Audience research
Part of How to research a multilingual audience without labelling people by language
Defining a multilingual social marketing audience in England: Census 2021 language figures and UK GDPR consent
Define a multilingual audience from a recent task, using Census 2021 main-language figures for England and the ICO's UK GDPR consent and withdrawal rules.
Define a multilingual social marketing target audience through a recent task and an England eligibility rule. Do not start with an inferred language group. A useful definition tells the researcher who can answer the decision, how to recruit them and what the sample cannot represent. The audience research guide covers the wider method.
What to take away
- Start with the decision under review, not an assumed language group or identity.
- Census 2021 gives England-level context on main language, not the language a person wants for one service.
- Ask eligibility questions directly. Never infer language from a name, accent or postcode.
- A volunteer sample can reveal problems but cannot estimate how common they are in England.
- Consent under the UK GDPR must be freely given, specific, informed and unambiguous, and people can withdraw it at any time.
Write the decision before the audience label
Start with the action under review. One example: do adults who recently used a local service in England need a language-specific social explainer for one step? Record the service, source content, proposed language, channel, decision owner and review date.
Separate the people who may see the content from the organisations that commission it. Buyers, budget approvers and platform operators answer different questions. A buyer can describe procurement but cannot stand in for audience comprehension.
Use Census 2021 for context only
The ONS bulletin on language covers England and Wales, with figures for England alone as well. In ONS Census 2021 language data, 90.8% of people in England (49.7 million out of 54.7 million) had English as a main language, against 96.7% in Wales.
Across England and Wales together, 91.1% of usual residents aged three and over had English (English or Welsh in Wales) as a main language. The England figure is the one to use for an England audience.
| Main language other than English (England, 2021) | Share of usual residents | People |
|---|---|---|
| Polish, most common in England | 1.1% | 591,000 |
| Romanian, next most common | 0.9% | 466,000 |
| Panjabi, highest in the West Midlands | 1.4% of that region | 83,000 |
| Urdu, highest in the North West | 0.8% of that region | 59,000 |
Regions differ widely. The North East had the highest share with English as a main language (96.5%), and London the lowest (78.4%, 6.7 million people). These figures show the scale of main languages. They do not show which language a person wants for one service, whether they use a social platform in it, or whether they agree to be contacted. Use them to choose which languages to study, not to size an audience.
Set observable eligibility
For an audience study, specify:
- residence or recent service use in England, with a stated way of checking it;
- a recent task that matches the proposed content;
- age or other criteria only where the decision needs them;
- self-described spoken, written and preferred research languages;
- the format, technology or help the person asks for;
- exclusions, and what each one does to the decision.
Ask people directly. A name, accent, country of birth, postcode or social profile does not establish language, literacy, vulnerability or consent, and guessing from them raises data protection questions of its own.
Build a recruitment frame
Choose a frame tied to the task: recent service users who agreed to research invitations, a documented support route, or eligible members reached through several community organisations. Note the frame date, source, coverage and missing groups.
The GOV.UK Service Manual guidance on finding user research participants is written for government services, so a commercial researcher must justify its own method. Its advice is still useful. Recruitment criteria should clearly specify the people you want to participate. Agencies are generally best for recruiting the general public. Allow up to a month to find participants.
Track invitations, undelivered messages, screened people, eligible people, participants and withdrawals by route and language. Use the customer interviews article for the consent and neutrality questions.
Make taking part understandable
Provide the information sheet, consent route, incentive terms and withdrawal process in the person's chosen language and in a format they can use.
Contact details and recordings are personal data, and most processing of personal data is subject to the UK GDPR. The ICO guidance on valid consent defines consent as a freely given, specific, informed and unambiguous indication of a person's wishes. People must be able to refuse consent without detriment and to withdraw it easily at any time.
So the information sheet should say what the purpose is, what data is collected, who sees the results, how long data is kept and that taking part is voluntary. Write it in language your participants can understand.
Use a qualified native-level reviewer for each language and a subject specialist for technical claims. Bilingual identity alone is not proof of translation skill. Test the source and translated materials with eligible people before fieldwork, and use the quality checklist to record who checked what.
Taking part gives no permission to reuse a testimonial, target the person or contact them to sell. Ask a privacy reviewer to decide any later contact.
Decide within the evidence
Report the population, frame, recruitment dates, response flow, questions in each language, access arrangements, analysis and exclusions. Say which findings are observations, which are interpretations and what counter-evidence would change the conclusion.
- Proceed only when the evidence supports one bounded content or research decision.
- Redesign if the language versions are not equivalent or key groups are missing.
- Stop if people cannot take part safely, voluntarily and accessibly.
Assuming an identity first is one of the common strategy mistakes because it recruits people who cannot answer the decision. Pair the results with buyer personas that expire when the evidence does.
Before you act
- Record the service, content, language, channel, owner and review date.
- State how residence or recent service use in England is checked.
- Ask about spoken, written and preferred research languages directly.
- List the frame date, source, coverage and missing groups.
- Give consent materials in the person's chosen language and format.
- Report exclusions and what counter-evidence would change the conclusion.
Common questions
Can Census 2021 tell me how big my multilingual audience is?
No. The ONS figures cover main language for residents of England and Wales, with England reported separately. They do not show which language a person wants for one service, how they use social platforms, or whether they would agree to be contacted.
Why avoid starting from an assumed language group?
A group label does not establish language, literacy, proficiency or consent. Start from the decision and a recent task, then ask people directly.
What must happen before fieldwork on translated materials?
Have a qualified reviewer check each language and a subject specialist check technical claims. Test the materials with eligible people, and provide consent information in the person's chosen language and format.
