adv

Airbnb uses big data and machine learning to enhance user experience

Airbnb uses big data and machine learning to enhance user experienceAirbnb is an online lodging rental platform for home-sharing hosts and visitors. The company has been instrumental in using big data and machine learning to create better user experience. Hosts can maximize their rentals whereas visitors can find their most preferred accommodation.

“Airbnb is a data informed company. We think data is the voice of our customers,” said Ricardo Bion, data science manager at Airbnb who spoke at the Cloud Expo Asia in Hong Kong end last month. “From the moment customers visit our website, search and book lodgings, review and share information about their trips, we have been collecting a lot of information about our customers.”

The company logs more than 15 billion events on its website and handles over 15 petabytes of data everyday. More than 150 people are working on data science at the company. Over 3.5 million lodging listings in 191 countries worldwide are posted on the Airbnb website.

Big data and machine learning have helped Airbnb to design new services or features for home-sharing hosts or visitors. By collecting and analyzing its massive trove of data, Airbnb aims to make better recommendations and match the right hosts and visitors together.

Bion shared two examples of how big data and machine learning are used to enhance user experience.

Price tips for hosts

The first example is about how Airbnb provides price tips to hosts who want to rent out their homes.

Hotels often change their room prices based on factors such as seasonality, events calendar and alike. Unlike hotels, Bion noted most of the hosts use the same price for their homes throughout the year.

“We explore ways to help them get more bookings and make more money,” he said. “They could charge higher prices when there is a high demand and lower prices when the demand is low.”

Airbnb built a pricing algorithm model based on an open-source machine learning tool Aerosolve. This model predicts the likelihood of a lodging listing getting booked for a certain day at a certain price based on factors like demand, listing location, listing type and quality. Airbnb can then use this model to recommend how much hosts should charge every night for their listings on its website.

In addition to this machine learning model, Airbnb also created a user interface component for hosts. Through the user interface, hosts could tell Airbnb the minimum and maximum price they are willing to have their listings booked as well as how often they would want their listings to get booked. Besides giving their recommendations on their lodging prices for each night, hosts could allow Airbnb to take over and automatically adjust the prices for them.

Understanding host preferences

The second example is about understanding host preferences to predict if hosts accept visitors’ accommodation requests.

According to Bion, most hosts in big markets generally prefer higher occupancy of their homes whereas hosts in smaller markets prefer a breather between stays. Sometimes hosts in the same market may be different in their preference.



Comments