07
Automotive · MMM
Bayesian marketing mix modelling for car brands with Google Meridian
Cycle & Carriage (Kia, Mercedes-Benz) · Havas Media Singapore · 2025
- Model
- Bayesian MMM, Google Meridian
- Data
- ~900 daily points per brand
- Period
- Sep 2022 – Mar 2025
- Outcomes
- Walk-ins, test drives, sales
The challenge
Media spend swung heavily between channels and years. The client wanted to know which channels were actually bringing people into showrooms, and how to plan the next year's budget.
The approach
- Tested daily against weekly data; weekly gave the reliable fit
- Modelled TV, search, online video, display, social and radio against walk-ins and sales
- Used priors and credible intervals so the uncertainty was clear to the client
- Turned channel contribution, ROI and saturation into planning guidelines
My role
I set up the analysis in Meridian, ran and refined the models, evaluated the results with my manager, and wrote the report.
The outcome
The client got channel-by-channel guidance on where to add, hold or trim spend. That included which channels were near saturation, which were under-invested, and how to flight campaigns, with the model's limits stated plainly.