Selecting a model
There are some considerations to take when selecting a final model:
- Experimental calibration: Integrating experimental results into the MMM is the gold standard approach
- Business insights: Evaluate the model by investigating if the outcomes match your business context
- ROAS convergence: Looking at the distribution of ROAS over multiple iterations and how it evolves can be a good indicator of higher confidence in results if the distributions are peaky
- Statistical parameters: Looking at the model fit statistics, such as R2, RMSE, AIC, BIC, and so on, can be a good indicator of higher confidence in results if the values are good
After creating the model, you need to see how accurate it is. In MMM, the gold standard for this is via experimenting and calibrating the model.
Experimenting and calibrating
Using experiments to calibrate the model is the best way to achieve accurate results. Several methodologies can be used:
- People...
