Practice · 02 · Modelling

    MMM
    Advisory

    Make MMM outputs usable in planning, budgeting and real investment decisions.

    § Intro

    The goal of MMM is clarity, not consensus.

    The most dangerous model is not a broken one. It is a plausible one that quietly reinforces the wrong decisions. Plausibility is not validity.

    When outputs align with expectations, they rarely get challenged. Budgets follow the model, and the model reflects the budget that came before it.

    This work looks past statistical fit to test whether the model reflects past decisions, or reveals what could happen next.

    The Two Tensions · Intellectual Spine
    Tension 01

    Plausibility vs.
    Validity

    The most dangerous model is not a broken one. It is a plausible one that quietly reinforces the wrong decisions. Plausibility is not validity. When outputs align with expectations, the hard questions about what the model actually measures rarely get asked.

    Tension 02

    Reflection vs.
    Potential

    Budgets follow the model, and the model reflects the budget that came before it. This work looks past statistical fit to test whether the model reflects past decisions, or reveals what could happen next.

    What This Covers

    Capabilities · 04
    01

    Breaking the Feedback Loop

    Most models reflect historical allocation. If TV was constrained, it appears weak. If digital was overfunded, it looks efficient. This isolates where the model is describing past decisions, not future potential.

    02

    Scale vs. Effect

    Channels often appear inefficient simply because they were underfunded. This tests whether weak results are real, or a function of insufficient scale.

    03

    Cross-Model Validation

    Different frameworks will give you different answers. The insight is not in forcing alignment, but in understanding what the disagreement reveals about risk and robustness.

    04

    From Output to Decision

    MMM does not allocate budget. People do. This translates model outputs into clear, defensible planning decisions, including what to scale, protect, or reduce.

    Experience across Meridian (Bayesian) · Robyn (Open-Source) · Strata QED

    § When to call

    Moments that start the conversation

    01

    The Planning Gap

    You have results, but no clear way to turn them into a plan.

    02

    The Reality Mismatch

    The model looks right, but the market does not reflect it.

    03

    Conflicting Answers

    Different models or partners are pointing in different directions.

    04

    The Investment Shift

    You are preparing for a major budget reallocation.

    05

    The New Build

    You are commissioning a model and want it designed for decisions, not reporting.

    These moments tend to surface the same questions

    § Where conversations often begin

    Common questions

    Q · 01

    Do you work with different MMM frameworks?

    Yes. We work across Meridian, Robyn, PyMC, constrained OLS, Adaptive MMM and Strata QED. The focus is not the framework itself, but whether the model is appropriate for the question, how robust the outputs are, and how they translate into decisions.

    Q · 02

    Why do different MMM models produce different results?

    Because they make different assumptions about how marketing works. Variations in data treatment, carryover effects, saturation curves, and prior assumptions can all lead to different conclusions. The difference is not an error. It is a signal that needs to be understood before making decisions.

    Q · 03

    Which MMM model is the most accurate?

    There is no universally correct model. Accuracy depends on the question being asked, the data available, and how results are interpreted. The priority is whether the model is valid for the question being asked, sufficiently reliable, and useful for decision-making, not simply statistically well-fitted.

    Q · 04

    Can MMM results be trusted for budget allocation?

    Yes, but only when interpreted correctly. MMM provides directional guidance on contribution and diminishing returns, but it should be stress-tested against commercial reality before being used to make major allocation decisions.

    Q · 05

    Do you build MMM models or review existing ones?

    Both. This includes reviewing existing models, challenging assumptions, and translating outputs into planning decisions. In some cases, this also involves rebuilding or recalibrating models where they are not decision-ready.

    Related Perspectives

    Make your model usable for decisions.

    If your model is shaping decisions, or about to, you need to know what to trust, what to challenge, and what to change before budgets move.