How Procter & Gamble rebuilt its media mix modeling capability and what it actually changed

In 2017, Procter & Gamble's then-CMO Marc Pritchard stood in front of the IAB Annual Leadership Meeting and delivered one of the most candid indictments of digital advertising the industry had heard from a major advertiser. He cited viewability fraud, murky agency contracts, and what he called a "crappy media supply chain." What got less attention at the time was the deeper measurement problem underneath: P&G, like most large advertisers, had gradually handed over its understanding of what actually drove sales to a patchwork of platform attributionattributionA framework for assigning credit to the touchpoints that contributed to a conversion, so you can measure which channels and interactions actually drive results.View full definition → tools, each of which had a structural incentive to claim credit for conversions. The company's investment in econometrics and media mix modeling had atrophied through the early 2010s, replaced by last-click and multi-touch attributionmulti-touch attributionA method that distributes conversion credit across all marketing touchpoints in the customer journey, rather than crediting only the first or last interaction.View full definition → models that were faster and cheaper but increasingly disconnected from business outcomes.

By the early 2020s, the limits of that approach had become impossible to ignore. Cookie deprecation timelines were accelerating. Walled garden reporting from Meta and Google remained opaque. And P&G's own dataown dataData collected directly from your own customers and prospects through your own channels: your most reliable and privacy-compliant source.View full definition → showed that aggregate sales responses to media spend were not matching what the attribution dashboards predicted. The company had grown its digital share of media spend significantly, trimmed its traditional TV presence, and was not seeing the brand health metrics it expected. It needed a methodology that could hold the full picture together across channels, time horizons, and geographies.

What P&G actually did

The rebuild was not a single initiative but a multi-year shift in measurement philosophy. P&G began investing heavily in what the industry now calls augmented or continuous MMM, which differs from the quarterly or annual econometric studies that were standard practice in the 1990s and early 2000s. Rather than commissioning a retrospective model once a year, the company built infrastructure to run models on a rolling basis, updating inputs weekly or bibiTechnologies and processes that turn raw data into actionable insights via reporting, dashboards and analysis, so teams can decide based on facts rather than intuition.View full definition →-weekly as new sales and media data came in.

This shift required genuine investment on the data side. P&G had to clean and standardize media delivery data across dozens of markets, which is less glamorous than the modeling work itself but entirely determines whether the outputs are trustworthy. The company also worked to integrate first-party retail sales data more tightly into the models, so the dependent variable (actual sales) was more granular and timely than syndicated panel data alone.

P&G also made a structural decision that many advertisers avoid: they brought meaningful modeling capability in-house rather than relying entirely on agency or vendor-run black boxes. This does not mean they stopped working with external partners. Consultancies and specialist analytics vendors still contributed, particularly for market-specific work. But the internal team could interrogate assumptions, stress-test model specifications, and push back when an output looked implausible. That interpretive capability matters as much as the statistical machinery.

A related move was the deliberate reconnection of MMM outputs to media planning decisions at the investment stage, not as a post-hoc audit. Results fed into budget allocation discussions across the portfolio, with category teams expected to show how their channel mix choices were grounded in marginal return estimates from the models.

The results, as far as they are publicly known

P&G's overall marketing efficiency story through the early 2020s is well documented in broad strokes, even if granular MMM-specific metrics are not publicly disclosed. The company reported in its 2022 annual results that it had achieved what it called "balanced growth" across categories even as it reduced certain forms of media spend and shifted mix. Over several years, the company cut back roughly $2 billion in what Pritchard described as ineffective marketing, while maintaining or growing market sharemarket shareThe percentage of total industry sales your company captures in a given period. It measures competitive position relative to rivals in a defined market.View full definition → in key categories. Analysts covering the company generally credited tighter measurement discipline as a contributing factor.

What is harder to attribute specifically to the MMM rebuild is the precise revenue impact, and any vendor or consultant claiming to know that figure with precision should be viewed skeptically. What is more verifiable is that P&G became increasingly vocal publicly about the limitations of platform-reported attribution, which is consistent with an organization that had built enough internal modeling capability to see the discrepancies. That institutional confidence does not come from buying a dashboard; it comes from years of methodological investment.

What transfers, and where it does not

The P&G case holds several practical lessons for CMOs who are not running a $80 billion consumer goods company.

The most transferable point is the decision to treat MMM as an ongoing analytical function rather than a periodic project. A model run once a year produces findings too stale to act on. The infrastructure cost of continuous modeling has dropped significantly in 2026 compared to even five years ago, with cloud computation making it feasible for mid-sized advertisers.

The in-house capability question is context-dependent. P&G has the scale to justify senior data scientists working exclusively on media econometrics. A company with $500 million in annual revenue probably does not. The realistic alternative is building enough internal literacy to be a credible client of external modeling partners: knowing what questions to ask about model structure, understanding what the confidence intervals actually mean, and having a clear link between model outputs and budget decisions.

The data qualitydata qualityThe degree to which data is fit for purpose: accurate, complete, consistent, timely, valid and unique. Poor quality data undermines analytics, reporting and AI.View full definition → point is universal and consistently underestimated. Before any organization spends on modeling software or external consultants, it needs a rigorous audit of whether its media delivery data and sales data can actually be joined cleanly. In most companies, they cannot, at least not without significant work. That foundational step is where most MMM projects fail quietly.

One area where the P&G experience does not transfer cleanly is brand investment scale. MMM produces more statistically reliable outputs when there is substantial variation in spend levels over time and across markets. Companies with relatively flat media budgets and limited geographic variation will get noisier results and need to calibrate their expectations accordingly.

The broader shift P&G represents is a move away from letting platform-reported attribution define marketing's contribution to the business. That shift matters regardless of company size. Attribution tools built by the same companies selling the media have a structural problem that no technical sophistication resolves. MMM, with all its limitations, at least operates from an independent analytical position. P&G's investment in that independence is what made its measurement credible internally and externally. That credibility is what ultimately gives a CMO the standing to defend or redirect a media budget.