How do you research the audience for connected TV advertising?

What to take away

Why does connected TV advertising need its own audience research?

Connected TV is not linear television with a different plug. The audience is assembled from households that may never watch a shared broadcast feed, and from individuals inside those households who are hard to identify. A single account can serve several people, and a single person can use several services.

That fragmentation breaks the assumptions behind older planning tools. A panel built for broadcast measurement may under-represent ad-free tiers or short-form app viewing. A survey built for desktop display will miss co-viewing entirely. Research has to be designed for the medium rather than borrowed from it.

There is also a commercial reason. Streaming inventory is often bought programmatically, with targeting applied at the point of sale. If your audience definition is vague, the platform will fill the gap with its own assumptions. Good research is how you keep control of that definition.

What does the UK viewing data actually show?

Start with volume. BARB publishes a Total Identified Viewing summary that sets out how much television and video content people in the UK watch, split by device and platform type. It is the closest thing the market has to an agreed baseline, and it is the number agencies will reach for when your plan is challenged.

Then look at the split between streaming and scheduled television. Nielsen's The Gauge tracks monthly share of total TV time by category. In its July 2025 release, streaming reached 46.4% of US television time while broadcast and cable together took 44.8%. Treat that as a signal about behaviour, not as a UK forecast.

For the UK, the practical move is to combine a total viewing baseline with platform-level research. The IAB UK insight archive carries market research on digital advertising, including connected TV and streaming, and is useful for sizing and for understanding how buyers and viewers describe the channel.

Which research methods fit which decision?

Different questions need different instruments. Reach and frequency questions need panel or platform data. Attitude questions need survey work. Depth questions need interviews. Behavioural questions need log or sales data.

A useful rule is one method per decision. Running a survey to answer a reach question wastes money and produces a number nobody should trust.

Sampling decides whether any of this is worth reading. Quotas should match the population you care about, and weighting variables should be the ones that shift viewing behaviour. Weighting on age alone rarely fixes a sample with too few ad-free households.

How should you segment a streaming audience?

Segment on behaviour you can act on, not on demographics you cannot buy. Platform, device, co-viewing pattern, ad tolerance and category involvement are all workable in connected TV planning. Age and region matter, but only when they change the buy.

A practical starting set is four or five segments. More than that and the media plan becomes unmanageable; fewer and you lose the differences that change creative or placement. If you need a worked structure, the buyer personas for connected TV planning in England show how to build segments that map onto targeting options rather than onto abstract labels.

Co-viewing deserves its own treatment. A household of four watching one screen is not the same as four individual impressions, and research that ignores this will overstate personal reach and understate shared attention.

What can surveys tell you that panels cannot?

Panels tell you what happened. Surveys tell you what people think happened, and why. Both are useful, and neither replaces the other.

Surveys are strongest on attitudes: tolerance for ad frequency, recall of creative, perception of a category, and stated reasons for skipping or staying. They are weakest on precise exposure, because self-report is unreliable at the level of individual impressions.

The ASA and CAP publish research on advertising issues, including consumer attitudes to online and video advertising. That material is worth reading before you write your own questionnaire, because it shows how regulators frame the questions and where public concern concentrates.

Question order and scale design are where surveys quietly fail. Asking about ad tolerance after showing three creative examples will bias the answer. Keep attitude batteries short, randomise where the platform allows it, and pilot the questionnaire before the main fieldwork.

If you are choosing a fieldwork platform, the survey tools used for connected TV research in England covers panel sourcing, weighting and quality checks in more detail than a general guide can.

When are interviews worth the cost?

Interviews earn their place when you need to understand a decision, not measure its frequency. They are good for explaining why a campaign underperformed, how a media buyer evaluates streaming inventory, or what stops a viewer from finishing an ad.

They are poor for sizing. Ten conversations cannot tell you what share of a market behaves a certain way, and treating them as if they can is a common error.

The customer interview method for connected TV advertising sets out recruitment, question design and how to code responses so they feed a plan. Budget for analysis, not just fieldwork: an unanalysed transcript is an expensive recording.

How do you size a connected TV opportunity?

Sizing runs in three steps. First, define the population: households or individuals, UK-wide or a nation, and which services count. Second, estimate reachable share from panel or platform data. Third, apply a realistic frequency assumption to convert reach into impressions.

Be explicit about what you are excluding. Ad-free tiers, logged-out viewing and some app environments will not be addressable, and a size that ignores them will disappoint at buying stage.

Triangulate before committing. When panel data and platform data disagree, the gap usually sits in definitions. What counts as a view? Which devices are in scope, and are logged-out sessions counted? Resolve those differences before a number reaches a plan.

Where a number comes from a third party, name the source and its date in the same sentence. Where it does not, label it as an illustrative example. For instance, a plan assuming a household sees eight ads a week is an assumption you can test, not a finding.

How do you research the competition alongside the audience?

Audience research rarely stands alone. To judge whether a reach target is ambitious, you need to know what rivals are buying and where inventory is tight. That is a separate exercise with its own data sources, covered in the connected TV competitor research approach.

The two work best in sequence. Audience work tells you who is worth reaching; competitor work tells you what it will cost and how crowded the placement will be. Running them in the opposite order tends to produce a plan shaped by what is available rather than what is needed.

How do you prove the research paid off?

The IPA publishes an Effectiveness Databank of advertising case evidence, including video and digital channels. It is useful less for benchmarks and more for study design: it shows what a convincing proof of effectiveness looks like when a client asks for one.

The simplest credible test is a holdout. Split the audience into a group that can be served connected TV advertising and a matched group that cannot, then compare outcomes on a measure you chose before launch. Pre-registering the measure prevents the result being chosen after the fact.

If a holdout is impossible, use a stepped rollout across regions or time periods and be honest about the limits of the design.

What legal and ethical limits apply?

Audience research collects personal data, so the UK GDPR applies from the moment you recruit a participant. You need a lawful basis, a clear privacy notice, and a retention period you actually follow. Consent for research is not the same as consent for advertising, and using one for the other is a compliance risk.

Panel and platform data carries its own obligations. If you onboard first-party data for targeting, the permissions attached to that data travel with it. The UK rules and compliance guide for connected TV advertising sets out the consent, transparency and record-keeping points that apply to UK campaigns.

Special category data, children's data and inferred characteristics all need extra care. If your segment definition relies on any of them, get the legal position confirmed before fieldwork begins, not after.

Record keeping is part of the design, not an afterthought. Keep consent records for as long as you hold the data, and note which sample each finding came from so a later reviewer can trace it.

How should findings be presented to a buying team?

Present the decision, the evidence and the uncertainty, in that order. A buying team needs to know what to do on Monday, which numbers support it, and how confident they should be.

Agree the headline number before the meeting. One reach figure, one frequency assumption and one cost estimate give a buying team something to plan against. Everything else can live in an appendix that gets checked when someone questions the plan.

Keep one page per segment: who they are, where they can be reached, what they respond to, and what would change the plan. Attach the method note separately for anyone who wants to audit it.

Flag the gaps. If you could not measure co-viewing, say so. A stated gap is manageable; a hidden one becomes a post-campaign argument.

Common questions

How much does audience research for connected TV cost?

Costs vary with method rather than channel. A nationally weighted survey with a reputable panel typically runs into several thousand pounds, while a short programme of depth interviews can be delivered for a fraction of that. For example, a team commissioning a 1,000-response UK survey might budget £6,000 to £12,000 as an illustrative range, excluding analysis time.

Can I rely on platform-reported audience data alone?

No. Platform data is strong on delivery and weak on cross-platform comparison, because each platform defines its audience differently. Use it for in-flight optimisation and reconcile it against an independent baseline such as BARB before you use it in a plan.

How often should audience research be refreshed?

Annually for segmentation and attitudes, and more often for anything tied to platform mix, which can shift within a quarter. If a major service changes its ad tier or pricing, revisit the reachable population rather than waiting for the next annual cycle.

What sample size do I need for a UK survey?

For a national read, aim for at least 1,000 weighted respondents so subgroup estimates remain usable. If you need reliable reads on Scotland, Wales or Northern Ireland separately, size each nation on its own terms rather than relying on a UK-wide sample to carry them.