Airbnb is only 7 years old but it’s one of the most disruptive and fastest-growing companies out there.
More than 25 million people in 190-plus countries have used its service, helping it reach a massive $US13 billion valuation.
One of the biggest reasons for its rapid growth is its strong data science technology. By collecting and analysing its massive trove of data, Airbnb is able to make better recommendations and match the right people together.
On Wednesday, Mike Curtis, Airbnb’s VP of engineering, shared some of the cool things his team has been doing to go beyond just regular data analysis. Here’s what he highlighted:
- A/B testing: A/B testing is a common method used to find product/market fit. It tests multiple configurations or designs of a product or web site to figure out how people respond to certain products or promotions. At Airbnb, users are exposed to different ranking algorithms or recommendation algorithms, and their behaviour will be tied back to the actual reviews or star rankings they leave to test its effectiveness. “It’s to find out: Did we do a better job matching this person?” Curtis says.
- Natural language processing: Curtis says the nature of Airbnb, where the guest and host have a real life interaction, tends to put some form of pressure to leave better reviews. So its star ratings often get overly inflated and reviews turn out to be more positive. In order to decipher the users’ true feelings, Airbnb has been deploying a natural language processing technology to review the text on the message threads or review boards to lift some sentiment out of it as well. “We’re trying to get a little more behind what the feeling was behind it,” he says.
- Photo analysis: The first form of contact Airbnb makes with its users is photos. The guests ultimately make their decision based on what they see in the photos. So Airbnb has done a bunch of analysis on what photos work the best, what types of photos attract the most clicks, and what features of those photos make them most desirable. Curtis says it’s still in the early stages, but ultimately, it’s intended to create a feedback loop so hosts can get the best photos for their listing. Curtis says it could also lead to automatically recommending Airbnb’s free professional photography service.