MMM is back, and it still isn’t enough for challenger brands

Thirty-odd years ago, if you ran marketing at a big consumer goods company, there was a good chance a team of econometricians was building you a marketing mix model. Barcode scanners had arrived in the supermarket, and for the very first time you could line weekly sales up against weekly spend and ask what moved what.

Then digital arrived, every click could be tracked, and MMM started to look like a slow, expensive relic. For the best part of fifteen years it was a luxury for the very biggest advertisers.

Now it’s back. Privacy rules, Apple’s tracking changes and the long goodbye to the cookie broke user-level attribution, and MMM, which never needed to follow individuals, walked back into the room. In an EMARKETER and TransUnion survey, nearly half of US marketers (46.9%) said they planned to invest more in it, and more of them named it their most reliable measurement method than any other.

I understand the appeal. But I’d like to make a case that won’t be popular in every effectiveness circle: for most challenger brands, MMM still isn’t good enough.

Comebacks don’t fix the original flaws

MMM didn’t come back because it got better. It came back because the thing that replaced it broke.

That’s worth sitting with. The methods have improved, and the new Bayesian tools are a genuine step forward. But the weaknesses that made marketers drift away twenty years ago are structural, and they’re still there.

It looks backward, at the speed of a quarter

A typical MMM is built on two or three years of history and refreshed once or twice a year. By the time it tells you what worked, the moment to act has usually gone.

That’s fine for settling next year’s budget split. It’s no help at all when a competitor launched on TikTok three weeks ago and is already showing up on your shelf at Ulta.

It’s built to see the short term

Most models are designed to catch the sales response that lands within weeks of the spend. But the IPA’s Profit Ability research found advertising effects can persist from four months to more than two years after a campaign ends.

That long tail usually ends up in a bucket called “base sales”, or gets added back with a multiplier. A WARC paper on econometrics noted that many practitioners simply multiply the short-term return by four.

In other words, the most important number in your brand case is often an assumption someone typed into a slide.

It can’t see your competitors

MMM models your spend against your sales. Competitors show up as a control variable, if they show up at all.

It can tell you whether your TV flight paid back. It can’t tell you that the brand next to you on the shelf is gaining momentum, or why. For a challenger, that’s half the picture. Competitors will never hand you their media plan, but they do show up in the same behavioral data you do.

It only models what you paid for

MMM is built on spend. But a great deal of what moves a challenger brand isn’t bought: earned conversation, search demand, a creator who picked you up unprompted, how often you’re recommended when someone asks ChatGPT what to buy.

Those are hard to put in a spend column, so they tend to get left out of the model. They don’t get left out of the market.

It needs a scale most challengers don’t have

Open-source tools made the math free. They didn’t make the team free. A credible model still needs clean data, enough variation in spend to learn from, and people who know what they’re doing.

The Marketing Science Institute’s John Lynch described the boom as “a modern-day Wild West”, full of new players making big claims. For a $100M brand with lumpy, concentrated spend, the model has less to learn from, and the uncertainty shows.

And it rarely turns into action

Only 28% of marketers say their organization is very effective at turning MMM insights into action, according to eMarketer. That doesn’t surprise me. A model that tells you how last year’s money performed, a few months late, struggles to tell you what to do on Monday.

What MMM does that an index doesn’t

In fairness, MMM does something an always-on view of the market can’t. It estimates incrementality: the sales your marketing actually caused, net of price, promotions and seasonality. It models diminishing returns, so it can tell you where the next dollar stops paying back. Finance trusts it, and it’s privacy-safe by design.

Grace Kite, the economist who founded the econometrics consultancy Magic Numbers and has supported twelve IPA Effectiveness Award winners, has spent years arguing that econometrics belongs to mid-sized brands, not just the giants. She’s also clear about its weak spot: “a lot of econometrics gets done and then not used.” Her own view is that it’s one key piece of the evidence, not the whole of it.

That’s the right way to see it. MMM is a verification layer: once or twice a year, it tells you how much of what happened was really yours. What it can’t be is the monitoring layer that tells you, in between, where to look and what’s moving.

What a challenger actually needs to know

The questions I hear from challenger CMOs aren’t really allocation questions. They’re: are we gaining or losing momentum against the brands we compete with? What’s driving it? And what happens to sales if we hold spend, or cut it?

Answering those takes something that’s behavioral, competitive, always-on, covers what you didn’t pay for as well as what you did, and looks forward rather than back.

If you already run an MMM, keep it for the annual budget split. Just don’t expect it to answer this week’s questions. If you’re weighing your options, we’ve laid out how MMM compares with always-on brand intelligence.

MMM is a very good answer to last year’s question. Challengers need an answer to this week’s.