Sourcing Data Scientists and Market Mix Modellers
Hiring for measurement is harder than hiring for almost any other analytics discipline. Here is what actually works.
Every measurement team eventually hits the same wall: there are not enough good MMM people, the ones who exist are paid by Meta or Google, and the candidates who come through standard data-science pipelines often cannot tell you what adstock is.
Why the hiring market is broken
Three things make MMM hiring uniquely hard:
- It is a hybrid discipline. A good modeller needs econometrics, marketing fluency, and the patience to clean a decade of media spend data. Most data-science programmes train two of the three at best.
- The training pipeline is private. The big consultancies (Brightblue, Gain Theory, Analytic Partners, Ekimetrics) used to train the market. Many of those programmes have shrunk or moved in-house at the big platforms.
- The job title hides behind dozens of others. "Marketing science", "media analytics", "effectiveness", "commercial analytics" - same job, four LinkedIn searches.
What to look for instead of credentials
Stop asking for "5 years of MMM experience" and start screening for:
- Statistical maturity. Can the candidate explain multicollinearity, regularisation, and why R² is a lousy single metric for an MMM? If yes, the rest is teachable.
- Marketing intuition. Show them a channel-level response curve and ask what is wrong with it. The good ones will spot the missing seasonality or the implausible saturation point in 30 seconds.
- Communication. The output of MMM is a conversation with a CMO, not a notebook. Candidates who cannot translate a coefficient into "spend less on this, more on that, here is the risk" will never get their work used.
- Engineering hygiene. Code in git, models in a reproducible pipeline, no one-off spreadsheets. This filters out 60% of senior candidates and is worth every rejected CV.
Build, do not buy, the middle
Senior modellers will always be expensive and scarce. The leverage is in building your own juniors: hire smart econometrics or applied-stats graduates, give them six months of structured exposure to the codebase and the client conversations, and you will outproduce a team that only hires laterals.