Marketing Mix Modeling for Mid-Sized Companies
Marketing mix modeling estimates how much each channel contributes to sales, using aggregate data rather than individual tracking. It used to be a big-company tool, but free open-source models have changed that.
Key takeaways
- Marketing mix modeling uses statistics on weekly spend and sales to estimate each channel’s contribution.
- Google made its open-source Meridian model available to all marketers in January 2025.
- MMM needs about two years of consistent data and some variation in spend to work well.
Digital attribution is getting harder, because privacy rules, browser restrictions and long buying cycles leave large gaps in click-based tracking. As a result, many companies make budget decisions from reports that miss much of what drives sales.
Marketing mix modeling, often shortened to MMM, takes a different approach.
What marketing mix modeling is
MMM is a statistical method that looks at aggregate data over time: spend by channel, sales, prices, seasonality and outside factors such as economic conditions. It estimates how much each input contributed to sales, without tracking any individual person.
Large consumer brands have used it for decades, since it does not depend on cookies or clicks. The method suits brand advertising, offline media and long sales cycles, which click-based attribution handles poorly.
Why MMM is now within reach
Until recently, MMM required expensive specialist firms. That has changed with open-source tools. Meta released an open-source package called Robyn, and Google made its own model, Meridian, available to all marketers in January 2025.
These tools are free, although they still require someone comfortable with data and statistics. Many companies run them with a part-time analyst or a measurement partner.
When marketing mix modeling is worth it
- You spend across several channels, including some that clicks cannot measure, such as radio, events or out-of-home.
- You have enough history. About two years of weekly data is a common starting point.
- Your spend varies. The model learns from changes, so a perfectly flat budget gives it little to work with.
- The budget is large enough that a better allocation would pay for the analysis.
However, if you spend on only one or two digital channels, simpler tests may serve you better. A regional holdout test, for example, can answer one specific question cheaply.
What data you need
- Weekly spend and impressions by channel.
- Weekly sales or leads, ideally by region.
- Prices, promotions and product launches.
- Seasonal patterns and major outside events.
Most of the effort goes into collecting and cleaning this data. So start organizing it now, even if the model comes later.
Using the results
Treat the model as one input rather than a final verdict. Compare its findings with holdout tests and with what customers tell you. When all three agree, you can move budget with confidence.
MMM often credits brand advertising more than click-based reports do, which supports the case we make in brand building vs sales activation. To see whether marketing mix modeling suits your business, book a free diagnosis.
Related guides
Find out where your marketing is bleeding.
Brand Doctor is based in Vancouver and offers a free marketing diagnosis to businesses across Canada, in person or remotely. You leave knowing what’s holding growth back and what to fix first.
