Geo Experiments and Causal Impact in Incrementality Testing
summary
Uses geo experiments on DMAs and a Python implementation of CausalImpact to estimate incremental conversions and incremental cost per conversion. Builds Bayesian structural time series models for both conversions and spend, runs a controlled pre / test / cool-down design, and returns a counterfactual versus observed lift with a 90% confidence interval. Reports a positive causal impact at HelloFresh and computes incremental CPA so stakeholders can compare lift versus test cost.
AI-written summary · about
chapters (8)
- 0:00Geo experiments intro
- 1:29Incrementality in marketing
- 6:54Traditional attribution flaws
- 9:43Bias solved with geo experiments
- 13:17Designing geo experiment steps
- 20:46Measuring lift with Causal Impact
- 25:20Higher confidence methods
- 27:37Q&A sample size modeling
related talks
36mMeasuring Media Impact- Practical Geo-Lift Incrementality Testing
32mCausal inference and scenario generation within Just Eat Takeaway.com
43mBeyond Just Prediction: Causal Thinking in Machine Learning
33mCausal Inference Framework for incrementality
39mCausal Effect Estimation in Practice: Lessons Learned from E-commerce & Banking
38mHow to Enhance Customer Targeting in Marketing
31mPython and the Holy Grail of Causal Inference
36mWhat-if- Causal reasoning meets Bayesian Inference