Marketing Mix Modeling (MMM) Making a Comeback
Content Provenance
Think about how you consumed media five years ago compared to today. You probably watched some cable TV, listened to the radio during your commute, and maybe scrolled through Facebook. Now? You might be streaming three different platforms, listening to podcasts, getting ads on TikTok, clicking through YouTube pre-rolls, and seeing sponsored posts on Instagram — sometimes all within the same afternoon.
Now imagine trying to figure out which of those touchpoints actually made you buy something.
That’s the challenge marketers are wrestling with every single day. The number of advertising channels has multiplied faster than most measurement tools can handle. And somewhere in that chaos, a classic approach that many people had written off as outdated is quietly becoming one of the most important tools in a marketer’s toolkit again — Marketing Mix Modeling, or MMM.
This isn’t just nostalgia. There are real, structural reasons why MMM is being taken seriously again, and understanding those reasons can help you think differently about how your organization measures marketing effectiveness.
Before we talk about the comeback, it helps to understand what MMM actually is and why it fell out of favor in the first place.
A Quick Explanation of MMM
Marketing Mix Modeling is a statistical analysis technique that looks at historical sales data alongside various marketing inputs — like TV spend, digital ads, promotions, and even external factors like weather or economic conditions — to figure out how much each of those inputs contributed to sales outcomes.
In simple terms, you gather a bunch of data over a long period of time, run it through statistical models, and the output tells you something like: “For every dollar you spent on paid search last quarter, you got back approximately $2.40 in revenue.” Or, “Your TV campaign drove about 12% of your total sales during the holiday season.”
It’s not perfect, but it gives marketers a way to look at the big picture without having to track individual customer journeys.
Why It Fell Out of Fashion
In the early 2010s, digital advertising came with a promise that felt almost too good to be true: perfect measurement. You could track exactly which ad a person clicked, follow them across the web, and tie their purchase directly back to a specific campaign. Compared to the slower, more approximate world of MMM, this felt revolutionary.
Marketers started pouring budgets into digital channels, partly because they could be measured so precisely. Last-click attribution models became the norm. If Facebook could show you exactly how many conversions came from a specific ad set, why would you bother with a statistical model that takes weeks to build and uses aggregate data?
So MMM got pushed to the sidelines. It became something big consumer packaged goods companies did once a year, almost as a formality. Many smaller brands and digital-first companies never even considered it.
The tools we use to measure marketing often shape the decisions we make — sometimes more than the actual marketing does. When precise digital tracking became available, it created a bias toward channels that could be easily tracked, regardless of whether those channels were actually the most effective
Here’s where things get interesting — and a little uncomfortable for anyone who has been relying purely on platform-native analytics.
More Channels Than Ever Before
The number of places a brand can advertise today is genuinely staggering. Consider just a partial list of where a mid-sized consumer brand might be running ads right now:
- Google Search
- Google Display Network
- YouTube
- TikTok
- Connected TV platforms (Hulu, Peacock, Paramount+, etc.)
- Streaming audio (Spotify, Pandora)
- Podcasts
- Programmatic display across thousands of publishers
- Retail media networks (Amazon, Walmart Connect, Target Roundel)
- Out-of-home digital boards
- Email marketing
- Influencer partnerships
Each of these channels has its own measurement system, its own attribution logic, and its own incentive to show you that it was responsible for your sales.
The Double-Counting Problem
This is one of those things that took the industry embarrassingly long to talk about openly, but it’s a real issue: when you add up the conversions that each platform claims credit for, the total is almost always higher — sometimes dramatically higher — than your actual sales.
Every platform wants to take credit. Facebook might say it drove 1,000 conversions. Google might say it drove 900. Your email platform might claim 400. But if you only made 1,200 total sales, something doesn’t add up.
This is because each platform is measuring things differently, often attributing a sale to itself if someone even saw an ad on that platform within a certain window before buying. When you’re running ads everywhere, almost every sale gets claimed by multiple channels simultaneously.
Privacy Changes Broke the Old Playbook
On top of channel fragmentation, the privacy landscape shifted underneath everyone’s feet.
Apple’s iOS 14.5 update in 2021 required apps to ask users for permission before tracking them. The majority of users said no. This significantly reduced the data flowing into Facebook’s ad measurement systems almost overnight. Advertisers who had built their entire measurement strategy around Meta’s pixel suddenly found their reported numbers were off — sometimes wildly so.
Then there’s the ongoing march toward a cookie-free web. Third-party cookies, which have been the backbone of cross-site tracking for decades, are being phased out across browsers. Google has been delaying its own deprecation timeline, but the direction of travel is clear: the ability to track individual users across the internet is going away.
When I talk to marketers who’ve been in the industry for more than a decade, there’s a common reaction to the current measurement environment: a mix of frustration and grudging respect for how complicated things have become. The tools that felt like they’d solved everything turned out to have real cracks in them.
Given everything above, it starts to make a lot of sense why MMM is back in the conversation. But it’s not just that the old alternatives are breaking down — MMM itself has genuinely gotten better.
MMM Doesn’t Rely on Individual User Tracking
This is probably the single biggest reason for the renewed interest. Marketing Mix Modeling works with aggregate data — total sales, total spend, macroeconomic variables, seasonal patterns. It doesn’t need to follow an individual person across devices or websites. It doesn’t care about cookies or app tracking permissions.
In a world where privacy regulations are tightening and individual-level tracking is getting harder, MMM is essentially immune to those problems. The input data is the kind of information most companies already have: their own sales figures and their own media spend numbers.
Modern MMM Is Faster and More Accessible
The old knock on MMM was that it was slow, expensive, and required a team of data scientists and a six-figure consulting contract. That criticism was largely fair for most of the 2000s and 2010s.
That’s changed. A few factors have contributed:
- Cloud computing has made it possible to run complex statistical models in hours rather than weeks
- Open-source tools like Meta’s Robyn and Google’s Meridian (both freely available) have brought serious MMM capabilities to teams that couldn’t afford proprietary solutions
- Bayesian modeling approaches have made it possible to update models more frequently with new data, rather than rebuilding from scratch each quarter
Organizations that used to need a large consulting engagement to run an MMM project can now run ongoing models with smaller internal teams.
It Captures What Digital Attribution Misses
Here’s something worth sitting with: a lot of marketing activity drives sales through channels that digital attribution models completely miss.
Someone sees a TV commercial for a brand they’ve never heard of. Three days later, they search for that brand on Google, click a paid search ad, and buy. The digital attribution model gives 100% credit to paid search. The TV campaign gets nothing.
Or consider influencer marketing. A person sees a sponsored post from a creator they follow on Instagram. They don’t click the link in bio. They close the app, think about it for two days, and then go directly to the brand’s website to buy. Again, digital attribution sees only a direct visit — no marketing credit given anywhere.
MMM, because it looks at the statistical relationship between total marketing activity and total sales over time, can pick up these indirect effects. If every time TV spend goes up, sales go up shortly after — even when there’s no direct click path — that relationship will show up in the model.
Investment in MMM Is Growing Rapidly
The market is voting with its dollars. Investment in cross-channel measurement approaches has been climbing steadily, with MMM solutions seeing particular interest from both enterprise brands and mid-market companies.
Major consulting firms have been expanding their marketing measurement practices. Technology companies like Google and Meta have released their own open-source MMM frameworks — which, yes, is partly self-interested, but also signals that they believe MMM is where serious measurement is heading. Several well-funded startups have emerged specifically to make MMM faster and more accessible.
It’s worth noting something slightly ironic: two of the biggest digital advertising platforms in the world have invested heavily in building MMM tools. That’s partly because they’ve recognized that their own in-platform attribution systems over-report effectiveness, and they’d rather help brands build more credible models than lose trust entirely.
It wouldn’t be fair to write about MMM’s comeback without acknowledging what it still can’t do well. No measurement approach is perfect, and MMM has real limitations worth understanding.
It Needs a Lot of Historical Data
MMM works best with at least two to three years of weekly data — ideally more. This means brands that are newer, that have significantly changed their business model, or that have gone through major shifts in their marketing mix may not have enough stable historical data to build reliable models.
If you launched a brand last year and want to understand your marketing effectiveness, MMM isn’t going to help you much yet.
It’s Not Great for Granular Decisions
MMM is fundamentally a macro-level tool. It can tell you whether TV or paid social delivered better returns over the last year. It’s not going to tell you whether your Tuesday email campaigns outperform your Thursday ones, or whether a specific creative variation drove better results than another.
For those kinds of granular, tactical decisions, other tools — A/B testing, incrementality testing, platform analytics — are still necessary and valuable. MMM works best as part of a broader measurement toolkit, not as a replacement for everything else.
Model Quality Depends on Data Quality
The old phrase “garbage in, garbage out” applies very directly to MMM. If your historical spend data is messy, if you’re missing data from certain channels, or if your sales figures include anomalies that aren’t accounted for (like a major supply chain disruption), the model outputs will be unreliable.
Building a good MMM requires careful data preparation, thoughtful variable selection, and honest calibration. It’s not as simple as feeding numbers into a tool and trusting the output.
It Still Involves Assumptions
All statistical models are built on assumptions. MMM is no different. Modelers have to make choices about which variables to include, how long a “carryover” effect to assume for advertising (the idea that advertising effects linger for some period after the ad runs), and how to handle interactions between channels. Different assumptions can produce meaningfully different results.
This doesn’t mean MMM is unreliable — it means it requires expertise and transparency about methodological choices.
The best measurement setups don’t pick a single approach and abandon everything else. They layer multiple methods that each have different strengths.
MMM + Incrementality Testing
MMM can tell you that paid social appeared to contribute significantly to sales over the past year. Incrementality testing — where you hold out a group of customers from seeing ads and compare their behavior to those who did see ads — can help validate whether that relationship is actually causal, and not just a correlation.
These two approaches work well together. MMM gives you the big picture over long time periods; incrementality testing gives you a more controlled, direct read on specific channels or campaigns.
MMM + Multi-Touch Attribution
Multi-touch attribution (MTA) tries to assign fractional credit to different touchpoints in a customer’s path to purchase using individual-level data. It’s been struggling with the same privacy issues that have affected all individual-level tracking, but where it works, it can complement MMM by providing more granular path-level insights.
Think of it this way: MMM tells you roughly how much of the pie belongs to each channel. MTA, where possible, helps you understand what’s happening within those channels in more detail.
The Role of First-Party Data
As third-party tracking disappears, brands’ own first-party data — email lists, customer purchase histories, CRM data — becomes increasingly valuable. First-party data can be used to improve the quality of MMM models, either directly as inputs or to calibrate and validate model outputs.
If you’re a marketer or marketing decision-maker reading this, there are some practical things worth thinking about.
Start Taking Data Collection More Seriously
MMM is only as good as the data that feeds it. If your organization has been sloppy about tracking marketing spend consistently across channels, or if your sales data lives in disconnected systems that don’t talk to each other, now is a good time to fix that.
Building a clean, consistent historical record of what you spent, where, and when — alongside your sales and revenue data — is foundational work that will pay off whether you run MMM models internally or work with outside partners.
Experiment With Open-Source Tools
If you have data scientists or analytically capable people on your team, tools like Meta’s Robyn and Google’s Meridian are free and worth exploring. They come with documentation, community support, and active development. They’re not plug-and-play for non-technical users, but they make serious MMM work accessible in a way it wasn’t five years ago.
Be Skeptical of Single-Source Measurement
If your current measurement approach is primarily “look at what each platform reports,” it’s worth stepping back and thinking about what that might be missing or distorting. Platform self-reported data has real value for optimization within a channel, but it’s a poor basis for cross-channel budget decisions.
Even running a simplified, rough version of MMM — or just triangulating your platform data against total revenue trends — can reveal surprising things about which channels are actually pulling their weight.
Think in Longer Time Horizons
One of the underappreciated things about MMM is that it forces you to think about marketing effects over longer time periods. Platform analytics tend to create a bias toward short-term, direct-response thinking. MMM, because it models relationships over months and years, is better at capturing how brand-building activity pays off gradually over time.
If there’s one mindset shift that MMM tends to produce in marketing teams that use it seriously, it’s this: a greater appreciation for the slow, compounding effects of consistent marketing activity, versus the temptation to chase whatever channel is easiest to track this week.
Marketing Mix Modeling never really went away — it just got overshadowed by tools that promised more precision than they could actually deliver. Now that the limitations of those tools are becoming harder to ignore, and as the privacy changes reshaping digital advertising continue to accelerate, MMM is finding a new audience.
It’s not a perfect solution. It requires good data, thoughtful methodology, and realistic expectations about what it can and can’t tell you. But as a way to understand the actual return on marketing investment across a fragmented, multi-channel media landscape — without depending on the ability to track individuals across the internet — it offers something genuinely valuable.
The fragmentation of ad channels and data sources isn’t going away. If anything, the media landscape is likely to get more complex before it gets simpler. Organizations that invest now in building solid cross-channel measurement capabilities — with MMM as a core component — will be in a much better position to make smart, defensible decisions about where their marketing dollars should go.
If you haven’t taken a serious look at MMM recently, now is a good time to start. The tools are better, the need is clearer, and the companies that figure this out will have a real advantage over those still reading dashboards that tell them only what they want to hear.