidentally collect tax references, bank details, payroll records and information about other people. Design sessions so participants can describe their work without sharing live credentials or unredacted files.

Create a data plan covering purpose, lawful basis, minimisation, access, storage, recording, transcription, retention, deletion, processors and international transfers. The ICO's guide to lawful basis explains that the organisation should determine and document an appropriate basis before processing.

Use participant codes in research notes and store contact details separately. Train moderators to pause if unexpected sensitive material appears. Provide a defined incident route instead of improvising after a screen recording captures confidential information.

Anonymisation requires more than removing a name. A rare trade, town, filing event and quotation can identify a business when combined. Review proposed case studies with this risk in mind.

Conduct interviews around a recent event

Ask the participant to reconstruct what happened from trigger to outcome. Useful prompts include:

What happened, and what first prompted you to seek help?
What did you do next, and who else was involved?
Which records or information did you need, and what was missing?
Where did you look for help, and what alternatives did you consider?
What made the process difficult, delayed or confusing?
How was the provider chosen, and what happened after the work was completed?

Recent-event interview prompts

What made you act at that time?
What did you need to finish?
Which organisation and accounting period?
What records and software did you start with?
Who helped, approved or delayed?
How did you find and compare options?
How did you recognise completion?

Use follow-ups that seek events and evidence. "What happened after that?" is less leading than "Would faster replies have helped?" Do not pitch the planned product during discovery.

Make clear that a research session is not tax or accounting advice. If the participant raises a live urgent problem, follow an agreed signposting route without allowing the interview to become an unrecorded engagement.

Invite relevant team members to observe within privacy controls. Ask them to record what they saw and heard before interpreting it. Analyse promptly while the sequence remains clear.

Build personas from patterns, not fiction

Extract observations with source codes, then group cases around decision variables. A useful persona may be a deadline-led adviser switcher with incomplete handover, a landlord preparing for digital reporting, or a finance employee researching providers for director approval.

The title is shorthand. The persona record should state:

Build personas from patterns

supporting and contradictory cases;
recruitment routes and sample limitations;
trigger and job to be completed;
authority and dependencies;
record condition and support needs;
qualification and exclusion questions;
evidence date and owner.

Avoid stock photographs, demographic decoration and invented quotations. Those elements make an artefact feel human while weakening its evidential basis.

Test each profile against edge cases. If the team uses the persona label to skip checking entity, records or authority, withdraw it. Retire attributes that never change a service decision.

Use MTD as a recruitment criterion carefully

HMRC's MTD for Income Tax guide sets out the staged process. For 6 April 2027, it identifies a qualifying-income threshold above £30,000 based on the 2025 to 2026 tax year, subject to detailed scope and exemptions. The Income Tax (Digital Requirements) Regulations 2021 set out digital requirements. HMRC guidance explains the exemption process, including the route for people who cannot reasonably use digital services because they are digitally excluded. Check HMRC's criteria and process; the threshold alone does not establish that someone is in scope or exempt.

Do not recruit a generic "landlord" sample and assume it represents this cohort. Screen the relevant income sources and stage, while avoiding a research moderator making an eligibility determination beyond their role.

Separate people already using compatible software, people with digital records in another form, those requiring substantial clean-up and those seeking exemption or support. Their service needs and costs differ markedly.

Select survey methods by inference needed

Surveys can measure structured questions after qualitative work has clarified the language and answer options. Choose the sampling frame before the platform.

Possible routes include a website intercept, client email, telephone questionnaire, research panel, assisted paper form and follow-up after a service event. Each excludes or favours different people. Combining routes requires consistent administration and a plan for duplicate responses.

Before data collection, define:

the target population and the decision the research will support;
inclusion and exclusion criteria, plus the sampling frame and recruitment routes;
the research questions, method and measures;
fieldwork dates, analysis approach and reporting limitations;
data handling, access, retention and deletion.

Define before survey collection

Target population and sampling frame
Question wording and reference period
Numerator, denominator, partial responses
Minimum subgroup evidence
Handling of free text
Accessibility testing
Personal-data controls
Response and non-response reporting

Pilot with intended users. A person may interpret "accounts service" as bookkeeping, statutory accounts, management reports or audit. Fix ambiguity before scale.

Do not report an open web poll as representative of businesses in England. Publish fieldwork dates, counts, response rate, source and exclusions beside the result.

Evaluate survey platforms on control

Compare software only after the collection route is selected. Assess accessible input, supported logic, authentication, data location, subprocessors, permissions, export, retention, deletion, backup, incident notification and contract exit.

Preserve the questionnaire version and raw export. Restrict identifiers and document every transformation. Supplier dashboards may change calculations or suppress details; retain enough information to reproduce the published figure.

A simple telephone script with controlled entry may be more inclusive than a complex online tool. Conversely, an online form may be proportionate for a digitally defined population. Match the system to the research rather than buying features first.

Research competitors symmetrically

Set the competitor field from the audience's real alternatives. It may include local practices, remote firms, bookkeeping specialists, software-led services, directories and self-service. Record why each route is included.

For named firms, capture current legal identity, service claims, professional designations, supervision statements, delivery channel, scope, pricing basis and evidence date. Check designations in the relevant body's directory, such as those of ICAEW, ACCA, AAT or CIOT, and verify the specific status rather than relying on a website badge. For statutory audit, check Companies House filings for the stated auditor and verify the auditor's registration with the responsible recognised supervisory body. Check designations with issuing bodies and confirm statutory-audit registration where claimed. A website badge is not enough.

Normalise commercial comparisons using the same entity, transaction level, worker count, work scope, contract period and software assumptions. Mark missing prices as undisclosed, not costly. Never submit invented business information to force a quotation.

Treat testimonials and public review scores as prompts for questions. They do not establish service quality, compliance or the likely outcome for another customer.

Apply the same language to every provider. If one unsupported claim is labelled unverified, an equivalent claim from a favoured firm receives the same treatment. Offer organisations a correction route before publishing material factual conclusions.

Combine evidence without false certainty

Create a matrix showing what each source can answer:

Evidence source	What it can answer	Limitation
Interviews about recent events	Triggers, tasks, decisions, barriers and handovers	Explain experiences; do not estimate national incidence
Existing service and enquiry records	Patterns among people who reached the practice, including recorded behaviours and questions	Exclude people who never entered that route and reflect the practice's existing reach
Survey of a defined population	How common specified answers or behaviours are within the population sampled	Inference depends on the sampling frame, response and questionnaire
ONS UK business activity, size and location data	Registered enterprise and local-unit structure	Does not identify individual firms' needs, buying roles or record quality
HMRC MTD for Income Tax guidance	MTD scope, stages and exemption information relevant to recruitment	Does not reveal demand for an accounting service
Competitor research	The provider and self-service alternatives an audience can consider	Public claims do not establish service quality or buyer incidence

What each evidence source answers

Source

Interviews: Explanation and chronology
Survey: Structured measures
Service data: Observed use and outcomes
ONS data: Population and industry context
Competitor records: Current stated offer

Strength

Interviews: Small purposive sample
Survey: Question and non-response effects
Service data: Existing users only
ONS data: No individual need or role
Competitor records: No proof of customer outcome

Limitation

Interviews
Survey
Service data
ONS data
Competitor records

Triangulation does not mean averaging incompatible numbers. Look for findings that converge through different evidence and preserve disagreements.

Distinguish observation, interpretation and recommendation. "Four participants uploaded the same document twice" is an observation. "The labelling was unclear" is an interpretation to test. "Rename the upload step" is a proposed action requiring usability evidence.

Turn findings into bounded changes

Give every finding a source, confidence, affected audience, counter-evidence, expiry and decision. Prioritise problems by consequence and frequency within the researched population, not by how vivid one quotation sounds.

Convert findings into qualification questions, engagement wording, directory filters, support routes, product boundaries or experiments. Assign an owner and success measure.

Run the changed service with matching users. Measure record correction, repeated questions, completion, customer effort and exceptions. Satisfaction alone cannot show that a submission was accurate or the engagement scope was met.

Maintain a research repository with approved access and retention. Revisit the audience after an HMRC process, company-filing route, software dependency, pricing model or service channel changes.

The result should be a smaller set of defensible decisions. Good audience research shows whom the service can help, what evidence it needs and where it must decline or refer. That is far more valuable than a colourful portrait of an imaginary typical accountant customer.

Before you act

Write the action the research will inform before collecting data.
Define audience segments using delivery variables that alter the service.
Map all roles involved in the purchase decision.
Recruit actual or likely users from diverse sources.
Create a data plan covering minimisation, storage and deletion.
Extract observations with source codes and group cases by decision variables.

Common questions

What should accounting audience research reveal?

It should reveal who has a defined task, what records and authority shape it, and where the present route fails. It should not decorate a broad SME segment with invented personas or turn a convenient client list into evidence about England.

Why is small business owner an insufficient research segment?

It hides factors that materially alter the service. These include entity, activity, work and trigger. They also include record condition, authority, dependency and access. A research segment should contain these delivery variables so that the service can meet specific needs.

How should researchers handle personal and financial data during accounting research?

Design sessions so participants can describe work without sharing live credentials or unredacted files. Create a data plan covering purpose, lawful basis, minimisation, access, storage, retention and deletion. Use participant codes and store contact details separately. Train moderators to pause if sensitive material appears.