For a company that may not have technology at the forefront of its offerings, Booking.com is a sleeping giant when it comes to developing and scaling innovative customer service solutions.
“Before we called it AI, we were doing AI stuff,” explains Stuart Frisby, director of design for Booking.com.
Back in the 2000s, the accommodations platform realized they knew enough about customers to begin tailoring content towards them. At the time, it was called personalization, but it resembles what a lot of AI models are working towards right now. Back then, it involved ranking results to offer dynamic packages that better suit a potential customer.
Fast forward to modern times, and the team can safely call it AI, though not before saying it’s all in the name of the customer.
“We always say we are a customer-first company but we innovate through technology,” says Gillian Tans, co-founder and CEO of Booking.com. “We make sure to keep delivering through mobile, but now you see machine learning and AI.”
“Booking.com tests around the question of ‘Where do I want to go on my vacation?’ and strives to make that decision easier for customers,” she adds.
As Tans mentions, the company uses AI and machine learning, among other techniques, to offer customers the best experience possible when booking a vacation. Many companies nowadays have some form of AI to aid in customer service, so that revelation is not shocking. But Booking.com has served over one billion customers and continues to innovate AI largely in-house, as opposed to leveraging third-party data to implement new services.
This type of curated knowledge is shown front and centre in Booking.com’s new Booking Experiences predictive learning feature. If customers book travel to one of the cities undergoing testing, such as Amsterdam, Paris or London, they can download a QR code that will grant them access to 30-50 attractions in the city. That’s it—no gimmicks or holdups. The code is associated with a credit card and customized to whoever is travelling, meaning it will purchase one ticket to the Louvre if travelling alone, or five tickets to the London Eye if travelling as a family. The code even grants priority access in some cases.
The technology uses machine learning to read millions of reviews on the site and sees what travellers mention most often as being favorable (or not so favorable). This is combined with the customer’s previous travel information as well as location data to create a list of possible things to do. Booking Experiences presents problems if it’s going to expand to new cities.
“Some of the complexities of scaling out are that it’s a super-fragmented industry, and finding out how we integrate with the inventory and pricing systems is sometimes problematic,” says Frisby. “Finally, we need to see where we add the most value. Is it the big attractions, or smaller ones we want to point tourists in the direction of?”
The Booking Experiences process is done all on a smartphone, the processing method of choice for many Booking.com users. Two out of every five transactions made through the site are done on a mobile device.
Booking.com also employs a chatbot for customer service needs. Chatbots are not revolutionary in the industry, but the site has learned to master the technology by leveraging the platform to serve millions of customers while still upholding their standards of service.
“We have been successful with a strong customer support function, and a strong consideration in building the bot is making sure it mirrors the positive qualities of that customer service experience,” says Frisby. “We just have to figure that out, then roll it out in 40-odd languages.”
The chatbot will aid a customer in figuring out anything they may need, whether it relates to finding wifi or parking. If it can’t answer a question, a real-life rep will take its place.
“Most of the complexity around AI and the chatbot is not related to the technology, but the narrow scope of things our business does,” says Frisby. “The reason we lean to doing everything in-house because no one in the world has a better data set then us.”
The company does call for a little outside help now and then when it comes to innovation, as Booking.com purchased (or acqui-hired as it might be described) the Israel-based Evature. Evature develops natural language and chatbot related technologies, and a Booking.com spokesperson said the purchase will “support research-and-development efforts generally, but also specifically in the area of deep learning and artificial intelligence,” all part of Booking’s “ongoing dedication to testing new areas of technology innovation.”
Looking past chatbots or machine learning-fuelled mobile offerings, Booking.com still tests the relatively-old-fashioned way: live on the site itself. Except these tests are run at a scale almost any other company might see as overbearing.
At any given time on the live site, there are over 1,000 tests being run to gauge how consumers interact with small changes. On the first visit, a button may be blue instead of red; another click and you may see the call-to-action closer to the review rating than it was before. Reactions to slight tweaks are all tracked by machine learning and analyzed at the highest customer service levels. These changes may result in miniscule engagement differences, but when a company is dealing with bookings by the billion, miniscule adds up real quick.
“That’s how fast innovation goes at Booking.com,” says Tans. “I think we’re one of the biggest testing companies in terms of how we do audience behavior experimentation.”
In terms of how far innovation reaches, the company is always looking forward. Frisby explained how Booking.com constantly tried to be ahead of the curve, whether it be the possible use of VR or the incorporation of virtual assistants.
“Our principle has always been that we’ll put our services wherever customers are,” he says.
Does that mean customers may at one point be able to say, “Alexa, I have a $1,500 budget, set up a vacation for me next week!” and Booking.com will do the legwork?
“I would fancy our chances at being able to do a pretty good job at that,” says Frisby. “We’ve never tried, but I would be first in line to do it.”