My Starter Project on the Lyft Rider Data Science Team

Summary

The article outlines a data science project undertaken by a new hire at Lyft, focusing on the Rider Experience Score (RES) tool to analyze the long-term effects of rider experiences on retention. It discusses the challenges of estimating causal effects without A/B testing and introduces the Augmented Inverse Propensity Score Weighting (AIPW) methodology to mitigate bias in observational data. The author shares insights gained from working with internal data sources, selecting confounders, and collaborating with colleagues to enhance the RES pipeline, ultimately contributing to improved rider experiences.

Key Learnings

• 1Understanding the limitations of A/B testing for long-term effect measurement and the necessity of alternative methodologies like AIPW.
• 2The importance of identifying and managing confounders in causal inference to avoid biased estimates.
• 3How to leverage machine learning models to estimate treatment effects in complex scenarios.
• 4The role of collaboration and internal resources in enhancing data science projects within an organization.

Who Should Read This

Data Scientists with experience in causal inference methodologies looking to enhance their understanding of practical applications in a real-world setting.

Test Your Knowledge

Topics