What Is Marketing Mix Modeling?
Marketing mix modeling uses aggregate sales and spend data to estimate what each channel contributes. How it works, what it's good for, and where traditional MMM falls short.
Marketing mix modeling (MMM) estimates how much each part of your marketing contributes to sales. It looks at your sales over time alongside your spend in every channel, and at the outside factors that also move sales, such as seasonality, pricing, promotions and the economy. Statistical analysis then separates the effect of each one.
Because it works from aggregate data rather than tracking individual people, MMM doesn't depend on cookies, clicks or device IDs. That's a large part of why it has come back to the center of marketing measurement.
How a marketing mix model works
A marketing mix model is built in three steps.
- Collect the data. Usually two or more years of weekly or daily sales, spend or impressions by channel, and the non-marketing factors that affect demand.
- Fit the model. A regression estimates the relationship between each input and sales, while holding the others constant. The result is a demand model: how much of your sales came from your base business, and how much from each channel.
- Use it to plan. With those relationships in hand, you can run "what if" scenarios. What happens to sales if the TV budget rises 10 percent? What if search spend moves into connected TV?
Two ideas make the model realistic:
- Adstock. Advertising keeps working after it runs, and its effect fades over time. The model credits a campaign for the sales it drives in the following weeks, not just the week it aired.
- Diminishing returns. Each additional dollar in a channel does less than the one before. This is what lets the model find the point where spend stops paying off.
What MMM is good for
MMM looks at your whole media portfolio at once, including channels that are hard to track, like TV, radio, podcasts and out-of-home. That makes it well suited to strategic planning: setting budgets, comparing channels on the same basis and showing finance where the money is working.
Where traditional MMM falls short
On the Data Gurus podcast, I described the problem with how marketing mix models were traditionally used:
“Prior to bottom-up multi-touch attribution marketers only had top-down marketing mix models. If you were a progressive company you would get a new MMM model done every year and that model would tell you to put ‘X’ percentage of your budget to digital, ‘X’ to TV, ‘X’ to direct mail and then it was up to each agency to determine how to allocate. With MMM, there are no clear-cut goals in terms of a ‘feedback loop’ if any of the media was actually hitting your numbers until you did another model next year.”
That gap matters more today than it did then. Classic MMM has struggled to keep up with digital marketing for several reasons:
- It's slow. A model rebuilt once a year can't guide decisions that change weekly.
- It's coarse. Traditional models measure channels like "digital" or "social" as one line, with little to say about specific platforms, campaigns or tactics.
- Digital data is fragmented. Each platform reports different metrics in different ways, which makes clean, consistent inputs harder to assemble.
- It assumes simple relationships. Older models often treated channels as independent and responses as straight lines, when channels actually influence each other and returns diminish.
- Correlation isn't causation. A model that only finds patterns in history can mistake coincidence for impact unless it is built and validated carefully.
What modern marketing mix modeling looks like
None of these are reasons to give up on MMM. They are reasons to modernize it. Today's models can be refreshed weekly or monthly, measure individual platforms and campaigns, account for how channels influence one another, and be validated against holdout data to show they predict sales they haven't seen.
That combination gives marketers what classic MMM couldn't: a privacy-safe view of what every channel contributes, updated often enough to act on.
If you're weighing approaches, our comparison of multi-touch attribution and marketing mix modeling covers where each fits.
Frequently asked questions
What data does a marketing mix model need?
Sales or conversions over time, spend or impressions by channel, and the outside factors that also move sales, such as seasonality, pricing, promotions, distribution and economic conditions. Two or more years of weekly or daily data is typical.
How is marketing mix modeling different from multi-touch attribution?
MMM works top-down from aggregate data and estimates each channel's contribution to total sales, including channels people don't click, like TV. Multi-touch attribution works bottom-up from individual user journeys and splits credit between the touchpoints it can track, which privacy changes have made much harder.
What is adstock?
Adstock is the idea that advertising keeps working after it runs, with its effect fading over time. MMM models this carryover so that a TV flight, for example, gets credit for sales in the weeks after it aired, not just the week it ran.
Does marketing mix modeling need cookies?
No. MMM uses aggregate data, not user-level tracking, which is why it has become central to measurement as cookies and device identifiers disappear.
How often should a marketing mix model be updated?
Traditional models were rebuilt once a year. That is too slow for most marketers today. Models refreshed weekly or monthly let you see whether changes are working while there is still time to act.