Artificial Intelligence at Airbnb – Two Unique Use-Cases

Ryan Owen

Ryan Owen holds an MBA from the University of South Carolina, and has rich experience in financial services, having worked with Liberty Mutual, Sun Life, and other financial firms. Ryan writes and edits AI industry trends and use-cases for Emerj's editorial and client content.

Artificial Intelligence at Airbnb

Airbnb started with two roommates, three air mattresses, and expensive San Francisco rent. In late September 2007, Joe Gebbia hatched the idea to Brian Chesky: rent out air mattresses to young designers in town for a conference. They would include internet, a workspace, breakfast, and a roof over their heads.

They made $240 that weekend. The concept only grew from there.    

Since 2007, Airbnb has evolved beyond its initial air-mattress-as-a-b&b model. Today, Airbnb connects guests with hosts who open their homes and properties to them in exchange for a fee. Airbnb then earns money from booking fees charged to both hosts and guests.

Airbnb went public in a December 2020 IPO, raising $3.5 billion. Today, its market cap exceeds $95 billion. Airbnb reported revenues exceeding $3.3 billion in 2020. According to its 2020 10-K filing, the company now boasts some “4 million hosts who have welcomed over 800 million guest arrivals” since its 2007 founding. Those hostings span some 100,000 cities across the globe.  

Like much of the hospitality industry, Airbnb has suffered from travel disruptions during the COVID-19 pandemic. Revenues dropped 29.6% year-over-year for 2020. Business has snapped back in 2021, however, with revenues growing 88% year-over-year for the six months ended June 30, according to Airbnb’s 10-Q August 2021 filing. Although Airbnb posted some quarterly profits pre-pandemic, it still hasn’t posted a full-year profit. 

In this article, we’ll look at how Airbnb has explored the applications of AI for its business through two unique AI use-cases:

  • AI-Enabled Guest Background Checks – how Airbnb uses AI to vet guests before they stay at host properties 
  • Smart Pricing Algorithms – how Airbnb uses a machine-learning algorithm to help its hosts set competitive rates that maximize occupancy rates

We begin by examining how Airbnb uses AI to perform background checks on its guests in order to ensure the safety of its hosts and their continued participation in the company’s split-fee revenue formula. 

AI-Enabled Guest Background Checks

Since its founding in 2007, Airbnb’s hosts have logged hundreds of millions of guest arrivals. Not all of them have been ideal experiences. In January 2019, a woman in London-suburb Kensington rented her five-bedroom luxury apartment valued at £2.5m (US $3.45 million) to another woman who claimed she would host a surprise baby shower, according to the London Evening Standard. The surprise was on the host, who stayed on the property in her guest house. Hundreds of partygoers arrived for the party, which damaged the property and unnerved the host so much that she thought she might end up “dead or stabbed.”

An Airbnb spokesperson later told the Daily Mail that the company has “zero tolerance” for this behavior and promised to investigate. Later that year, Brian Chesky, Airbnb CEO tweeted:

A tweet from Brian Chesky, CEO of Airbnb, that says the company will ban party houses
Source: Twitter

With a business model built on earning fees from hosts who rent and guests who pay, calming hosts’ frayed nerves by vetting guests seems to make business sense.  

To resolve its party house problem and calm its hosts, Airbnb looked to enhance background checks by using AI to scan social media, blogs, search engine results, and other public information to search out individuals with indicators that suggested that they might be capable of abusive guest conduct.

The tool they had in development, according to the patent filed with the United States Patent Office, verifies the “identity and trustworthiness” of guests by ferreting out fake profiles that do not reconcile with profiles found online and in other public sources by the tool. Airbnb claims that the tool then considers activity found for the guest on “third-party online systems” and calculates a trustworthiness score. If that trustworthiness score indicates a high-risk reservation, the potential guest is pended for manual screening by a human.

Another patent filed by Trooly, which was acquired by Airbnb in 2017, goes further, describing these sources as:

  • Information on web pages
  • Database information
  • Social media posts 
  • Blog posts 
  • Comments posted by the person online 
  • Company or association directory listings   

“Keeping our Airbnb family safe is one of our top priorities,” the company writes on its website. The company goes on to say that they will conduct background checks on US and India guests within ten days of their reservation.

 The company does not disclose the number of background checks it has executed or the number of guests who have been turned away. However, Airbnb does state that problematic background checks can result in booking delays or even removal. They offer a chance for reinstatement among its guest rolls and clearly state the criteria for appealing a removal.

Smart Pricing Algorithms

Airbnb hosts aren’t likely to be hospitality industry experts. For people who want to make money from their otherwise vacant properties, trying to find the right rate to charge guests involves time, research, and risk. Among their challenges:

  • Competing properties enter and leave the market
  • Demand shifts across seasons and time
  • Special events affect lodging supply and demand

If hosts price the property too high, it sits vacant. If the price is set too low, they lose money. A machine-learning-informed algorithm that helps set prices for hosts seems to make sense and make it easier for hosts to rent their properties.

This AI tool makes business sense for Airbnb too. Due to its booking-fee revenue model, Airbnb follows the fortunes of its hosts. High rates and higher occupancies mean more revenue.

In November 2015, Airbnb officially launched Smart Pricing, which built upon a lighter, less robust version of the tool launched a few months earlier as Price Tips. At the time, Airbnb would not disclose how much additional revenue the Smart Pricing technology might generate. Instead, Joe Zadeh, their Vice President of Product, told CNBC, “Since we have rolled out Price Tips [in June 2015], the precursor to Smart Pricing, 20 million nights have been priced using it and hosts have seen a 13 percent increase in their revenue.”

When it set out to implement Smart Pricing, Airbnb began by talking with humans. The company asked their hosts about their challenges and concerns related to pricing. They soon determined that hosts formed two distinct groups: those who rented their space to earn needed income and those who did so for extra money.

Armed with these very human insights, Airbnb set out to create its AI-based Smart Pricing tool. When Airbnb’s engineers and data scientists designed the Smart Pricing tool, they sought demand-side data like seasonality, unique features of a property, and price, wrote Hector Yee and Bar Ifrach for the Airbnb Tech Blog. They elaborate:

These features interact in complex ways and can result in machine learning models that are difficult to interpret. So we went about building a package to produce machine learning models that facilitate interpretation and understanding.

That package became Aerosolve, a machine learning package designed to interpret complicated data with intuitive models. Aerosolve drove Smart Pricing, a dynamic pricing feature that color-codes a calendar to show hosts the chance of renting a room to guests at different price points.

Airbnb’s Smart Pricing tool went live in November 2015.

A graphical depiction of how Airbnb price tips for hosts works
Airbnb’s Smart Pricing tool color-codes a calendar to show hosts the chance of renting a room to guests at different price points (Source: Airbnb Tech Blog @ Medium)

The Smart Pricing algorithm appears to have grown revenue by 8.6% for hosts while decreasing rates for guests by 5.7%, according to research conducted by a team of researchers from Carnegie Mellon, Harvard, and the University of Toronto and published in the journal Marketing Science in October 2020.

With Airbnb’s split-fee revenue formula, this appears to be a win-win-win situation across hosts, guests, and the company. 

However, the same study also examined differences in daily revenue earned by white and Black hosts. The study found that, while Smart Pricing reduced the $12.16 daily revenue gap by 71.3% among those who adopted the tool, a significant number of Black hosts did not adopt it, which led to a widening of the daily revenue gap across the population of hosts. The researchers  write:

The algorithm, if adopted, can substantially reduce the revenue gap between Black and white hosts. However, because Black hosts were 41% less likely to adopt the algorithm than white hosts, the introduction of the algorithm widened the racial revenue gap.

Airbnb has touted its commitment to anti-discrimination both to Emerj in 2017 and in the press in general. Airbnb made headlines when it created Project Lighthouse, the company’s anti-discrimination team that partnered with both civil rights organizations and data scientists. For now, it appears that Smart Pricing accomplishes its goal of setting competitive rates that drive overall revenue growth. The challenge appears to be getting all of its hosts onboard with the tool. 

Subscribe
subscribe-image
Stay Ahead of the Machine Learning Curve

Join over 20,000 AI-focused business leaders and receive our latest AI research and trends delivered weekly.

Thanks for subscribing to the Emerj "AI Advantage" newsletter, check your email inbox for confirmation.