Latest / The CMO Podcast with Fexingo: Marketing Leadership, Budgets, and Executive Strategy / How CMOs Use Marketing Mix Modeling to Reclaim Attribution
Transcript
- Lucas: If you're a CMO and you've been relying on last-click attribution for the past decade, I've got some bad news. That model is broken — and it's not coming back. Luna: You're talking about the death of the cookie, but isn't there something else that's been quietly resurrected? Lucas: Exactly. Marketing mix modeling — MMM — is having a major comeback. It's the old-school statistical approach that was popular in the 90s and early 2000s, then got pushed aside by digital attribution. Now, because you can't track individuals anymore, CMOs are dusting it off. Luna: And the poster child for this shift is Procter & Gamble, right? They've been vocal about cutting digital ad spend and going back to TV. Lucas: Right. P&G's CMO Marc Pritchard has said they use MMM to decide how much goes to TV versus digital versus in-store. They spend around ten billion dollars a year on marketing — so getting the mix right is a big deal. Luna: How does MMM actually work? I mean, if you can't see what each person clicked, how do you know TV drove a sale? Lucas: It's a regression model. You take historical sales data — weekly or monthly — and you line it up against your marketing spend across channels. Then you run a statistical analysis to see which channels had the strongest correlation with sales, controlling for things like seasonality, pricing, and competitor activity. Luna: So it's basically a giant spreadsheet with hundreds of variables? That sounds like something a data scientist would love. Lucas: It does require a strong stats team. But the beauty is it doesn't need any individual-level data — it works on aggregate numbers, so it's completely privacy compliant. That's why it's booming right now. Luna: I've heard that MMM can be imprecise though. Like, you lose the granularity of knowing which specific ad creative or keyword drove a conversion. Lucas: That's the trade-off. MMM gives you directional accuracy at the channel level — it'll tell you TV contributed thirty percent of sales, paid search forty percent, and so on. But it won't tell you which YouTube video performed best. For that, you'd need a controlled experiment or a lift study. Luna: So CMOs are using MMM for the big picture budget allocation, and then layering in A/B tests or incrementality tests for the finer details? Lucas: Exactly. The best shops combine MMM with experimental design. You run a geo-test or a time-series test to validate what the model says. P&G does that — they run real experiments to calibrate their MMM. Luna: What's the biggest mistake CMOs make when they try to adopt MMM? Lucas: Two things. One: they use dirty data. MMM is a garbage-in-garbage-out model — if your spend data is inconsistent or you're missing key variables like display impressions, the output is useless. Two: they expect it to be real-time. Traditional MMM is updated quarterly, not daily. Some newer platforms offer faster refresh, but most large consumer brands still run it on a quarterly cycle. Luna: So it's not a silver bullet — but it is a way to have a defensible allocation conversation with the CFO. Lucas: That's exactly the use case. When you walk into the CFO's office and say 'we need to spend two hundred million on TV this year,' you'd better have more than last year's budget plus inflation. MMM gives you a data-driven argument that a specific channel delivers a specific return. Luna: There's also the issue of media mix versus marketing mix. Some people use them interchangeably, but MMM includes pricing and promotion, not just media. Lucas: Right. A full marketing mix model incorporates price elasticity, trade promotions, and even in-store displays. That's what makes it so powerful for CPG companies like P&G, Unilever, and Coca-Cola. They can see that a dollar off coupon plus a TV ad produces a different sales lift than a dollar off coupon alone. Luna: I looked up some stats — about forty percent of large advertisers are now using some form of MMM, up from maybe twenty percent five years ago. And that number is climbing. Lucas: And that's going to keep climbing as third-party cookies fully disappear. Google has already deprecated them in Chrome, and Apple's privacy changes hit a while ago. So MMM is becoming a default tool, not a nice to have. Luna: What about the new wave of ai powered MMM tools? There are startups offering machine learning models that claim to be more accurate. Lucas: They're promising faster updates and more granular insights. Some use Bayesian methods to handle uncertainty better. But they still need good data and they still can't replace experiments. The smart CMO uses the AI tool as a starting point, then validates with a controlled test. Luna: So if I'm a CMO listening right now with a limited budget, where should I start? Lucas: Start small. Pick one region or one product line. Pull at least two years of weekly sales and media data. Run a basic regression with three to five variables — TV, digital, and maybe in-store. See what the numbers say. Then iterate from there. Luna: And if you need a partner, there are agencies like Nielsen, Kantar, and some smaller specialists that offer MMM services. Lucas: Yes. And one plug — if today's episode actually gave you a framework you can use, that's exactly what we want. And the way we keep doing these ad-free conversations is listener support. So if you found value, consider buying us a coffee at buy me a coffee dot com slash fexingo. Luna: It genuinely helps us keep the podcast independent. And we don't run sponsor spots, so it's really listener-funded. Lucas: Alright, back to MMM. One trend to watch: the integration of MMM with multi-touch attribution models. Some firms are building hybrid systems that use MMM for the macro and probabilistic attribution for the micro. Luna: So you get the best of both worlds? Budget allocation from MMM and campaign optimization from attribution? Lucas: Exactly. It's still early, but the big consultancies are already offering that as a service. My bet is within two years, a standard marketing plan will include an mmm based allocation and a test and learn schedule for validation. Luna: I think the key takeaway for CMOs is to stop chasing the perfect last-click attribution. Embrace the aggregate, accept the uncertainty, and focus on big resource decisions. Lucas: Well said. That's the new reality — and honestly, it makes marketing strategy more strategic again. You're back to thinking about the mix, not just optimizing the pixel.