Marketing Mix Modeling (MMM)

What Marketing Mix Modeling Actually Does

As Patrick Gilbert explains in Never Always, Never Never, marketing mix modeling operates like viewing your business from 30,000 feet. It's a top-down model that asks a fundamental question: when spend goes up in certain places, does the business tend to grow? MMM uses historical data across your channels and attempts to control for the real-world noise around them—seasonality, pricing, promotions, economic conditions, competitor activity, distribution changes, PR, and everything else that influences revenue whether marketing is good or not.

This makes MMM a powerful budgeting and planning tool. It helps you decide how to allocate a finite pool of money across major channels and levers over a quarter or a year. It sets direction and helps you make smarter allocation decisions at the strategic level. However, MMM moves slowly. It needs enough history to learn from, and it can't reliably tell you what to do this week. It can tell you that TV or YouTube tends to pay off, but it can't tell you which specific creative is driving that effect.

The NFL Roster Construction Analogy

Gilbert uses an illuminating comparison between MMM and NFL roster construction under the salary cap. Just as an NFL general manager doesn't get an unlimited budget and must allocate a fixed dollar amount across different positions, marketers face finite budgets and competing priorities. You have to build a complete marketing "roster"—brand awareness, demand generation, conversion optimization, retention—knowing that most of these investments will never directly "score" in attribution reports.

Like a smart GM who knows you can't evaluate performance by looking only at who scored touchdowns, effective marketers understand that MMM reveals how each marketing function can be greater than the sum of its parts. The offensive linemen of marketing—brand campaigns, upper-funnel media, creative that changes perception—rarely appear in attribution reports but quietly shape outcomes everywhere else.

MMM's Strengths and Limitations

MMM excels at strategic allocation decisions because it captures the full customer journey and accounts for factors that attribution models miss. It can measure the impact of offline advertising, account for view-through effects, and identify the compounding value of sustained market presence. This makes it particularly valuable for understanding how brand marketing contributes to business outcomes over time.

However, MMM has significant limitations. Because it relies on historical data, it's vulnerable to bias. If a particular channel has underperformed for years, the model may conclude that the channel simply isn't valuable—when the real issue might be poor execution or strategy. As Gilbert notes, an MMM model may be accurate about the past and still wrong about the future. This is where judgment and experience become critical.

The Measurement Triangle: MMM with Attribution and Incrementality

No single measurement tool can do everything. As outlined in Never Always, Never Never, MMM works best as part of a three-tool system alongside attribution and incrementality testing. MMM sets the guardrails and allocation strategy at the macro level. Attribution provides fast, directional signals for tactical optimization within channels—which creative, which audience, which keyword. Incrementality testing pressure-tests the assumptions and separates correlation from contribution through controlled experiments.

These methods aren't competitors—they're complements, like different roles on a football roster. Each tool makes the others better when used in the right sequence and for the right purpose. MMM informs your strategic budget allocation, attribution guides your tactical execution, and incrementality validates your biggest bets.

From Scoreboard to Film Room

Gilbert emphasizes a crucial distinction: measurement tools like MMM are film rooms, not scoreboards. They're designed for learning and optimization, not for evaluating teams and partners. When MMM becomes a "scoreboard"—a clean, definitive verdict on performance—the incentives become distorted. Teams start optimizing for the model rather than for business outcomes.

The right approach separates two activities: evaluating campaigns and tactics (the film room) versus evaluating teams and partners (the scoreboard). MMM belongs in the film room, helping you understand what worked, what didn't, and what to change next. Business health—revenue growth, profit, market share, brand demand—serves as the real scoreboard.

All models are wrong, but some are useful.

George Box, as referenced by Patrick Gilbert

Related Terms