AI is coming for MMM jobs
We recently interviewed a candidate for MMM role and were shocked to hear about their bloated MMM process.
I have some good news and bad news.
Good News – AI will cut a lot of fluff around MMM process.
Bad News – This will not augur well for analysts working in these bloated systems.
For years, some parts of the industry became bloated around processes that frankly should never have required so many layers in the first place.
– One team to clean the data.
– Another team to “run the model” by pushing a button on a platform, while having little understanding of what is actually happening under the hood.
– Another team to build decks and beautify charts.
– Then a project manager whose primary role is to sit between the client and the actual analysts.
End result?
15 people working on a dataset that rarely exceeds:
– 50 columns
– 250 rows
Yet somehow the project takes:
▪️3 months to deliver
▪️$150K – 200K per model
This operational fluff became normalized.
And in many cases, clients unknowingly paid for organizational inefficiency. From our very inception, Aryma Labs operated like a special ops team. MMM does not need a battalion.
AI and automation is going to compress a huge part of this layer further.
– The analyst who only knows how to click buttons on a black-box platform is at real risk.
– The analyst who only beautifies slides is at risk.
Because if you do not understand how the model actually works underneath, then you are effectively competing against an automated open source library.
And the reality is:
An automated system can already generate 1000 MMM models faster than a human team.
The real skill is no longer generating models.
The real skill is:
▪️Knowing which model is wrong
▪️Knowing what parameters to tweak
▪️Understanding why coefficients flipped signs
▪️Identifying when adstock assumptions break
▪️Spotting hidden endogeneity
▪️Knowing which business story is statistically believable
And guess what, the smart analysts are realizing this. The above are few reasons why they want to join Aryma Labs.
Ironically, for MMM analysts, the antidote to AI may actually be going back to the old ways.
Not by learning how to Claude code a MMM. Because honestly, almost anybody can do that now.
The safer long-term skill is going back to fundamentals:
▪️Understanding regression deeply
▪️Understanding adstock
▪️Understanding multicollinearity
▪️Understanding causality
▪️Understanding parameter tuning
▪️Understanding business constraints
▪️Understanding why the model behaves the way it behaves
The future analyst will not be valued for producing code. They will be valued for the inference skills – translating statistical speak to business speak and vice versa.
For those interested, our Foundational and Advance MMM courses equips analysts with first principles MMM skills. Link in comments.
And yes, if you already know how things work under the hood of an MMM, we are hiring !!
