Before you commission marketing agencies audience research

What to take away

Why audience research briefs go wrong

The typical failure is not a bad agency. It is a brief that asks for "insight into our audience" with no named decision attached. Agencies then optimise for the deliverable they can sell, usually a presentation, rather than the answer you needed.

A second failure is treating audience research as a one-off. Audiences shift, channels shift and competitor positioning shifts with them. A single study in year one becomes a stale assumption by year three.

A third is ignoring what is already known. Many client teams hold search data, CRM records and sales call notes that would answer half the brief. Commissioning fieldwork before reading your own data wastes money.

Fix all three by writing a one-page decision statement before you approach anyone. State the choice you are making, the date it must be made and the cost of getting it wrong.

Start with the decision, not the method

Write down what will change once the research lands. If nothing changes, stop.

Useful decisions look like this: which two sectors to prioritise next financial year, whether to move budget from paid search into trade press, or whether a proposition needs rewriting before a spring campaign.

Each of those needs different evidence. Prioritising sectors needs sizing and growth data. Reallocating budget needs channel behaviour. Rewriting a proposition needs message testing.

Once the decision is written, the method largely picks itself. You can then ask agencies to challenge your method rather than invent one, which shortens proposals and reduces the padding in them.

Sizing the market properly

Start with published population and household data rather than a vendor's estimate. The census data published by the Office for National Statistics gives counts by area, age and household type, which is enough to build a defensible top-down size for most consumer briefs.

For business audiences, layer in company counts and sector splits from Companies House filings, then apply your own conversion assumptions openly. Label every assumption in the model so a reviewer can challenge it.

Do not let an agency present a single market size figure without its workings. Ask for the base, the filters and the growth assumption separately. If the base is a paid panel estimate, say so in the deck.

For a worked example, a Leeds training provider targets operations managers in manufacturers with 50 to 250 staff. A top-down model shows 1,400 qualifying firms, averaging two relevant managers each, so 2,800 named buyers.

If it wins 1 per cent at an average contract of £3,600, that is £100,800 annual revenue from the segment. Those figures are illustrative, not researched.

Reading the numbers like a planner

Media planners lean on syndicated audience data because it is consistent and comparable across channels. The TouchPoints audience data published by the IPA is a standard reference for how people spend time across media, and it is a sensible starting point when you need to compare channel reach.

Syndicated data is strong on behaviour and weak on your specific proposition. Use it to narrow channel choices, then use your own research to test the message within those channels.

Watch the base sizes in any cross-tab you are shown. A segment of 40 respondents will move wildly on small changes, and agencies sometimes present those cuts without a warning.

Ask for confidence intervals or margins of error in writing. If the agency cannot supply them, treat the numbers as directional and say so when you report internally.

Choosing between qualitative and quantitative

Qualitative work answers why. It suits proposition development, objection handling and early category exploration. Samples are small by design, so never quote percentages from them.

Quantitative work answers how many and how much. It suits sizing, prioritisation and tracking. It needs a clean sample frame and a questionnaire that has been piloted before full field.

Most briefs need both, in sequence. Run a small qualitative stage to find the language, then a quantitative stage to measure how widely that language lands.

A practical rule for a mid-sized UK brief is six to ten depth interviews followed by 300 to 500 survey responses. Those ranges are illustrative, and the right numbers depend on how many segments you intend to report.

For a structured approach to the qualitative stage, our guide to running customer interviews that produce usable evidence covers recruitment, questioning and how to avoid leading participants.

The evidence an agency should show you

Ask for the sample source, the fieldwork dates, the incentive used and the weighting scheme. These four items separate a research supplier from a content producer.

Check whether the sample comes from a panel, a customer list or a river sample. River samples, where respondents arrive through open web links, are cheap but rarely representative, so they should not be used for sizing.

Look at the questionnaire before field. A question that offers seven brands in a fixed order will produce order effects unless it is rotated. Ask whether rotation was applied.

Finally, ask what the agency would do differently with twice the budget, and what it would cut with half. The answers reveal which parts of the method it actually believes in.

Building personas you can use

Personas are a summary device, not a research method. They should be built from the data you have collected, not invented in a workshop.

Keep them small in number. Three to five personas will be used; twelve will be ignored. Each one needs a trigger, a barrier and a proof point, not just a name and a stock photograph.

Tie each persona to a measurable behaviour, such as the search terms they use or the content they download. That makes them testable rather than decorative.

Our walkthrough on building buyer personas that survive contact with sales shows how to connect persona work to CRM fields so the sales team can validate or challenge them.

What the regulators are watching

Advertising research does not sit outside the rulebook. The research published by the ASA and CAP sets out the consumer attitudes and compliance themes that shape how claims are assessed, which is useful context when you test advertising concepts.

If your audience includes US consumers, the studies published by the Federal Trade Commission cover advertising and consumer protection topics that may affect claim wording and substantiation.

This matters for agencies because research findings often become marketing claims. A test showing that 62 per cent of respondents preferred a product is a claim that needs a base, a date and a method behind it.

Build a simple evidence file for every claim you intend to publish. Store the questionnaire, the sample description and the raw tables together.

Where advertiser collaboration helps

Some audience questions are too expensive for one brand to answer alone. Industry bodies exist partly to pool that work.

Looking at ISBA's member groups shows how advertisers organise around shared interests, and the same logic applies to commissioning research with two or three non-competing brands.

Shared studies work best when the audience definition is identical and the propositions are not. If two brands sell competing products, keep the questionnaire sections separate.

Agree the data-sharing terms before fieldwork starts. Who owns the raw data, who can publish, and what happens if one party withdraws are the three questions that cause trouble later.

Commissioning and quality control

Write a brief with a decision, a deadline, a budget range and a named internal owner. Agencies respond better to constraints than to open questions.

Ask for two methodological options at different price points. This forces the agency to explain trade-offs instead of defending a single approach.

Pilot every questionnaire on five to ten people who match the target profile. Piloting catches ambiguous wording and missing answer options at very low cost.

Review the raw tables, not just the report. Discrepancies between the two usually mean the report has smoothed over a weak segment.

Set a rule that any figure in a final deck must be traceable to a table. If it is not, it does not go in.

For a full picture of how these findings feed reporting cycles, the measurement and reporting guide for agency work explains how to connect research inputs to performance outputs.

Working with a specialist agency

Generalist agencies can run research, but the strongest briefs often go to a specialist with a defined method. Decide which you need before you shortlist.

The marketing agency directory on this site is a reasonable place to start when you want to compare how suppliers describe their research credentials, though you should still check the underlying sample approach yourself.

Ask for two recent examples with the decision they informed. A supplier that can describe the decision, not just the sample size, is usually the better partner.

Check insurance, data protection registration and how respondent data will be stored and deleted. These are basic checks, and their absence is a warning sign.

For a wider comparison process, our guide to running a supplier comparison before you brief covers the tools and questions that make shortlisting faster.

Costs and timescales in England

Costs vary widely by method, sample and sector difficulty. Treat the figures below as illustrative planning ranges for a UK brief, not quotations.

A mid-sized programme combining desk research, interviews and a survey typically lands between £15,000 and £30,000. A tracking study will cost more because it repeats.

Timescales slip when recruitment is hard. Business-to-business samples with narrow job titles can take twice as long as consumer samples, so build contingency into the plan.

Pay attention to the split between fieldwork and thinking. If 90 per cent of the quote is fieldwork, you are probably buying data rather than judgement.

Common questions

How much should a small firm spend on audience research?

A small firm can get a long way on free public data plus six depth interviews, which might cost £3,000 to £5,000 with a freelance researcher. Those figures are illustrative. Spend more only when the decision at stake is worth more than the research.

Can we use the agency's own panel data?

Yes, but ask who owns the panel, how members are recruited and how often they are reused. Panels that over-sample the same respondents can produce flattering but unreliable results, particularly for niche business audiences.

How often should audience research be repeated?

Repeat the sizing model annually and the qualitative work when the proposition changes or a new segment opens. Tracking studies usually run quarterly or twice yearly, depending on budget and volatility.

What is the single biggest warning sign in a proposal?

A proposal with no stated sample frame and no margins of error. If the method section cannot explain where respondents come from, the findings will not survive scrutiny from a finance director or an external auditor.