Peer-to-peer lodging marketplace Airbnb says its tenants make healthy profits on average, but supply and demand tend to skew year-over-year income. Seasonality is partially to blame — bookings of beach homes of Florida beach homes will inevitably slow down during the winter months, as will Indonesian bed and breakfasts during monsoon season — but so are one-off concerts, conventions, sporting events, and other not-so-predictable draws. Completely mitigating those fluctuations is likely impossible, but Airbnb’s unveiling a new tool does its best to intelligently anticipate them.
It’s called Price Tips and it starts with massive store of data, VP of Engineering for Airbnb Mike Curtis said on stage at the company’s Open Air event in San Francisco on Thursday. “Aerosolve,” machine learning tools of the company’s creation, pore over “hundreds of signals” — factors such as listing type, location, availability, and special amenities — to settle on the best price for a property during any given day, week, or month. It’s simple on the front end: Hosts see a calendar of dates and the price they’ve set for their listings, and Price Tips highlights in red the ones it deems a little too high or low. If the price is right — low enough to attract renters but not so low as to leave money on the table — the numbers will turn green.
Deciding on a price “has been a game of looking at hotels and seeing what they charge,” Curtis said, but Price Tips theoretically eliminates that guessing. “If you have an apartment in San Francisco, it could be worth twice as much what it is normally when Dreamforce is in town and the entire city is booked out.” In Airbnb’s tests, properties were four times more likely to be booked if a host sets their rate within 5 percent of Price Tip’s recommendation. And what’s good for them is good for Airbnb — the company takes a 9 to 15 percent cut of all list prices.
Price Tips isn’t the first attempt at applying computer intelligence to Airbnb listings. Beyond Pricing, a third-party startup, claims to help hosts make 10 to 40 percent more from their properties by recommending prices based on trends. Everbooked takes a similar approach but automates the process, changing listings in real time to reflect “algorithmically determined” demand. But both services have serious drawbacks — they leverage a 1 percent fee on rentals (Price Tips is free) and lack access to Airbnb’s vast trove of internal data.
Price Tips isn’t Airbnb’s only play at expansion, but it might be the company’s most forward-looking. Overuse may have diluted the term “machine learning,” but the underlying principle — computer programs able to adapt to new information independently, without being hand-held or explicitly guided by a watchful programmer — promises true artificial intelligence capable of easing nearly all of life’s burdens. Imagine diagnoses by a computer with access to the world’s medical data at its digital fingertips, or a virtual concierge that purchases, packages, and ships gifts it knows your family will love just in time for the holidays. Airbnb’s providing Aerosolve free of charge under an open source license.
Artificial intelligence may not be as flashy as, say, a floating houseboat on the river Thames, a gondola dangling 9,000 feet in the air, or properties in the long-verboten island nation of Cuba, but it’s undeniably Airbnb’s most important area of interest. Here’s hoping the company continues to devote development time and resources toward it.