Measuring Media Impact- Practical Geo-Lift Incrementality Testing
summary
Explains how running geo-lift incrementality tests on DMAs or similar geos gives a privacy-agnostic causal estimate, using an AA check, market stratification, power analysis and methods like difference‑in‑difference, synthetic controls and BSTS with statsmodels-style regressions; requires substantial daily spend, typical runs of 4–6 weeks plus a 2‑week cooldown, and often finds platform claims (example: platform 6x vs measured 2.5x) much higher than true incremental ROAS.
AI-written summary · about
chapters (17)
- 0:00Measuring media impact
- 1:05Incrementality testing importance
- 3:35Geo-lift privacy solution
- 5:23Geo-lift limitations
- 7:52Geo-lift test roadmap
- 10:22Scale vs go-dark tests
- 12:28Market selection for tests
- 14:34Test duration and power
- 16:20Difference-in-difference method
- 17:24Synthetic controls for lift
- 18:29Bayesian time series approach
- 20:57Cool-down and carryover effects
- 22:41Results and true ROAS
- 23:24YouTube campaign case study
- 26:59Pitfalls and contamination risks
- 28:45Short-term vs long-term gains
- 30:09Q&A
related talks
35mGeo Experiments and Causal Impact in Incrementality Testing
46mFrom correlation to causation: measuring ad effectiveness at scale
23mThe unreasonable effectiveness of data for better monetisation
31mMedia Mix Modeling:How to Measure the Effectiveness of Advertising
38mHow to Enhance Customer Targeting in Marketing
30mPredictive Analytics in Social Media and Online Display Advertising
19mRobert Moakler, Data Science Intern, Integral Ad Science @ MLconf SEA
13mOptimizing Subscription Services with Uplift Modeling