What Good Marketing Mix Modeling Actually Looks Like

Most Marketing Mix Modeling projects don’t fail because the math is wrong. They fail because the output lands in a presentation and stops there.

The model gets built. The charts get made. Someone says “interesting” in a meeting. And then the next year’s budget gets set the same way it was set the year before.

Here is what we have learned from running MMM engagements across several CPG portfolios, and what makes the difference between a model that changes how a team spends money and one that doesn’t.

The question that matters is rarely the one being asked

The stated brief is usually some version of: “We want to understand our media ROI.” That is a reasonable starting point. But the actual question — the one that needs answering before any budget decision gets made — is almost always narrower than that.

Which channels should we cut if we need to find 15%? What happens to sales if we double down on digital and pull back on TV? How much of last quarter’s sales lift was the promotion versus the media running at the same time?

A model that answers “what is our ROI” is not the same as a model that answers those questions. The first is a measurement. The second is a decision tool. You want the second one, and building it requires knowing what decisions are actually in play before you start.

The data problem comes first

Before any modeling happens, someone needs to sit with the data and establish a single version of the truth. This takes longer than most clients expect, and it is not the interesting part of the work, but it is the part that determines whether the output is worth anything.

In practice this means:

• Sales data that aligns to the same geography and time period as the media data
• Media spend and impressions that have been cleaned and reconciled against actual invoices, not just platform exports
• Pricing and distribution data that captures changes the brand made during the measurement period
• Seasonality and competitive data for the category

If any of these are inconsistent or missing gaps, the model will find an explanation in the noise rather than the signal. The output will look clean. It will be wrong.

The model is not the deliverable

This one takes some pushback to get right. At the end of an MMM project, the thing the team needs is not a model. It is a set of budget scenarios, explained clearly enough that the person setting the budget understands what they are trading off.

That means the final output is a conversation, not a report. Someone needs to walk the marketing director through: “If you take the budget as it stands today and reallocate 20% from trade promotion into digital, here is what the model says happens to base sales, here is what happens to incremental volume, and here is the confidence interval around that projection.”

The model is the thing that makes that conversation credible. It is not the thing that gets handed over.

What “optimizing” media spend actually means

There is a tendency in this work to treat higher ROI as the goal. It is not. The goal is the right level of spend in each channel, which is a different question.

Every channel has a saturation point. Below that point, additional spend generates incremental return. Above it, you are paying for reach that has already done its job. The job of an MMM is to find those curves, channel by channel, brand by brand, and tell the team where they are sitting on each one.

In one engagement, the finding was that the brand’s highest-ROI channel was also significantly underfunded relative to saturation. The lowest-ROI channel was significantly overfunded. The reallocation was straightforward once the model showed where the curves were. The result was 3 to 5 percent incremental lift on campaigns that were already running. The budget didn’t change. The allocation did.

When it works and when it doesn’t

MMM works well when:

• The brand has been spending consistently across multiple channels over at least 18 to 24 months
• Sales data is available at a weekly or bi-weekly level
• The business is willing to actually change allocation based on the output

It works poorly when:

• The brand has been making major structural changes mid-period (new distribution, major pricing resets, category disruption)
• The team already knows what they want to hear and the project is validation, not exploration
• Nobody is willing to share what decisions are actually on the table

The second list is not a reason not to do it. It is a reason to have the honest conversation about scope before the engagement starts.

If you are considering an MMM project and want to talk through what the data requirements would look like for your portfolio, get in touch.