The most reliable way to measure TV incrementality is a geo holdout test: run your TV campaign in some regions, suppress it in matched regions, and compare business outcomes. This gives you an estimate of the revenue your TV actually caused.
TV advertising is one of the hardest channels to measure. Unlike digital ads, you cannot suppress TV from individual users. You cannot retarget people who saw an ad. And there is no click to attribute a conversion to. But you can still run a proper incrementality test.
Why TV measurement is challenging
Traditional TV buying is done at a regional or national level, which makes audience-level holdouts impossible. A viewer watching ITV in Manchester cannot be randomly excluded from a national TV buy. Geography is the only lever available to create a controlled comparison.
Media mix models (MMMs) have historically been the primary tool for TV measurement, but they rely on statistical assumptions that are hard to validate without an experimental benchmark. An incrementality test gives you a real-world check on what the model says.
Designing a TV geo holdout
To run a TV geo holdout, you need to work with your media agency or TV broadcaster to suppress your campaign in a set of holdout regions. This typically means buying TV in a subset of available regions and leaving others dark. Not all broadcasters or buying arrangements support this, so plan early.
- Select holdout regions that match your test regions on historical sales and demographics
- Suppress TV in holdout regions for the full test period
- Keep all other channels, such as digital, OOH, and radio, running consistently in both groups
- Measure revenue, website visits, or footfall in both groups across the test period
Connected TV offers new options
Connected TV (CTV), which includes streaming services and smart TV platforms, offers audience-level targeting and suppression. This makes traditional audience holdouts possible for the first time in a TV-like environment.
CTV incrementality tests can be designed the same way as digital audience holdouts: randomly assign a portion of your audience to a control group and compare conversion rates. The same statistical requirements apply: adequate conversion volume, sufficient test duration, and a pre-test balance check.
Be careful not to mix linear TV and CTV results without accounting for reach overlap. A household watching both a linear TV broadcast and a CTV stream may be counted in both channels, which inflates apparent reach and makes the holdout comparison unreliable.
How long does a TV incrementality test take?
TV effects tend to build over time because brand advertising works cumulatively. A minimum of four weeks is recommended for linear TV tests. If your TV campaign runs intermittently, extend the test to capture at least two full flight periods and the decay between them.
Allow for a post-test observation period to capture lagged purchase behaviour from people who saw the ads but did not convert immediately.
Using TV results to calibrate your MMM
Once you have an experimental lift estimate from a TV geo holdout, you can use it to calibrate your marketing mix model. If the model's TV coefficient is significantly higher or lower than the experimental result, you have evidence to adjust the model parameters and produce more accurate budget recommendations.
Can I run a TV incrementality test without an agency?
For linear TV, you almost always need agency support to suppress advertising in specific regions, as this requires changes to the media buy. For CTV, some platforms offer self-serve testing tools that allow you to set up holdouts without agency involvement.
What outcome metric should I use for a TV test?
Use business outcomes rather than media metrics. Website visits from TV regions, direct sales, or store footfall are more meaningful than reach or GRP (gross rating points). Ideally, use the same conversion metric you would use for digital incrementality tests to enable cross-channel comparison.
Is it possible to measure TV incrementality with small budgets?
It becomes progressively harder as TV budgets shrink. Regional TV buys below a certain spend level do not deliver enough GRPs in each region to create a meaningful exposure difference between test and holdout. If your TV budget is small, consider pooling several regions into test and holdout groups rather than splitting at a very granular level.