Glossary · Marketing Engineering

Geo-Lift Testing

also: geographic split test · geo experiment · market holdout

Definition

Geo-lift testing is an experimental design that randomizes marketing treatment across geographic markets (DMAs, metros, countries) to measure incremental impact when user-level randomization is infeasible. Synthetic control and open-source libraries like GeoLift construct matched control markets from pre-period covariates.

User-level A/B tests fail for TV, radio, OOH, and brand campaigns because exposure is unobservable per user. Geo-lift solves this by treating markets as units: run the campaign in treatment markets, withhold in synthetic or matched controls. Meta's open-source GeoLift library automates market matching via time-series regression; Google's CausalImpact adds Bayesian uncertainty. Minimum detectable effect depends on market variance and campaign size — typically 8–15% lift is detectable at large weekly spend with 4 treatment and 20 control markets.

Essays on this concept