Effectiveness
Last updated 8 August 2026
Do we need marketing mix modelling?
Founder, Front Foot Marketing
The question behind the question
Which MMM should we buy?
What decision do we need better measurement for, and is modelling the right way to get it?
The straight answer
Marketing mix modelling is a tool, not an answer. Start with the decision it's meant to support.
MMM can help estimate what each channel contributes over time, including channels attribution can't see. But it's built on assumptions, history and judgement. It doesn't remove the need for any of them.
The deeper answer
Different measurement answers different questions.
Attribution shows the journeys you can track. Incrementality tests show what one change actually caused. Mix modelling estimates how the whole budget works together, using past patterns.
None of them tells the whole story. The useful question isn't which one is right. It's which one fits the decision in front of you.
If you're reallocating a large budget across many channels, modelling may earn its cost. If you're deciding whether one campaign works, a well-designed test is often faster, cheaper and more convincing.
However, if you're spending under $500,000 on marketing and comms, in one or two channels only, then don't fall for the MMM silver bullet. The returns may not be worth it.
You'd also need at least two years' worth of historical sales data, with three years being better. That gives a more accurate view of impact over time, takes into account seasonality, lags and wider economic impacts, and ensures statistical confidence.
The uncomfortable bit
A model can be very precise and still wrong. And the people building it rarely have to answer for the budget it moves.
What we know
- Where the budget currently goes, by channel
- What attribution and platform reports claim each channel delivers
What we don't
- How much of those claimed results would have happened anyway
- Whether our data history is long and varied enough for a model to separate cause from coincidence
What we need to decide anyway: Whether to commission modelling, run targeted tests, or make the allocation call with the evidence already in hand.
What I'd challenge
I'd challenge the idea that buying a model is the same as improving a decision. A model nobody acts on is just a more expensive dashboard.
When I'd say don't do this
Should you commission marketing mix modelling? Maybe. But not until you can name the budget decision it will change, and who has agreed to act on what it finds.
The Anton Test
If the model told us to move 20% of the budget tomorrow, would we actually do it?
What I'd do next
I'd write down the two or three allocation decisions we'll make in the next twelve months. Then I'd ask which form of evidence each one needs, before talking to any vendor.
The decision
This is ultimately a decision about how much evidence a budget call needs, and what form it should take.
This connects to the broader decision on Marketing effectiveness, if you want the fuller picture.
Keep asking
Related answers
Decision
What happens when the numbers disagree?
Don't automatically believe the number that looks most precise.
Decision
How much evidence is enough?
You rarely get perfect information. The goal isn't certainty. It's confidence good enough to act.
Effectiveness
What should our marketing dashboard actually tell us?
A dashboard should help you make a decision. If it doesn't, it's probably just reporting.
Still working out what the real question is?
Sometimes the hardest part isn't finding the answer. It's knowing which question to ask.