FROM THE MODELLING DESK
Measurement, budget allocation, and Bayesian MMM - written for the people who own the number.
Last-click attribution hands the credit to whoever touched the customer last. That is rarely the channel that created the demand, and branded search is the biggest beneficiary of the con.
Read the post →Most scaleups have two or three years of weekly data and a dozen channels. Without priors, a model will happily hallucinate certainty it has not earned. Priors are the guardrails.
Every channel has a point where the next pound buys less than the last one did. Most teams find it by overshooting. A response curve finds it before the money is gone.
There is no click to track when someone hears your ad on a podcast. Good news: you never needed the click. Aggregate data measures offline channels better than promo codes ever did.