Decoding Data & AI: Why is everyone talking about MMMs (Marketing/Media Mix Models)?

What we need here is what economists call a model - a description of how people will react to different marketing choices, but slightly simplified. Much like an architectural model will show me the general outline of the building but not include every bathroom installation, my Marketing Mix Model will show me in general terms how quickly the credit cards are coming out of the wallets when I change my marketing mix.

In the end, this is a mathematical formula that will predict my revenue based on the money spent on different media and marketing channels. Since this model can get a bit complicated, like in most complex data analysis, it is useful to let AI do the heavy lifting. This is a typical machine learning problem, where the idea is to feed lots of data to a self-learning algorithm, which will construct our model by finding patterns in the data that a human analyst won't be able to see.

Luckily, both Google (LightweightMMM) and Meta (RobynMMM) believe this problem is important enough that they have recently created Open Source versions of this self-learning algorithm that everyone can download, use, and adapt to their own purposes.

You "only" have to bring your data to the table. However, as always, more and higher quality data is a lot better than less. Data quality and proper data management are absolutely essential to leverage the power of such AI-driven solutions. The more data we can throw into those models - timing, size, and location of your campaigns, as well as revenue data on the most granular level - are crucial. The cherry on top would be any data about competitor promotions you may have collected.

Feeding that input data into a marketing mix model allows the latter to address the million-dollar question of marketing: how should ad spend be distributed across the different marketing channels, in order to maximize conversions and therefore revenue?
 

Let's take a look at the output from Robyn - the MMM package by Meta: It tells you both what you are currently spending in your marketing mix and how much the different channels contribute to your ROAS - already quite helpful. This is shown on the left. However, since this is a full model, it will also show you how much more you could make with the same spending level, just by optimizing your channel mix.
 

What you can see in this example is that just by optimizing allocation, the client could generate an additional 560,000 conversions, an increase of 22.6% while keeping total spending the same. Robyn suggests to substantially reduce investment in TV and Out-of-home (billboards) advertising, and increase search and print media.

You can see that this is a powerful way to supercharge your marketing decisions by moving beyond simple measures to really leverage the value of your data.

Want to learn more about OMMAX's expertise in data & AI? Get in touch with our experts through the form below and sign up for our Decoding Tech, Data & AI series!