Geo-Lift Experiments
What you'll learn: How to use entire geographic regions as test and control units to measure the true incremental lift of marketing campaigns.
The Core Idea
A geo-lift experiment divides geographic regions (cities, states, DMAs) into test and control groups. You increase marketing spend or introduce a new campaign in the test regions while keeping control regions unchanged. By comparing outcomes between the two groups, you measure the incremental impact—the lift directly caused by your marketing, not just correlation.
Think of it like A/B testing, but instead of splitting users randomly, you split places. If you boost TV ads in Denver but not Phoenix, the difference in sales growth between these cities reveals TV's true effect.
Why Geography Works
Unlike digital holdout groups (where individuals are randomized), geo-experiments leverage natural separation: people in Boston don't see ads meant for Seattle. This makes it perfect for channels like TV, radio, billboards, or regional promotions where individual-level randomization is impossible.
The key assumption: parallel trends. Control regions should behave similarly to test regions in the absence of treatment. You validate this by checking pre-experiment trends.
The Process
- Match regions: Pair test and control geographies with similar historical sales, demographics, and seasonality
- Run the experiment: Increase marketing in test regions for a fixed period
- Measure lift: Compare actual outcomes in test regions vs predicted outcomes (based on control region trends)
- Calculate incrementality: Lift = (Test performance - Expected baseline) ÷ Expected baseline
Unlike attribution models that assign credit to touchpoints, geo-lift directly answers: "What would have happened without this campaign?"
Key Takeaway: Geo-lift experiments use geographic regions as treatment and control units to isolate the causal, incremental impact of marketing spend—offering ground truth for channels where individual-level holdouts aren't feasible.