Paul Sangrey

I build Bayesian causal models and forecasting systems. At GoodRx, I architected, built, and run a causal system connecting marketing to revenue — powering planning, optimization, and reporting across the business. It is fully multivariate and has automated experiment ingestion, going beyond state-of-the-art systems like Google's Meridian. I've published in the Journal of Econometrics and presented forecasting-at-scale work at ISF that outperformed both Prophet and neural-network approaches under structural breaks. I have a PhD in economics from Penn, specializing in forecasting and finance.

Before GoodRx, I owned peak and major-event forecasting on Amazon's consumer forecasting team. I produced the forecast for the first major US sales event Amazon ran outside of Prime Day — with no prior history, my forecast beat the accuracy achieved on events that did have history. One memory that stays with me: late nights in March of 2020 when COVID-19 broke the automated forecasting system, and I was one of a handful of people who built the ad-hoc replacement overnight as Amazon sorted out its essential-worker needs.