Writing

Ideas in progress

Practical notes on moving from accurate predictions to better decisions with causal machine learning.

Article idea

Your Best Customers Might Be the Worst People to Target: Using Causal ML for Promotion Optimization

Why high purchase propensity is not the same as incremental lift—and how uplift modelling can separate persuadable customers from those who would buy anyway.

Causal MLUplift ModellingMarketing

Article idea

My Churn Model Had 90% AUC—and Targeted the Wrong Customers

A strong ranking metric can still produce a poor retention strategy when predictions ignore treatment response, intervention cost, and customers who cannot be persuaded.

ChurnModel EvaluationDecision Science