Tidy tools and extensions for MMM (Media/Marketing Mix Modeling)
Motivation
This package has two main objectives
- Extend the tidymodels modeling framework with transformation steps required for MMM
- Adstock & Carryover Transformations
- Saturation
- Provide pre/post modeling tools for MMM projects.
- Model Evaluation
- Model Tuning
- Media/Marketing Optimization
- Visualizations
Why Tidymodels and Tidymmm
Tidymmm takes inspiration from
-
A few great projects that have been developed in the recent past (see related work below).
Modeling methodology and design from these projects (and research behind them) are used as inspiration, however there is no objective of re-implementing thier exact functionality.
The biggest difference is that tidymmm aims to provide modelers with felexible tools (a la tidymodels philosophy) and not lock users into a specific method. The disadvantage is that tidymmm is less automated out the box, but on the other hand has all the flexibility that comes with tidymodels.
-
The Tidymodels infrastructure in the form of packages like
parsnip(modeling),recipes(data prepossessing),tune(hyper-parameter optimization),workflows(combining everything together).- The aim of tidymmm is to expend and tap into these tools + provide MMM specific tools implemented the tidy-way.
