Radio advertising attribution methods without the guesswork

What to take away

Why does attribution feel harder on radio than on search?

Search hands you a click. Radio hands you a listener who may act later, in a shop, on a phone, or not at all. That is why radio advertising attribution methods vary so much in cost and confidence.

A station can tell you when a spot ran and roughly how many people it reached. For UK listening estimates, buyers can also refer to RAJAR (Radio Joint Audience Research), the UK's radio audience measurement service; its audience data provides reach context, not proof that a campaign caused an outcome.

Data volumes keep rising across digital channels, making modelled and panel approaches more workable than a decade ago, as the worldwide data growth figures tracked by Statista show. More data does not mean more certainty.

For how these methods sit alongside reach, frequency and response reporting, the radio advertising measurement and reporting guide sets out the framework.

Which method fits a direct response campaign?

Promo codes and tracked phone numbers are the workhorses. A code read out on air is countable, and a number routed to a unique line is countable. Both are cheap to set up and explain to a finance team.

For call tracking, buyers can consider UK providers such as Infinity and Mediahawk; check whether the service can assign numbers and report calls by station or region.

Coverage is the catch. Only listeners who use the code or number appear in your data. For example, a campaign spending £8,000 over four weeks might log 300 code redemptions and still miss every listener who drove to a branch.

Codes and numbers also age badly when the same creative runs across several stations. Give each station its own code, or you lose the ability to compare performance.

Add a simple control: run the code in one region and not another, then compare. That gives a rough counterfactual without paying for a panel.

When should you use panel or modelled attribution?

Use them when the campaign has no direct response mechanic, or when awareness is the objective. Panel data asks a sample of listeners what they heard and what they did. Modelled attribution estimates sales from exposure patterns.

Both produce estimates, not counts, which is the nature of the method rather than a flaw. Report it as a range with assumptions stated.

Buyers sometimes read panel results as a ranking of stations. It is safer to treat them as an indication of recall within the survey window.

Panel work carries a cost per wave, so it suits larger campaigns. A small local buy rarely clears that bar.

How do you score the methods against each other?

Use the rubric below. Score each method from one to five on every row, then weight rows to match the objective.

These are starting-point judgements, not market benchmarks. A score of five means a more favourable result: lower cost, stronger attribution confidence, faster reporting or broader coverage, depending on the column. Check actual supplier costs and adjust the scores for the campaign design.

The ratings are comparative planning scores, not guarantees. Codes and tracked numbers are quick and relatively inexpensive but record only people who use them; panel and modelled approaches can cover brand outcomes but produce estimates. A regional holdout offers a counterfactual, although the result depends on how well the regions can be compared.

Weight direct response campaigns towards cost, confidence and speed, and brand campaigns towards coverage.

What legal duties sit behind the data?

Audience data used to target or measure a radio campaign falls under UK data protection law. The UK GDPR and Data Protection Act 2018 set the framework; guidance from the Information Commissioner's Office (ICO) explains lawful basis, retention and individual rights.

Retention and transparency obligations apply to measurement data as well as targeting data. Consent is required where it is the appropriate lawful basis or another applicable rule requires it. If a station shares listener lists with you, agree in writing what happens to them once the campaign ends.

There is a competition dimension too. The Competition and Markets Authority oversees competition in advertising markets, which matters when comparing vendors that share data or pricing.

Errors are common. The radio advertising measurement mistakes in England piece covers the mistakes that show up most in post-campaign reviews.

Common questions

Can I use more than one attribution method at once?

Yes, and most campaigns should. A promo code covers direct response while a panel wave covers brand lift.

Is modelled attribution accurate enough to report to a board?

Accurate enough as a range with stated assumptions. Presenting a modelled figure as a precise count is where campaigns get into trouble.

How long should a radio attribution test run?

Long enough to gather a usable sample, usually at least two to four weeks for a local campaign. Shorter tests tend to produce noise that looks like a result.