Media Mix Modeling with PyMC Marketing

In the previous chapter, we discussed how to assign credit to marketing channels, understand their value, and optimize our budget. However, a big caveat of the topics discussed in the previous chapter is the availability of granular-level data that shows which ads the user saw and clicked, and where they ended up converting.

This data is only available in digital channels, leaving out activities such as sponsorship or offline marketing, which, by definition, cannot be easily tracked. Adding to this issue, cookies are being slowly deprecated across major vendors, implying that even on digital channels, the availability of granular data is reducing. For example, many advertisers struggled when Apple released iOS 14.5 with enhanced privacy protections as this prevented companies from having access to the type of data needed for multi-touch attribution.

The solution to these measurement issues is a type of modeling called media mix modeling...