Large-scale Business Forecasting at Uber: Successes and Failures in Adapting to a Changing Business
Can a machine learning framework keep up with a business changing as rapidly as Uber has over the last 4 years? The core mission of Uber’s finance data science & machine learning engineering team is to provide high-quality forecasts of key business metrics for executives to use in planning upcoming growth and optimizing investments. As Uber has evolved, so too have those algorithms, becoming more complex or simpler as-needed. In some cases, we have failed to keep up with the pace of change. In this talk I discuss where and how we’ve been able to evolve our algorithms to drive enormous impact. I also share my failures — both technical and organizational — so that others can learn from them without the pain of experience.
Emily Bailey leads data science for Uber’s rides finance organization. Her team focuses on accurate short and long-term forecasting and optimizing Uber’s investment strategy. Prior to joining Uber in 2015, she solved forecasting problems at cleantech startup Opower (acquired by Oracle in 2016). Emily's educational background is in Economics (Duke University, BS) and Computer Science (Columbia University, MS).