Marketing mix modeling measures what your marketing delivers as a whole, by comparing results over a longer period against what you spent per channel. Not one visitor’s click path, but the bigger picture across every channel. The abbreviation is MMM, which as the owner of Making Marketing Matter I obviously appreciate.
Why it is coming back
The method is old. Big brands were already calculating this way in the sixties, long before there was a click to count. Once everything moved online and became measurable, the need faded. You could see exactly who clicked on what.
That picture is breaking down. Cookies are disappearing, privacy rules are tightening measurement and more people arrive without a traceable trail. What remains is a dashboard that still shows something, only less and less of the whole.
At the same time the technique became accessible. Where you once needed an agency with econometricians, the models now sit online as open source tools. Computing power stopped being the barrier. The only question left is whether your situation makes it worthwhile.
What a dashboard misses
Attribution follows one visitor’s journey. Someone sees an ad, clicks, returns, fills something in. That path can be traced as long as it stays online and the visitor allows tracking.
Marketing mix modeling does the opposite. It looks at totals over time and searches for correlation. If radio spend went up, you see what happened to applications weeks later. That means everything you can never click on counts too: a billboard, a television campaign, an ad in a local paper.
The model also accounts for what you do not control. Season, weather, current events, what a competitor did. Exactly the things a dashboard credits to your campaign while they stood entirely apart from it.
Attribution zooms in. Modeling zooms out. They answer different questions, so you do not need one of the two but both, each in its own place.
Where it really counts
The organisations where this weighs heaviest are the large and slow-moving ones. Government, healthcare, semi-public institutions. There a campaign runs across ten channels at once, spread over departments that barely speak to each other.
The campaigns for Amsterdam show this well. Different channels, different departments, each with its own goal and its own accountability. Everyone dutifully delivers their own figures. Only the effect never comes together anywhere, because there is no button anyone gets judged on and no place where the full picture arrives.
That is the gap marketing mix modeling closes. It gives one view where the separate efforts meet, instead of ten reports each showing a fragment.
It does require someone who oversees the whole and translates it into what you do differently tomorrow. A model delivers numbers, not a decision. That is precisely the work of the architect who oversees the entire chain, alongside the specialists doing the calculations.
When it pays off
It is worth doing once you run many channels alongside each other, some of them offline, with a budget that matters and a history of a few years. Without that history there is nothing to find patterns in.
There is one more condition that often gets forgotten: there has to be variation. Anyone who put the same amount into the same channels every month for three years gives the model nothing to work with. Difference is what makes it measurable.
For smaller organisations with one or two channels it is overkill. There you get further with measurement that is properly set up and someone who understands the figures. Modeling becomes an expensive way to confirm what you already saw.
The best use is not the model itself, but what it unlocks. Once all your effort sits in one view, a conversation starts that never happened before: what are we actually doing here, and what for.
A model never tells you what to do. It only shows what happened, so you can make that choice better.
Making things simple is sometimes the complicated part.
Frequently asked questions
What exactly is marketing mix modeling?
Marketing mix modeling is a statistical method that measures what each channel contributes to a result, by comparing outcomes over a longer period against what was spent per channel. It looks at the whole picture over time, not at one visitor's click path.
How is it different from attribution?
Attribution follows the individual click journey and only works where that journey is measurable. Marketing mix modeling looks at totals over time and includes offline spend and external factors such as season and current events. Attribution zooms in, modeling zooms out.
How much data do you need for marketing mix modeling?
As a rule of thumb, two to three years of weekly figures, with enough variation in spend per channel. Without that variation the model has no differences to work with. Anyone who spent the same amount everywhere every month has little to model.
Is marketing mix modeling useful for smaller organisations?
Often not. With one or two channels and a manageable budget you get further with well-built measurement and common sense. It pays off once you run many channels alongside each other, some of them offline, and nobody can say anymore what delivers what.
