Theresa Johnson explains how Airbnb is building its next-generation end-to-end revenue forecasting platform, leveraging machine learning, Bayesian inference, TensorFlow, Hadoop, and web technology. The revenue forecasting platform is the core technology in producing the company forecast of all business-relevant metrics such as nights booked, signups, and listings. It has produced the monthly forecast and growth scenarios consumed by Airbnb’s CEO, e-staff, the board of directors, and planning units within the organization.
There are two components to producing the forecast: interpolation and extrapolation. The interpolating component uses statistical models to understand the underlying patterns of the business. The extrapolating component uses the estimated coefficients from the interpolating piece and combines them with domain knowledge to produce an extrapolation about the future. The strength of the forecasting lies in the combination of statistical modeling and human judgment to produce an accurate and reasonable forecast. Airbnb is building a web-based UI for users to make changes of some of the forecast fields during extrapolation. The backend compute engine will recompute the forecast curve to fit in the user input. Users can interactively adjust the metrics and see the impacts and results.
Theresa Johnson is a product manager for metrics and forecasting products at Airbnb. As a data scientist, she was part of the task force and cross-functional hackathon team at Airbnb that worked to develop the framework for the current antidiscrimination efforts. Theresa is a founding board member of Street Code Academy, a nonprofit dedicated to high-touch technical training for inner city youth, and has been featured in TechCrunch for her commitment to helping early-stage founders raise capital. Theresa is passionate about extending technology access for everyone and finding mission-driven companies that can have an outsized impact on the world. She holds a PhD in aeronautics and astronautics and dual undergraduate degrees in science, technology, and society and computer science, all from Stanford University.
Comments on this page are now closed.
For exhibition and sponsorship opportunities, email strataconf@oreilly.com
For information on trade opportunities with O'Reilly conferences, email partners@oreilly.com
View a complete list of Strata Data Conference contacts
©2018, O'Reilly Media, Inc. • (800) 889-8969 or (707) 827-7019 • Monday-Friday 7:30am-5pm PT • All trademarks and registered trademarks appearing on oreilly.com are the property of their respective owners. • confreg@oreilly.com
Comments
Hi Can you please share the slides
Can you please share the slides? The talk was really interesting :)