AI wants to know how you plan a trip
At the Expedia product testing lab, facial recognition software gauges travelers’ feelings as they go through the process of booking hotel rooms online. The Swiss cruise company MSC Cruises is starting to use a virtual assistant to answer passengers’ questions. Designers from the boutique design and research firm Gettys Group can show hotel executives new room layouts using virtual reality goggles, so hotels don’t need to build a full-scale model.
Travel companies are adopting artificial intelligence and other new technologies to look more deeply into what customers want and to use that information to find faster, cheaper ways to improve their offerings. And as sophisticated research tools become less expensive and more widely available, even startups in the industry are using them.
Competition in every aspect of the travel sector is extremely stiff, and travel companies “need these mechanisms to reach their target markets,” said Alex Susskind, associate dean of academic affairs at Cornell’s School of Hotel Administration. “They need to know who wants to see pictures before they buy, who decides mainly on price, who likes to speak to someone” and why.
Tammy Snow, the director of user experience research at Expedia, said researchers still use product testing, customer surveys and data analysis. “What has shifted significantly,” she said, “is how we are combining technology and methodologies.”
In Snow’s product-testing lab at Expedia headquarters in Bellevue, Wash., travelers test aspects of online travel purchasing, sitting down in front of a computer in a small room and going through the process of booking a flight or planning a vacation. In a room next door, researchers and members of the product creation team watch a large video screen that displays what the test subject is doing and a scrolling readout of the subject’s reactions, generated by software that plots points on that person’s face and then uses a coding system to identify the person’s emotions.
Alex Hopwood, a director of product management at Expedia, said it helped to get an outside perspective. “We get so close to a feature that we start making assumptions and think we know what the customer wants,” he said. Seeing how a real traveler uses the website and reacts to it “is almost like a slap in the face — in a good way.”
Scott Wainner, chief executive of Fareness, a 15-person startup that lets travelers compare flight prices from a specified departure point to a set of destinations like European cities or beaches, said feedback tools are now affordable even to small travel companies. Internet-based services like Usability Hub and UserTesting.com allow companies to test different aspects of a product, showing two navigation screens to a sample set of users, for example, to see which they prefer. Testers, who have been chosen based on their demographic profile, answer questions online or record themselves on video doing the task and giving feedback.
Through user testing, Wainner said he found that his customers were most interested in learning about the cheapest flight, the shortest flight and the flight that offered the best combination of saving time and money. Displaying search results based on this information, even though the changes were small, made a big difference in sales, Wainner said. He noted that the company “didn’t have to set up our own internal testing facility” to get that information.
Website analytics software has been improving both in the amount of data it collects and in the ways it makes that data understandable and useful. Fareness.com uses Mixpanel, a tool that helps analyze trends like the number and demographics of Wainner’s customers who view company ads, explore the site and book airline tickets. Using Mixpanel, he said, he can, for example, test different messages focused on customers using a specific iPhone operating system or living in a certain region of the country, or compare actions of repeat customers in different demographic groups.
Some travel companies are creating systems that themselves generate research. MSC Cruises is rolling out a virtual assistant in passenger cabins called Zoe that will answer spoken questions. It can improve its answers with continual research based on the interactions it has. Zoe was programmed to answer the 800 most common questions — queries about excursions or onboard restaurants, for example — and variations of those questions in seven languages. The questions were gleaned from staff members and from data collected from the ships’ guest services desks. To work with an international clientele, the system “listened” to 400 people and improved its ability to understand different accents.
If a passenger needs to ask the same question repeatedly in different ways, that indicates the system didn’t understand the passenger’s meaning the first time. Questions that Zoe can’t answer will be sent as a text to researchers who can add appropriate answers to the system’s next iteration. “It’s a never-ending process,” said Luca Pronzati, chief business innovation officer of MSC Cruises. “As more people use the system, the pool of data gets larger and we do better.”
Booking.com, an Amsterdambased online travel agent, is incorporating similar learning into its hybrid chatbot, Booking Assistant. The assistant currently answers about 60 percent of users’ post-booking queries, like hotel checkout times or Wi-Fi availability. It also loops in humans when it can’t find the appropriate answer, the company said. The chatbot improves its repertoire much as the Zoe system does, adding questions that staff have tagged and answered.
James Waters, vice president of commercial operations at Booking.com, said that the company was expanding its offerings beyond hotel room reservations to transportation and leisure bookings, so more customers would need answers to more questions. “The system is a blend of technology and human help,” he said.
Susskind said that traditional marketing in the travel industry had evolved. “You run the risk of not being able to compete effectively in the marketplace without these enhanced approaches.”