ASK US ANYTHING
Can You Prove CTV Works Regionally?
We Just Did.
CTV often gets written off as a super expensive channel with limited reach, great for visibility, but fuzzy when it comes to impact. We wanted to challenge that.
So we ran a controlled CTV/ATV campaign for a client of ours in Bayern, while keeping Baden-Württemberg untouched as a comparison. The aim: prove that regional campaigns can drive measurable demand.
Our Approach (quite intuitive if I must say)
We combined old-school planning with modern analysis:
- Google search trends to track real-world interest in the brand advertised on CTV
- Bayesian uplift modeling to compare against a control region
- AI-powered counterfactuals to test multiple interpretations
- And human judgement to keep things grounded (that’s us!)
What we found
- During the campaign, brand interest in Bayern rose +58.9% vs. just +13.7% in the control region
- After the campaign, interest held at +70%; proving the effect wasn’t short-lived
- Even when using a conservative model, we saw a +50-60% uplift
- All methods agreed: CTV had a measurable impact
But we didn’t stop there
We wanted to dig deeper. Because, you know; with AI models, the first answer isn’t always the final word. 🙂
So we built a counterfactual model using only non-neighboring regions to remove potential spillover effects. This method showed an even higher uplift—up to +270%. But it came with tradeoffs: the statistical fit was weaker, and the model relied heavily on regions that weren’t well correlated with Bayern’s pre-campaign volumes.
So what did we learn?
- Removing spillover effects can reveal hidden impact, but introduces greater uncertainty
- AI-powered methods are powerful, but need human supervision to stay grounded
- The real impact likely sits between the conservative +50% and the upper bound of +270%
What this means
- CTV can be a both a performance and branding channel
- Search data, statistical models, and AI analysis can work together, not in silos – just have to know what to look for and how to connect the dots
- Measurement doesn’t require black-box models. It just needs clarity, structure, and the right signals
Bottom line: If you want to prove that your media works or not, start local. Pair smart planning with search data and with the available AI-powered analysis, and you’ll get more than a gut feeling. You’ll get proof.
Bonus Bottom line : When AI gives you an answer—ask it again.
