Game Data Pros CEO William Grosso joined Recast’s Coffee Breaks series to tackle one of the hardest practical questions in marketing measurement: Why most MMMs are probably wrong (and what to do about it)?

Recast’s Coffee Breaks series features deep-dive, practitioner-focused conversations on topics like MMM design, incrementality testing, and marketing experimentation with leading experts and operators across industries. (youtube.com)

Hosted by Recast co-founder Michael Kaminsky, the conversation digs into the tradeoffs between elegant models and usable ones. Rather than chasing a perfectly detailed representation of reality, Bill argues that teams should focus on whether the model’s assumptions and forecasts hold up in practice, especially on the channels that matter most for spend (LinkedIn).

A key theme of the discussion is the idea that what you do with the MMM, and how you use it over time, matters far more than a single static model artifact. Instead of treating defining and building an MMM as a one-time exercise that spits out a fixed “optimal allocation,” the conversation centers around:

• Weekly model updates that reflect the latest performance data.
• The idea of a “Spend flight” as a sequence of weekly budget spends, designed both for financial reasons (optimizing revenue while reducing risk) and as a learning exercise (carefully adjusting channel spends to learn and improve the model)
• Treating each major budget shift as a designed experiment, with the model’s forecasts as ex-ante expectations.
• Using each week as a checkpoint: did reality line up with the model, and if not, what does that tell youabout our assumptions? (LinkedIn)

This approach turns MMM into a self-correcting system rather than a static report—tightening the feedback loop between model, spend decisions, and observed outcomes. Over time, the model becomes more robust precisely because it’s being “tested in production” every week.

About Game Data Pros

Game Data Pros (GDP) partners with leading gaming, digital entertainment, and sports-betting companies to design and operate high-impact experimentation programs, Bayesian media mix models, and scenario-planning tools. GDP focuses on making advanced analytics operationally useful—helping teams turn sophisticated models into better product, marketing, and portfolio decisions.