Media Mix Modelling: The Budget Decision Framework Every Brand Needs

Every brand with a media budget faces the same fundamental question: how much should go to Google vs Meta vs SEO vs email? The honest answer, for most businesses, is that nobody actually knows — they’re guessing based on platform-reported ROAS figures that are designed to flatter the platform reporting them.

Why Platform ROAS Is Not a Reliable Signal

Every ad platform wants credit for every conversion it can plausibly claim. Meta uses a 7-day click, 1-day view attribution window by default — meaning if someone sees your Facebook ad on Monday and buys on Sunday after clicking a Google search ad, Meta claims the conversion. Google claims it too.

The result: the sum of all platform-reported ROAS figures is always higher than your actual blended return. In extreme cases, platforms collectively claim two or three times the revenue that was actually generated. Budget decisions made on these figures systematically over-invest in channels that are good at claiming credit and under-invest in channels that generate demand but rarely touch the last click.

What Media Mix Modelling Actually Is

Media Mix Modelling (MMM) is an econometric technique that uses statistical regression to estimate the incremental contribution of each marketing channel to revenue. It analyses your historical spend, revenue and external factors (seasonality, economic conditions, competitive activity) to produce a model of how each marketing input affects output.

Unlike last-click attribution, MMM doesn’t rely on individual user tracking. It works at an aggregate level — correlating changes in spend with changes in revenue over time — which makes it robust to privacy changes and better at capturing the contribution of upper-funnel channels like brand advertising that rarely appear at the last click.

The output: a set of revenue contribution curves for each channel — how much incremental revenue does each additional dollar of spend generate, and where does the curve flatten (the point of diminishing returns).

Reading an MMM Output

The most useful MMM output for budget decisions is the marketing contribution chart — a decomposition of total revenue into base (what you’d earn without any marketing) and incremental contributions from each channel.

Most brands are surprised by three things when they first see this: SEO and brand equity contribute far more to base revenue than they realised; paid social contributes less incremental revenue than platform reporting suggested; and email marketing is dramatically under-credited by last-click attribution.

The second key output is the response curve per channel — how ROAS changes as spend increases. This tells you which channels have headroom (ROAS still high at current spend) and which are at diminishing returns (ROAS declining, meaning reallocation would improve overall efficiency).

Running MMM Without an Enterprise Budget

Full econometric MMM traditionally required expensive analysts and large datasets. Recent years have seen the emergence of simpler approaches accessible to mid-market brands. Robyn (Meta’s open-source MMM tool), lightweight Bayesian MMM frameworks, and Meridian (Google’s open-source MMM) can all be implemented with internal data science resources or specialist consultants at a fraction of traditional MMM project costs.

The minimum viable dataset is 2 years of weekly spend and revenue data per channel — most brands have this in GA4, their finance system and platform exports.

From MMM to Actionable Budget Changes

MMM produces a recommended optimal budget allocation — the mix that would generate maximum revenue at your current total spend level. The distance between your current allocation and the optimal is your opportunity.

In practice, move gradually. Shift 20–30% of budget toward the recommended channels and re-run the analysis quarterly as new data accumulates. The most common MMM finding: brands are over-invested in Meta and under-invested in Google Search, organic content and email. The second most common: seasonal budget allocation doesn’t match demand — brands spend heavily in January when intent is low and pull back in Q4 when intent is highest.