What if your forecasted uplift isn’t real? In this episode of The Novi AI Roundup, we explore how standard machine learning models can mislead engineers by overestimating production drivers due to bias in the data. Drawing from our URTeC 2025 paper “When Numbers Lie: De-Biasing Horizontal Well Production Datasets with Causal Forecasting”, we discuss how causal AI helps separate correlation from causation, revealing the true impact of spacing, depletion, and geology. From false uplift to masked degradation, this episode is about building models engineers can trust.

This podcast episode is based on the technical paper “When Numbers Lie: De-Biasing Horizontal Well Production Datasets with Causal Forecasting”, authors: Dillon Niederhut and Kiran Sathaye. The paper was presented at URTeC 2025. Download the full paper here.