Marketing mix modelling, or MMM, is a statistical method for measuring how each of your marketing channels contributes to sales. Rather than tracking individual users the way pixel attribution does, it analyses the relationship between spend and revenue across channels over time, which makes it resistant to the tracking losses that have degraded platform attribution since iOS 14.

For years MMM belonged to enterprise brands with data teams. That has changed. Open source tools and AI-assisted modelling have pulled it within reach of ecommerce brands doing $5M to $30M, at exactly the moment platform dashboards have become least trustworthy.

The commercial appeal is blunt: your Meta dashboard and your Google dashboard both claim credit for the same orders. MMM arbitrates. This guide explains how marketing mix modelling works, when a DTC brand is big enough to use it, and how to run a first model without hiring a statistician.

What is marketing mix modelling?

MMM is regression analysis applied to your own trading history. It takes weekly or daily data on spend per channel, revenue, pricing, promotions and seasonality, and estimates how much each input actually moved sales. The output is a set of channel contributions and diminishing-return curves: what each channel really drove, and what the next pound spent on it is likely to return.

Because it works on aggregate data rather than user tracking, it needs no cookies, no pixels and no consent banners. That is why the method, decades old in FMCG, has come back into fashion for ecommerce.

How does MMM differ from pixel attribution?

Pixel attribution follows individual users and assigns credit for each order to the touchpoints it managed to observe. MMM ignores individuals entirely and asks a colder question: when spend on a channel rose or fell, what happened to total sales? Attribution answers quickly and granularly but sees a shrinking share of the journey. MMM answers slowly and at channel level but sees everything, including the sales your dashboard credits to nothing. They are complements rather than rivals: attribution for week-to-week optimisation, MMM for where the budget should live.

Why platform dashboards over claim (and how MMM arbitrates)

Every platform grades its own homework. Meta, Google and TikTok each attribute conversions using their own windows and their own view of the user, so adding up dashboard revenue routinely produces more sales than your store recorded. Branded search and retargeting exaggerate the effect by claiming demand that already existed, a pattern we unpack in our guide to what ROAS really tells you. MMM sits above the argument, measuring each channel's relationship with total revenue, and across the accounts we have modelled it consistently redistributes credit away from bottom-funnel capture channels towards the demand creation the dashboards undervalue.

When is an ecommerce brand big enough for MMM?

The practical floor is less about revenue and more about data: you need roughly two years of history, meaningful spend across at least three or four channels, and enough week-to-week variation in that spend for the model to learn from. In revenue terms that usually means brands from around $5M upwards. Below that, a disciplined incrementality testing programme, holdouts and geo tests, answers the same questions more cheaply. Above it, running blind on platform numbers starts costing real money in misallocated budget.

How to run a first model (tools, data, timeline)

The accessible route is open source: Meta's Robyn and Google's Meridian are both free MMM frameworks, and a growing set of commercial tools wraps similar models in a friendlier interface. The work is mostly data assembly: two years of weekly spend by channel, revenue, discounting and promotion calendars, and any known demand shocks. Expect a first usable model in six to eight weeks, then treat it as a living tool refreshed quarterly rather than a one-off report. Crucially, validate it: when the model claims a channel is under- or over-credited, run a holdout test and see if reality agrees. A model that survives that test earns a seat in budget decisions, which is how we use it inside our ecommerce paid media agency engagements.

What MMM cannot tell you

MMM will not tell you which ad, audience or creative worked; it operates at channel level. It struggles with brand-new channels that lack history, with very small budgets that never vary, and with long-horizon brand effects that unfold over years. And it is a model, not an oracle: poorly specified inputs produce confident nonsense. The healthy setup pairs MMM for allocation, incrementality tests for verification, and platform attribution for in-channel optimisation, all read by someone accountable for contribution margin. That triangulated view is what we build for clients as an ecommerce marketing agency: not a prettier dashboard, but a decision system the numbers can actually support.

Frequently asked questions

Is MMM better than triple pixel attribution tools?

They answer different questions. Attribution tools improve the granular, user-level view for daily optimisation; MMM measures true channel contribution without depending on tracking. Brands serious about measurement run both and let incrementality tests referee disagreements.

How much data does MMM need?

Roughly two years of weekly data across spend, revenue and promotions, with genuine variation in spend. More history and more variation produce tighter estimates; a channel that never changes budget gives the model nothing to learn from.

How much does marketing mix modelling cost?

Open source frameworks like Meta's Robyn and Google's Meridian are free, so the real cost is analyst time for data assembly and modelling, typically a few weeks of work. Commercial MMM platforms charge subscriptions that generally make sense from seven-figure ad budgets upwards.

How often should the model be refreshed?

Quarterly is the working rhythm for most DTC brands: frequent enough to steer budget shifts, infrequent enough for new data to accumulate. Refresh sooner after major changes to channel mix, pricing or product range.

Let's get in touch

If your dashboards all disagree and the budget meeting runs on folklore, we help founder-led Shopify and DTC brands in the UK and US measure and scale profitably. Book a growth call with Webtopia.