Marketing Mix Modeling: The Ghost of Measurement Past Returns—Again

You know it, I know it, and anyone who has spent more than five minutes in marketing knows it—measurement is a mess. It's always been a mess. And the only thing more persistent than bad attribution models is the industry's desperate need to convince itself that this time, we’ve finally figured it out.

And so, once again, like some washed-up rock band clinging to a nostalgia tour, Marketing Mix Modeling (MMM) is back. And this time, it’s bringing AI, open-source tools, and promises that sound suspiciously like the last three times it was hot. Only now, it has some big tech overlords slapping their names on it, so it must be legit, right?

MMM: The Lazarus of Measurement

For the uninitiated (or those who haven’t been forced to sit through a three-hour attribution workshop while silently Googling “how to fake enthusiasm in meetings”), Marketing Mix Modeling (MMM) is an old-school statistical approach that, in theory, tells brands where their ad dollars are actually making an impact. In reality? It’s the corporate version of a fortune teller reading tea leaves—except the tea leaves are ad spend reports, and the fortune teller is a consultant billing you by the hour.

Let’s break it down. MMM works by taking a pile of historical data—sales, media spend, weather patterns, inflation, whether a Kardashian launched a new skincare line that week, and whether Mercury was in retrograde—and attempts to connect those dots into something resembling logic. The goal? To tell you what parts of your marketing budget actually drive revenue and which ones are just paying for your agency’s fancy dinners.

Sounds useful, right? Except there’s a small problem: it’s all in the past. MMM is essentially the rearview mirror of marketing—it tells you what might have worked months ago but gives you absolutely no idea what you should do right now. It’s like driving a car using only your rearview mirror while flooring the gas pedal, hoping you don’t hit a tree. It doesn’t update in real time. It doesn’t help you optimize spend on the fly. And it definitely doesn’t tell you if that “game-changing” TikTok campaign your CMO keeps hyping up actually did anything beyond making an intern go viral.

So why, exactly, is MMM suddenly making yet another comeback in 2025, as if it’s the marketing version of bell-bottom jeans? Two words: Google and Meta. And let’s be real—when the two biggest ad platforms on the planet start aggressively pushing a new measurement framework, you can bet it’s not out of the goodness of their data-hoarding hearts. No, this is about making sure that when you run your next MMM report, it magically shows that you should be spending even more on—you guessed it—Google and Meta.

MMM’s comeback isn’t some organic return to data-driven wisdom. It’s a panic-fueled response to the slow collapse of cookie-based tracking and the increasing difficulty of measuring digital ads in a privacy-first world. With Apple nuking IDFA, Google dragging its feet on third-party cookie deprecation, and regulators circling the ad industry like sharks sensing blood, Big Tech needs a new way to justify ad spend. And what better way than a complex, opaque statistical model that conveniently reinforces the importance of their own platforms?

But sure, let’s all pretend this is about “advertiser empowerment” and not just another chapter in the long, ridiculous saga of ad platforms making the rules while marketers scramble to keep up.

Big Tech Wants You to Love MMM (For a Reason)

If you think Google and Meta are pushing MMM out of some deep commitment to helping brands make better decisions, let me introduce you to a bridge I’d like to sell you.

Google has open-sourced Meridian, its MMM tool, while Meta has Robin—because nothing says transparency like tech giants offering free tools that just so happen to justify more ad spend on their own platforms. But why are they so into MMM all of a sudden?

A few theories:

1. They need an alternative to Multi-Touch Attribution (MTA), which has been on life support ever since Apple nuked IDFA and Chrome keeps waffling on cookies. If advertisers can’t rely on last-click data anymore, then MMM, which conveniently favors upper-funnel channels like YouTube and Instagram, starts looking a whole lot more appealing.

2. It helps them “prove” they deserve more of your ad budget. Traditional digital attribution often undercounts view-based platforms like YouTube or Facebook’s feed ads. MMM, on the other hand, is perfectly fine making assumptions based on high-level trends rather than actual user behavior.

3. They need a hedge. With cookies and user-level tracking getting more complicated than a congressional hearing on TikTok, Google and Meta need to show regulators and advertisers that there are still ways to measure performance—even if those ways are conveniently skewed in their favor.

The Problem with MMM: It’s a Great Theory, But a Nightmare in Practice

Here’s the thing—MMM sounds great in PowerPoint decks, but it falls apart when you actually try to use it for, you know, real marketing decisions.

For starters, it’s slow. Really slow. Like “waiting-for-your-government-issued-rebate-check” slow. Unlike MTA, which at least gave you a semi-real-time look at performance, MMM is built on historical data, meaning you get answers after the fact. And if something changes (spoiler: something always changes), your model is already outdated by the time you actually try to use it.

Then there’s the issue of triangulation. You’ll hear this word thrown around a lot by consultants and execs trying to sound smart, but here’s what it actually means: “We have no idea what’s right, so we’re just going to take a bunch of different models and average them together until they tell us something we like.”

It’s like trying to guess the weight of a cow by asking 30 random people at a state fair. Someone’s gonna be wildly off, someone else might be close, but ultimately, all you’re doing is coming up with a number that feels right.

And don’t even get me started on the operational nightmare of actually implementing MMM. A lot of brands sign up for it because, well, it sounds like something serious companies should do. Then they get their first report and realize they have no idea how to actually act on the results.

Advertisers Are Running Blind, But MMM Won’t Save Them

If there’s one thing marketers love more than a shiny new attribution model, it’s completely ignoring how to use it. We’re talking about an industry that jumps from one buzzword to the next like an influencer chasing the algorithm—first it was last-click, then multi-touch, then incrementality, now it’s MMM—and at no point does anyone stop to ask, Wait, do we even know what we’re trying to solve?

Because let’s be clear: most brands don’t have a measurement problem. They have an “actually making decisions” problem.

That’s why every quarter, CMOs walk into executive meetings armed with deck after deck of MMM reports, regression analyses, and charts with more lines than a Wall Street cocaine party—only to turn around and dump another million into the same media mix because, well, that’s what they did last year. If MMM really worked the way these brands hope, the first thing it would tell them is to stop lighting money on fire.

But they don’t. Instead, they commission yet another study, set up another framework, and add more columns to their already bloated spreadsheet hellscape, all while their media buyers sit there thinking, Cool, but what do we actually change?

How to Stop the Madness

If you actually want to measure things in a way that helps you make decisions—real, tangible, move-the-budget decisions—you need to do something radical: use common sense. And that means building a measurement strategy that does not rely on just one method, or worse, whatever Google is pushing this quarter. Here’s what that looks like:

• A Day-to-Day Attribution System:
  No, it’s not perfect. No, it won’t capture everything. But it will give you a consistent read on performance so that when your CMO asks what happened to sales yesterday, you don’t have to say, Give me three months and a Python script to run an MMM model. Whether it’s Google Analytics, platform-reported data, or something like Rockerbox, you need a daily pulse.

• Incrementality Testing (aka, Grow Up and Run an Experiment):
  Want to know if your ads actually work? Turn them off. If you run a $10 million campaign and sales don’t budge, guess what? That channel is a big fat waste of money—and no amount of MMM hand-waving is going to change that. The best brands test constantly, whether it’s holdouts, geo-based experiments, or even ghost bidding (RIP, cookie tracking).

• MMM (If You Must, But Set Expectations Lower Than Your Intern’s Salary):
  MMM is not a Monday-morning decision-making tool. It’s not going to tell you whether to spend more on TikTok this week or if that influencer campaign actually moved the needle. What it can do is help validate long-term trends—if you use it right. The problem is, most brands use it as a CYA strategy instead of an actual planning tool.

Analytical Breakdown: Tweaking Multimedia Mix Modeling (MMM) for Maximum Accuracy

Subscribe to our premium content at ADOTAT+ to read the rest.