Marketing Mix Modelling Comes Back
Marketing mix modelling never stopped working. It stopped being used, because it cost a great deal, took months, and answered a question the business had already moved past by the time the deck arrived.
Attribution filled the gap with something faster and narrower, and the industry spent a decade optimising the channels it could measure at the expense of the ones it could not.
Why Marketing Mix Modelling Is Returning
Because the two practical objections have weakened. Running the models is now cheap enough to do monthly, and the data preparation that consumed most of the budget is largely automatable.
And because the alternative got worse. Identifier loss, walled gardens and platform-reported numbers made click-based attribution progressively less trustworthy, right at the moment the older method got faster.
What It Is Good At, And What It Is Not
It is good at the question finance actually asks: if we move money between channels, what happens to sales. It handles television, print and out of home, which attribution never did honestly.
It is poor at the tactical question a performance team asks on a Tuesday, and it will not tell anyone which creative to run. Treating it as a replacement for daily optimisation produces disappointment on both sides.
Attribution answered the question we could measure. Mix modelling answers the one the business was asking.
The Indian Consideration
Retail distribution. In a market where a large share of sales runs through trade, a model that ignores distribution shifts, price changes and competitor promotions will confidently credit advertising for a warehouse decision.
Which means the model is only as good as the inputs the client is willing to share, and the hardest part of a modern modelling project is still a conversation between marketing and sales about whose numbers go in.
Related: whether AI can fix measurement.
More in The AI Shift.