Audience research
Part of Audience research for a website project that keeps buyers and end users apart
Buyer hypothesis cards for web design agencies: fields and confidence
A blank hypothesis card for a web design buyer in England: the fields to fill, plain confidence labels and how each claim traces back to evidence.
A persona should be an editable evidence summary, not a made-up person decorated with a stock photograph and a quotation nobody said. Until fieldwork supports it, label the card "hypothesis" and show the gaps that could overturn it.
No buyer or end user has been researched for these cards, so the method holds no finding about England demand, demographics or purchasing behaviour. What it does hold is a way to keep a card honest once evidence arrives.
What to take away
- Give every card a plain confidence state, never a percentage invented by the team.
- Record who coded each item and keep a route back to the original evidence.
- Give each card an expiry date, and revise or delete it when evidence contradicts it.
- Publish the method and decision pattern rather than the person when identification risk cannot be ruled out.
- A card that changes nothing is decoration, so end it with a test of what the team will do differently.
Start from two cards
Start with two cards, because the commissioner may never use the public journey and the person using it may have no purchasing influence. A buyer card describes someone who shapes a website decision. An end-user card describes a task, setting and barrier.
Buyer vs end-user hypothesis card
Buyer hypothesis card
- Population
- authority over scope or spend
- Geography
- decision location in England
- Decision or task
- procurement or operating choice
- Evidence needed
- approval path and constraints
- Disproof
- fact changing buying assumption
End-user hypothesis card
- Population
- eligible to attempt a task
- Geography
- residence or eligibility rule
- Decision or task
- outcome the person seeks
- Evidence needed
- behaviour, barriers, language
- Disproof
- observation changing task assumption
Never add a personal name, age, portrait or verbatim quote unless it is necessary, authorised and traceable to consented evidence. Most planning decisions need a role and context, not a biography.
Fields that make a card traceable
Leave unknown fields blank rather than guessing. A card should carry a stable identifier, eligible population, source references, research date, recruitment route, contradictory evidence, excluded groups, confidence rationale, decision owner and expiry date.
Minimum evidence fields per card
- Stable identifier
- Eligible population
- Source references
- Research date
- Recruitment route
- Contradictory evidence
- Excluded groups
- Confidence rationale
Do not turn an ONS business category into a buyer personality. Administrative statistics can define organisational sampling frames, but they do not reveal budget authority, website frustration or willingness to appoint an agency. State the denominator behind any count, as the guide to measurement errors warns.
An untested buyer card for a website rebuild
Here is a card before any fieldwork. Every entry is a hypothesis, so the confidence field says so and the card names what would overturn it.
| Field | Entry on an untested card |
|---|---|
| Role | Person who approves scope, spend or acceptance for a website rebuild |
| Decision | Which agency to appoint and what counts as finished work |
| Eligible population | Organisations in England that commission an agency, defined before recruitment |
| Evidence | None yet |
| Confidence | Untested |
| What would overturn it | Interviews showing that a different role approves scope |
| Decision owner | A named person on the team |
| Expiry | A date written on the card |
The end-user card for the same project names a task instead, such as finding contact details on a phone. In a web design agency directory, the buyer card shapes which questions a profile should answer for the person approving scope. It rates no agency.
Coding evidence back to its source
Code each finding as soon as possible after a session. The GOV.UK Service Manual says to do analysis as soon as you can after each round of research, while it is still fresh in people's minds.
The Service Manual describes sorting notes into clusters until clear groups emerge, which it calls affinity diagramming or affinity sorting. Use that method, and name each group.
For a persona card, record who coded each item and retain a route back to the original evidence while access remains authorised. Separate what people said from what the researcher observed, and compare findings with operational records whose definitions and permissions are understood.
Then seek a challenge review from someone who did not moderate the sessions. The reviewer reads the card against the coded evidence and says which statements the evidence does not support. A card cannot move up a confidence state without that review.
Confidence states and expiry
Confidence is a plain state, not a number. State the population and limitation beside the label, so a reader can see what the label covers.
| State | What it means | What the team does |
|---|---|---|
| Untested | No fieldwork supports the card yet | Treat it as a hypothesis and plan the research |
| Supported within this sample | Evidence from eligible participants supports it | State the population and the groups the sample missed |
| Contradicted | Later evidence overturns it | Revise or delete the card |
| Expired | The expiry date has passed | Re-test the card or retire it |
Begin the cycle from the business assumption and its source. Write what future evidence will make the team revise or delete the card, and put the date on it.
Privacy and publication boundaries
Participant records should be minimised and access-controlled. The ICO's data minimisation guidance says to identify the minimum amount of personal data you need to fulfil your purpose. A named privacy reviewer should set the lawful basis, special-category condition where applicable, retention and withdrawal handling.
Publish a persona only if its intended use was explained, identification risk is assessed and the evidence does not expose a participant through a distinctive combination of details. Otherwise publish the method and decision pattern, not the person.
A useful card ends with a test: what will the team do differently? If there is no answer, the persona is decoration rather than research.
Before you act
- Label the card hypothesis until evidence supports it.
- Choose a plain confidence state and write the expiry date.
- Leave unknown evidence fields blank rather than guessing.
- Record who coded each item and keep a route to the evidence.
- Get a challenge review from someone who did not moderate.
- Assess identification risk before publishing anything.
Common questions
What should a hypothesis card contain before any research has been done?
It should carry a stable identifier, eligible population, source references, research date, recruitment route, contradictory evidence, excluded groups, confidence rationale, decision owner and expiry date. Unknown fields stay blank. It should also show the gaps that could overturn the assumption, and state who the sample missed.
How does a hypothesis card earn confidence?
Start with the business assumption and its source, then run authorised research with eligible participants. Separate what people said from what the researcher observed, and compare findings with operational records. Then seek a challenge review from someone who did not moderate the sessions. Confidence is recorded as a plain state, not a percentage.
When should a persona be published, and when should it not?
Publish only if the intended use was explained, identification risk is assessed, and the evidence does not expose a participant through a distinctive combination of details. Otherwise publish the method and decision pattern instead of the person. Participant records should be minimised and access-controlled, with a named privacy reviewer setting retention and withdrawal handling.
