What is Marketing Mix Modeling?
Media Mix Modeling or, interchangeably, Marketing Mix Modeling is a way to quantify the contribution of each marketing channel and known external factors to revenue or customer acquisition.
Back before the days of Performance Marketing MMMs were a way that big brands measured the impact of their media budgets including TV, print, billboards, radio, events, and seasonality.
I sat with the MMM team at Scotts Miracle Grow. Sometimes the data would be so granular that it was measured in basis points. The marketing department would use that data to justify budgets for the coming season. It was a year long cycle to make the report and to use it’s insights.
Now with most marketers diversifying away from last click and into channels with no clicks like podcasts or CTV MMMs are valuable again.
Open source Mixed Media Modeling with Platform Data
After Apple’s ITP privacy initiatives reduced tracking visibility for digital ads Google and Meta both started working on MMM projects to help advertisers justify spend.
Google’s Meridian project is launching today to a wider audience of users.
Meta’s Robyn project was launched in 2021 and continued to mature and gain wider acceptance.
Both projects have the ability to download platform data. Meta has made MMM data exports available to be used with any MMM. Google is set to release a treasure trove of data today known as Query Search Volume (QSV).
Historic search data for organic and paid
Via an API call to the MMM data platform it will be possible to pull in data for:
- Google Query Volume (GQV): The frequency of specific search queries over time.
- Paid Search Data: The performance of paid search campaigns.
- Reach and Frequency Data: How many unique users were exposed to ads and how often.
The big benefit is the ability to control for confounding variables, basically making sure that it really is the advertising driving and not some external factors.
This is a huge give from Google. A giant leap in the right direction to gain some goodwill with advertisers.