How to Test Incrementality in Pharma CTV
A feasibility matrix for household holdouts, geographic tests, and matched-control designs in pharma CTV, with the assumptions each relies on.
The short answer
CTV is one of the fastest-growing channels in pharma DTC, and one of the harder ones to measure causally. Ads are delivered to households, not individuals. Identity links between TV devices and prescription data are imperfect. And many CTV buys run through supply paths where the advertiser has limited control. Still, incrementality testing in CTV is possible. The key is choosing a design that fits the constraints.
Three designs
Household holdout
Households in the target audience are randomly split. Test households are eligible for ads; control households are suppressed.
Needs: a partner that can reliably suppress specific households, and a measurement link from household to prescription outcomes.
Watch for: control households seeing ads through other partners or devices, which biases lift down.
Geographic test
Markets or regions are assigned to test or control. CTV runs in test markets only.
Needs: enough markets to balance, and stable other media across regions.
Watch for: regional differences and events unrelated to media. Fewer units mean only larger effects are detectable. See HCP-level vs. geographic designs.
Matched control
Exposed households are compared with similar unexposed households after the fact.
Needs: good data on household characteristics that predict outcomes.
Watch for: unmeasured differences between exposed and unexposed households. This is the weakest design for causal claims.
Feasibility matrix
| Condition | Household holdout | Geo test | Matched control |
|---|---|---|---|
| Partner can suppress households | Required | Not needed | Not needed |
| Large audience | Helps | Helps | Helps |
| Many CTV partners in plan | Hard to coordinate | Easier | Possible |
| Other media varies by region | Fine | Problematic | Fine |
| Need strongest causal evidence | Best | Good | Weakest |
Before you start
- Define the outcome and confirm it can be measured at the household or regional level.
- Estimate the detectable effect. If the design can only detect very large effects, a null result will be uninformative. NIST's handbook on confidence intervals explains the relationship between sample and precision.
- Freeze other activity in control units as much as possible.
- Agree on the decision rule before seeing results.
Supply quality affects results
If CTV impressions are going to low-quality or misrepresented inventory, a test will measure that inventory, not CTV in general. Audit the supply chain before or during the test.
Frequency and reach context
A lift result depends on the reach and frequency delivered. Report them alongside the lift estimate so others can judge whether results would transfer to a different plan. See CTV frequency management.
Practical takeaway
Choose the strongest design your partners and audience size allow. Document the design's main assumption, estimate what effect it can detect, and publish the reach and frequency delivered with the result.
Sources
- NIST/SEMATECH, Confidence Intervals
- American Statistical Association, Statement on p-Values
- Media Rating Council, Standards and Guidelines
External guidance and platform documentation change. Links were current at publication; check them again before relying on them for a decision.
Editorial note. Analysis and frameworks are the author's own and do not represent Acxiom or any current or former employer, client, or named platform. Examples labeled hypothetical or illustrative are not results from real campaigns. Nothing here is legal, regulatory, or medical advice.
New pharma programmatic breakdowns, occasionally
One email when I publish something worth reading. Benchmarks, measurement teardowns, and case studies with the caveats attached. No cadence promises, no reselling your address.
Unsubscribe any time. See the privacy policy.
Working through this decision on a real plan?
I work on health and pharma data, identity, and activation, after five years running HCP and DTC programmatic agency-side. Happy to talk through how this applies to your situation.