{"id":2209,"date":"2015-11-18T22:12:27","date_gmt":"2015-11-19T03:12:27","guid":{"rendered":"https:\/\/digital.hbs.edu\/platform-digit\/submission\/airbnb-leverages-machine-learning-to-maximize-host-and-company-revenue\/"},"modified":"2015-11-18T22:12:56","modified_gmt":"2015-11-19T03:12:56","slug":"airbnb-leverages-machine-learning-to-maximize-host-and-company-revenue","status":"publish","type":"hck-submission","link":"https:\/\/d3.harvard.edu\/platform-digit\/submission\/airbnb-leverages-machine-learning-to-maximize-host-and-company-revenue\/","title":{"rendered":"Airbnb Leverages Machine Learning to Maximize Host and Company Revenue"},"content":{"rendered":"

Founded in 2008 by Brian Chesky, Joe Gebbia, and Nathan Blecharczyk, Airbnb has grown its lodging rental platform from a small Y Combinator startup to a $20 billion dollar unicorn. Airbnb\u2019s website allows individuals to list their housing units \u2013 either an extra bedroom or the entire space \u2013 for others to rent. In exchange, the company captures a portion of the value (via a 3% host service fee and a 6-12% guest service fee) for setting up the transaction, managing the payment process, providing customer service and insurance, etc. Given the scale economies they capture via their asset-light model, the company\u2019s goal is to maximize its revenue stream both by increasing the number of units available for rent on the platform and by ensuring that those units are priced to maximize the likelihood of being rented (as opposed to losing potential customers to hotels) \u2013 the latter of which is certainly aligned with users\u2019 motivations.<\/p>\n

While Airbnb has spent a lot of marketing money to increase the number of units available on its platform, the company has also begun hiring technical talent (e.g., data scientists) to analyze how the units already on the site could be better priced to satisfy all three stakeholders \u2013 the hosts, the occupants, and Airbnb. Historically, Airbnb hosts have priced their properties based on nearby comparables. However, this process is quite inefficient, as it not only forces users to spend time finding \u201ccomparable\u201d properties, but also, the person(s) establishing to \u201cbaseline\u201d that others use may have over- or underpriced the property, based on their financial goals and motivations. To help minimize these inefficiencies, Airbnb has recently launched Price Tips \u2013 a pricing guide that provides pricing suggestions for hosts, which helps hosts capture more revenue via increased likelihood of their unit being rented, while also providing Airbnb with more value capture by taking a portion of the rent. Airbnb\u2019s Price Tips functionality is built using Aerosolve \u2013 the company\u2019s open-source machine learning pricing tool.<\/p>\n

Aerosolve and Price Tips evolved (and continue to evolve) using the research from Airbnb\u2019s data scientists. These scientists have leveraged big data and statistical techniques (e.g., regression analysis) to help determine how to appropriate price a unit, based on supply, demand, and other factors. Further, via Aerosolve\u2019s machine learning techniques, the increasing points of data (Airbnb has already used over 5 billion points for their algorithm) will continue to make the pricing suggestions smarter and more dynamic. Specifically, Airbnb uses some of the following categories (not exhaustive) to determine pricing:<\/p>\n