Microsoft Harnesses Machine Learning for Sourcing

"The only way that we can serve our travelers is by turning them into data," travel management provocateur Eric Bailey told BTN. Bailey is better known as Microsoft's director of travel, venue sourcing and payment. He has pretty much ditched travel policy for Microsoft's 75,000 travelers in exchange for a traveler-centric approach. Machine learning has emerged as critical to the effort. "We can't know all of our travelers," he said. "What we can do is take their data and turn it back into travelers—or at least types of travelers. That's how we're using machine learning now."

Microsoft travel manager for compliance and analytics Marta Rodriguez explained more: "Machine learning allows us to cluster travelers and build persona types based on behavioral features: bookings, expenses, history, everything we know." That gives the Microsoft travel team a solid foundation for setting up its travel solutions and defining its sourcing strategies based on those personas and the needs of those types of travelers.

Rodriguez said Microsoft has four different traveler types. Letting machines do the number crunching is making the travel team "very efficient listeners." She said, "We have a lot of data streams that we can integrate together, and we can really get to the key messages that travelers are throwing to us."

Using Emirates as an example, Rodriguez said, "If we see that [the data] of very frequent travelers who have a dominant preference for Emirates is telling us certain things, we can take that information back to Emirates for more strategic sourcing and business decisions." Microsoft is looking at real patterns of a group of high-value travelers attached to actual dollars, and that gives the company a stronger negotiating position than anecdotal feedback.

Having that kind of data for the travel program overall has offered a totally different lens through which to view the structure the program, as well. Microsoft actually has hard data that shows its road warriors always book online; those travelers want to make changes on the go and want to be able to self-serve a lot of those changes. "Their problem is that it's really hard to make changes on the go," said Rodriguez. "That's completely different from other types of travelers, but we need to figure out how to solve for it." Infrequent travelers need more recommendations and the company is using machine learning to solve for that as well.

While Microsoft is well known as a first mover in pushing smarter data through its booking tools and increasing relevance and efficiency for its travelers, Rodriguez, who formerly worked for Pi Travel, sees sourcing innovation as the broadest machine learning opportunity for travel managers right now—if they'll take the leap.

"As a travel manager, you [need to] understand the potential of machine learning for your own benefit," she said. "People need to embrace it, instead of saying, 'Oh, you're doing that because you're Microsoft,' or, 'You're doing that and that's really cool, but I don't know anything about AI or machine learning and I'm too scared to dig in.'"

Rodriguez's former colleagues at Pi agree that while personalizing the booking and expense process has been a higher-profile effort, managing business intelligence on the supply side needs to get more attention. "Suppliers have a lot of data at their disposal, and travel managers have to keep up. Otherwise, they continue to get their data from suppliers and it's like the fox guarding the henhouse," said Pi managing director of North America Tom Tulloch. He said more companies are getting smarter with their travel data, pulling in additional sources beyond travel management companies, card and expense. And the good news, he said, is that travel data is getting more accurate and more real time.

"A lot of people think they have a data problem, but I'm not sure that's true. We all have lots of data. It's combining map and matching and enhancing that data in a way that makes it easy for people to consume, understand and go do something about. We need to focus more attention on that in order to effectively manage the travel category."

Microsoft sees the two efforts as sides of the same coin. The more the booking and traveler apps can help the company understand their travelers, the better data they have to understand what and how to source, not just by total volume but also by the travelers' relative value to the supplier.