Is your marketing effective? Let Bayes decide!
summary
A Bayesian media-mix model for WeRoad quantifies delays, saturation curves, and a positive baseline, using five channel investments to predict weekly new customers and fixing collinearity with informative priors. Implementation uses Python with PyMC, Streamlit, polars and arviz; MCMC sampling gives full posteriors and trains in about 1.5 minutes on weekly data. The project finds a practical MAE benchmark (~1,000 on synthetic test) and delivers immediate impact, cutting spend and saving €50,000 in six weeks.
AI-written summary · about
chapters (24)
- 0:00Marketing attribution challenges
- 1:09Offline and online touchpoints
- 2:15From attribution to contribution
- 3:40Marketing media mix model
- 4:44When to invest with cross-correlation
- 6:28Investment timing for travel
- 7:30Correlation metrics for nonlinear effects
- 8:14Effect duration with adstock
- 9:21Investment impact with adstock
- 10:48Where and how much to invest
- 11:10Baseline and seasonal effects
- 13:01High error with high promotion
- 13:45Saturation with transformed variables
- 15:15Collinearity and coefficient instability
- 16:18From frequentist to Bayesian regression
- 17:43Bayesian models with PyMC
- 19:29Defining priors for constraints
- 20:33Priors for customer acquisition cost
- 21:56Simulation for budget allocation
- 22:40Four modeling lessons
- 24:06Q&A on computational cost
- 24:27Q&A on business impact
- 25:32Q&A on prior cheating
- 27:58Final Q&A and closing
related talks
31mMedia Mix Modeling:How to Measure the Effectiveness of Advertising
39mHow to Allocate Marketing Budget - Bayesian Approach to the Rescue
42mSolving Real-World Business Problems with Bayesian Modeling
29mMedia Mix Modelling
01h 26mProbabilistic Programming and Bayesian Computing with PyMC
41mPyMC-Marketing: Customer and Marketing Analytics the Easy Way
27mUsing Causal thinking to make Media Mix Modeling
27mEmpowering Business to make model-based marketing decisions