Measuring Across TV, Radio, OOH and Digital: Attribution and Marketing Mix Modelling

When you run radio, billboards, TV, search and social at the same time, every channel claims credit, and the totals add up to more than your actual sales. Multi-channel attribution is the practice of working out what each channel really contributed. Here's how the main approaches work and a practical way to use them at any budget.

This guide is part of our guide to measuring billboard, TV and radio advertising.

The problem with platform reporting

Each digital platform reports conversions it touched, using its own rules. Offline channels don't report conversions at all. So:

• Search and retargeting look like heroes, because they're often the last click.
• Radio, TV and OOH look invisible, even when they drove the search.

We explain the dynamic in radio vs. digital ads.

Three approaches to attribution

Approach	How it works	Strengths	Limits
Multi-touch attribution	Splits credit across tracked digital touchpoints	Granular, fast	Can't see offline ads; weakened by privacy changes
Incrementality testing	Compares exposed vs. unexposed groups or markets	Measures cause, not correlation	One question at a time; needs planning
Marketing mix modelling (MMM)	Statistical model of spend, sales and outside factors over time	Covers all channels, including offline	Needs lots of historical data; less granular

Incrementality tests: the practical starting point

The simplest, most convincing approach for most businesses:

1. Pick comparable markets (similar size, demographics, sales history).
2. Run the channel in test markets only, keeping everything else the same.
3. Compare the change in sales, leads or branded search between test and control.

The difference is the channel's incremental effect. Repeat for each major channel over time. You can also run "holdout" tests in digital by excluding a random share of your audience from ads.

Marketing mix modelling

MMM uses historical weekly data (spend by channel, sales, price changes, promotions, seasonality, weather, competitor activity) to estimate each channel's contribution and diminishing returns. It answers questions like "what happens if we move 20% of search budget into radio?"

When it's worth it: you spend across several channels, have a couple of years of reasonably clean weekly data, and budgets large enough that better allocation pays for the analysis. Open-source MMM tools have made it more accessible, but it still needs someone who understands the statistics.

A practical approach for growing businesses

1. Track everything you can directly: vanity URLs, codes, call tracking. See QR codes, promo codes and vanity URLs.
2. Watch blended metrics: total new customers and blended customer acquisition cost across all spend.
3. Monitor branded search as the bridge between offline and online.
4. Run one incrementality test per quarter on your biggest uncertainty.
5. Ask customers: "How did you hear about us?" often reveals channels your analytics miss. See using surveys to measure offline ads.
6. Graduate to MMM when spend and data justify it.

Designing a fair market test

• Choose test and control markets with similar size, demographics and recent sales trends.
• Keep other marketing identical in both during the test.
• Run long enough to capture the sales cycle, usually several weeks or more.
• Measure the same outcome in both (sales, leads, new customers, branded search).
• Compare the change in each market, not the raw totals.

Metrics that keep everyone honest

• Blended CAC: total marketing spend ÷ total new customers
• Marketing efficiency ratio: total revenue ÷ total marketing spend
• Incremental lift from tests, per channel

If the channel dashboards say one thing and blended numbers say another, trust the blended numbers. For a full list of what to report, see advertising KPIs that matter.