Marketing budgets are under the microscope like never before. CFOs demand proof of ROI, privacy regulations are stripping away tracking data, and traditional measurement methods—marketing Mix Modeling (MMM) and attribution—are proving unreliable in an era of fragmented media.
Is marketing even working?
This existential question is what AdExchanger’s recent articles—"MMM Is Going to Fail Us Again Unless We All Think Bigger" and "From Theory to Practice: How Organizations Can Embrace Experimentation in Marketing Measurement"—attempt to answer. Both argue for a new approach to measurement. Still, they fall into the same trap: treating MMM and experimentation as silver bullets rather than recognizing the fundamental flaws in measuring marketing.
If marketing is going to prove its worth, it needs a framework—one that integrates financial rigor, pricing strategy, and real-time market intelligence.
Why MMM Is Failing (Again)
Marketing Mix Modeling (MMM) has long been a cornerstone of optimization, using statistical regression to estimate how different media channels contribute to business outcomes. But in today’s dynamic environment, MMM suffers from three fatal flaws:
It’s Too Slow
Traditional MMM is retrospective—it looks at months or years of data and provides recommendations long after making decisions. By the time a model suggests reallocating a budget, consumer behavior, economic conditions, or platform algorithms have already shifted.
It’s Too General
MMM operates at an aggregate level, missing granular, real-time insights that drive modern marketing. It assumes media effects are linear and stable, ignoring the nuances of personalized digital experiences, influencer marketing, and social proof.
It’s Easy to Game
Because MMM relies heavily on assumptions, it’s easy for agencies and media platforms to manipulate the inputs to make their channels look more effective. The result? Overinflated TV and digital video budgets at the expense of performance-driven channels.
💡 Example: When Procter & Gamble cut $200 million in digital ad spend in 2017, they saw no impact on sales. Their MMM models had overestimated digital’s contribution to revenue—proving that even sophisticated brands can be misled by flawed measurement.
But Experimentation Alone Won’t Save Us
The alternative to MMM—experimentation, including A/B testing, geo-experiments, and lift studies—has its issues. While experimentation offers causal insights (proving what works rather than just showing correlation), it falls short in key areas:
It’s Expensive. Running controlled experiments at scale requires costly infrastructure, dedicated control groups, and lost revenue from suppressed treatments.
It’s Hard to Generalize. An experiment that works in New York might fail in Chicago due to market differences.
It Doesn’t Capture Long-Term Effects. Short-term tests can measure immediate conversion lift but struggle to assess brand equity or long-term customer retention.
💡 Example: Groupon’s deep-discounting strategy drove short-term sales spikes but failed to generate long-term loyalty. Experimentation optimized for immediate revenue—at the cost of brand perception and repeat purchases.
A Smarter Approach: Measurement Beyond MMM and Experiments
Instead of choosing between MMM (slow, broad, flawed) and experimentation (precise but narrow and costly), marketers should adopt an integrated measurement framework that combines:
Low-Tech Marketing Math for Fast, Actionable Insights
Harvard’s Low-Tech Marketing Math (Dolan & Wathieu) proves that many marketing decisions can be made using fundamental break-even analysis and margin calculations—without waiting for a complex model to run. If a campaign requires an unrealistic lift to be profitable, it’s dead on arrival—no need for an MMM study or A/B test.
💡 Example: Before launching its new streaming service, Disney+ estimated the customer lifetime value (LTV) of a subscriber and used simple financial math to justify aggressive promotional pricing—without needing complex attribution models.
Value-Based Pricing as a Built-In Measurement Tool
Most marketing models ignore the most valuable signal of all: price sensitivity. Pricing strategy (Dolan & Gourville, Pricing Strategy) offers a real-time experiment: How much are customers willing to pay? If a marketing campaign increases willingness to pay, it improves brand perception. If it only drives volume at lower margins, it’s unsustainable.
💡 Example: Netflix routinely adjusts subscription pricing and monitors churn rates as a leading indicator of brand value. Marketing isn't reinforcing pricing power if a price increase leads to mass cancellations.
Market Intelligence: Why You Need Real-World Inputs
Marketing models often assume that past performance predicts future success. However, real-world market intelligence (Dolan & John, Marketing Intelligence) can continuously refine assumptions based on customer perception and competitor behavior.
Instead of relying on fixed conversion rates, track external factors—seasonality, economic shifts, and competitor pricing.
Instead of broad attribution models, use perceptual mapping to understand how customers see your brand vs. competitors.
💡 Example: Tesla spends $0 on traditional advertising but dominates earned media and brand perception. If measured using MMM, Tesla would look like a failed brand—but it’s a case study in marketing intelligence.
The debate between MMM and experimentation is missing the point. No single model will ever fully capture marketing effectiveness in today’s fragmented, fast-changing landscape.
Instead, marketers need to:
✅ Use Low-Tech Math for fast, directional decision-making.
✅ Leverage Value-Based Pricing to track marketing’s impact on willingness to pay.
✅ Incorporate Market Intelligence to refine models based on real-world behavior.
Marketing is no longer a game of optimizing for click-through rates or last-touch attribution. The real question is: Are we measuring what matters? The marketers who win in this new era won’t be the ones who stick to outdated models; they’ll be the ones who rethink measurement from the ground up.
Comments