Why Most mMMs Are Probably Wrong (and What to Do About It)
with Game Data Pros CEO William Grosso on Recast’s Coffee Breaks Podcast
Last week, Game Data Pros CEO William Grosso joined Recast CEO Michael Kaminsky on Recast’s Coffee Breaks podcast to tackle a deceptively simple — yet critical — question: how do you validate your media mix model (mMM), and and when is it good enough to actually use?
The key moment
In the discussion, Bill underscored a sharp trade-off many analytics teams face: chasing “perfect” versus embracing “usable.” Too often, organizations treat an MMM like a gold-standard artifact — carefully designed, the product of person-years of exhaustive analysis, lovingly handcrafted, but then almost untouchable — an academic deliverable sitting on a shelf rather than a living decision-tool.
Instead, Grosso said, the more valuable approach is to ask: do the model’s assumptions hold up in practice? Are you capturing the channels that matter for spend decisions? And are you updating your insights in a way that reflects what’s actually happening in performance?
What that looks like in practice
From the conversation, three operational patterns emerged as markers of a “good enough” MMM:
Weekly updates: Rather than one-time modelling runs, the model is refreshed with latest performance data, keeping the decision cycle tight.)
Spend “flights” as experiments: Rather than assuming the model will give the perfect allocation and then executing it blindly, each major budget shift is treated as a designed experiment — we act, observe, adjust.
Feedback loops: Each week becomes a check-point: did reality align with the forecast? If not, what assumptions broke? What channel coefficients need revision? The model becomes self-correcting.
In other words: a living updated-constantly MMM, not a static report.
Why this matters (especially for gaming & entertainment)
As you know from our ongoing discussions around advanced MMMs, scenario planning and business-cycle modelling in gaming, entertainment and fantasy sports, the challenge is not simply building a technically complex model — it’s operationalizing it. That means embedding the insights in planning, spend optimization, scenario management, and ensuring that the model is actually used in decision-making.
The Recast Coffee Break conversation hits home this point: No matter how sophisticated your Bayesian sampler, hierarchical channel splits, time-varying intercepts or interaction terms are — if the model isn’t feeding into week-to-week spend allocation decisions, it’s not fully delivering. The “good enough” threshold is when modelling becomes a tool not just a “project.”
Operational takeaway for teams
For anyone working on MMMs or planning to scale to multi-million dollar budget allocations (as many of your clients do), here are a few practical take-aways distilled from the talk:
Define the “usable” threshold up front.
Ask: What are the key channels? What minimum predictive performance or forecasting validity do we require to make spend decisions? It’s okay if less material channels remain lower fidelity.Build a weekly refresh cadence.
Don’t wait for quarterly modelling runs. The faster you put model updates into action, the faster you learn and adapt.Treat each budget shift as a designed experiment.
When you move spend, do it with the model’s forecast in hand — then track actual vs forecast and ask “why the variance?” Build the model’s correction loop into the workflow.Embed the model into decision-ops, not just analytics ops.
Ensure that planning, execution, measurement and scenario-re-planning are connected.Embrace your own “Suite of Truth.”
Most companies have many different marketing measurement systems, and they rarely agree. Use them in conjunction, and make sure that you understand the strengths and weaknesses of each system.
Accept “good enough” and focus on the loop, not perfection.
The aim is to get the model working in production, and then improve it. Over-fine-tuning that delays this is often a false economy.
Make Sure to Watch the Complete Podcast
If you missed it, you can watch the full episode on Recast’s YouTube channel under their Marketing Measurement Coffee-Break Sessions playlist.
Until next time —
The GDP Team