Marketing mix modeling (MMM)

Marketing mix modeling takes your sales week by week and puts them next to what you spent in each channel over the same period. The model then works out how much of those sales would have come anyway and how much each channel added. The method dates from the 1960s and was long used mostly by companies like Unilever and P&G, with their own analysts and large budgets.

What has made MMM relevant again is that tracking individual clicks has got worse. Cookies are disappearing, apps block tracking and the platforms report their own numbers. MMM needs none of them. Weekly spend and sales are enough, ideally two years back. Google's Meridian and Meta's Robyn are free and open source.

The weakness is that the model only sees what has varied. If you spent exactly the same on Meta every week for two years, it cannot tell what Meta does. An online shop in Borås with a steady budget gets a less certain answer than a company that has paused and tested. Start by collecting the data. The model is the easy part.