About three years ago, IBM's travel program stopped to look around and ask itself some big questions. Namely, what did it need to do to lead IBM travel into the future? What would the future even look like? An answer to the latter question wasn't clear, but IBM global category leader of T&E Shawn Busby said it was plain to see that the future centered around what IBM did with its travel data.
Busby served as the keynote speaker at the BTN Group's The Beat Live conference in Atlanta in early October, delivering a talk titled The Data Driven Travel Manager. He described how IBM transformed its travel program from one that made decisions based on static, incomplete data stored in silos to one that leveraged its combined data to identify actionable insights through its cognitive data and analytics platform, IBM Travel Manager.
To some extent, IBM's capabilities are expected; it's one of the global leaders in artificial intelligence and cognitive technology. What's perhaps more impressive—and more applicable to procurement departments not backed by IBM Watson—is the methodology IBM used to get to a new travel platform and how human-centered development and insight-driven procurement practices have changed the role of travel procurement within IBM.
Mind Meld Between Procurement & Research
There's a certain serendipity in the early stages of what would become IBM Travel Manager. Despite IBM's position as a leader in artificial intelligence and cognitive computing, it wasn't a given back in 2015 that IBM would build an internal platform. The travel team shopped around among the usual suspects for an external analytics solution, Busby said, but nothing on the market at the time "blew our socks off" in terms of future capabilities.
Meanwhile, a shift was underway at IBM Research. The 3,000 people there work across IBM's business units to create internal solutions or external offerings, and it has been doing so successfully for more than 70 years. But around 2014, IBM Research began experimenting with agile development and design thinking. This way of working unites subject matter experts across multiple disciplines under a focused mission to put users at the center of the development process.
That same year, IBM formed a partnership with Apple to bring IBM's Big Data and analytics capabilities to iPhone and iPad. According to Guang-Jie Ren, research manager at IBM Almaden Lab, that project sparked the IBM Research team that supports procurement and HR to consider how it could better support employees, particularly those traveling for business. When the travel team sitting in procurement approached IBM Research in 2015 to help it answer those big questions about the future, the sparks were already flying within the research team.
Big Data Is Too Big
IBM's travel program is massive. BTN ranked the tech giant No. 2 in its 2018 Corporate Travel 100 based on an estimate of IBM's 2017 U.S.-booked air volume, which rang in at $430 million. Like any other procurement department, though, IBM's procurement team each year is tasked with removing cost. But how do you effectively remove cost when the information required to make good decisions is not only massive but also fragmented across various spreadsheets and dashboards?
The Global Business Travel Association put a fine point on this struggle in a survey it conducted among 273 travel managers in 2015—around the same time the IBM team began excavating the power of its own travel data. Even though the industry was starting to play around with data visualization solutions like Domo and Tableau, travel manager respondents to the GBTA poll were swamped in incomplete and inaccurate data to feed into such tools.
According to the survey, 57 percent of travel managers pegged "managing data to track program performance and support decisions" as a major drag on their productivity. Nearly half of respondents said they spent too much time manually processing, cleansing and reconciling inaccurate data. More than half said multiple data sources and formats were particularly problematic.
Busby echoed those findings: "How do you take this to the next level to actually provide insights that you can action versus spending hours, days, weeks sifting through line items or dashboards and pulling reports out and twisting and tweaking them to arrive at a nugget that would allow you to say, ‘Hey, I see that something is going on in my program,'?"
2015 was too early for most to try to solve the problem. Big Data was just big, and it was getting bigger and more complex. GBTA survey respondents saw little light at the end of that tunnel, as 57 percent expected to devote more time to data tasks in the future, not less.
However, the IBM Research team and IBM's procurement department and HR department, which manages IBM's travel policy, envisioned a future in which machines would unify IBM travel data and then distill that massive data set with cognitive analytics to support smarter policy, better buying decisions and more effective supplier relationships.
Training the Machines: A Human-Centered Strategy
Design Thinking: What Is It?
Design thinking isn't a new idea. It’s been around since the 1960s and was popularized by, most notably, David Kelley, who founded both the global design and innovation firm Ideo and the Hasso Plattner Institute of Design at Stanford University. But in recent years, the term and the methodology behind it have gained huge traction outside the design world.
At its core, design thinking emphasizes bringing together people from various disciplines to solve a problem that focuses not on business interests but on the end user. It stresses empathy with those who will use what is created and allows development teams to move quickly and identify and solve problems early and often alongside these users. IBM has launched Enterprise Design Thinking, a framework for large, dispersed teams to use design thinking effectively.
"[Using] the old methodology, it takes months or years to see the outcome," said IBM Research senior technical staff member Pawan Chowdhary, who led the development process for IBM Travel Manager. "IBM Design Thinking helps us to get on to the requirements [of a project] very quickly. Our [procurement users] were able to use the tool within a few months." From there, users continued to be part of the process, staying hands-on for functionality tests. "It helps to not only get the buy-in but also they feel important to be able to make changes to the way they approach their customers and their daily activities," Chowdhary said.
"It’s very simple," said Guang-Jie Ren, research manager at IBM Almaden Lab. "If you do a good job in understanding user requirements and do a good job to bring in not only graphic design, the user flow, but also insights at the right moment from data and analytics, then you're going to drive adoption. And with adoption, you drive business value."
These days, travel managers can ask IBM Travel Manager questions and the system will deliver relevant data results. That’s an aspect of the platform, Ren said, that wasn’t originally discussed back in 2015 when this journey began. It came directly out of the user interaction studies conducted in 2016. "That’s that benefit of what we have at IBM: that our research team is working on something together with our business units," Ren said. As IBM Research continues to check in with users, the idea is that new insights and improvements to the platform will emerge.
IBM Research used its design thinking methodology to make users an integral part of the development process. Users, in this case, were procurement category and regional leaders, or in IBM speak, subject matter experts. "IBM is a big spender in travel, so we have a sizable set of users who are responsible for various aspects of travel," explained IBM Research senior technical staff member Pawan Chowdhary, who led the development effort. "We brought them all together; we did not want to have development separate and then pull in [subject matter experts]."
The development team built out a minimum viable product by late 2016, according to Chowdhary, starting with booking, expense, employee and hotel contract data, plus some analytics around airline and hotel that could be tested by users. "The key was to get something partially functioning so users could start playing around and actually help with the design and the development," Busby said.
Users initially functioned as a gut check to validate data being normalized and unified by machine learning and algorithms at work within the platform. Not only is IBM's data fragmented into silos, it also can be duplicated and triplicated among these silos. Think of a hotel booking: captured at the agency, again on a corporate card and a third time in an expense system. Moreover, that same hotel booking might be attached to different hotel name variations among those data feeds. Multiply all those variations by the number of hotel bookings at IBM and you start to understand the magnitude of the data normalization and unification challenge. And that doesn't even account for human variations and errors. "If 10 people stayed at the Marriott Marquis Atlanta," Busby said. "Do you think all 10 of those people would enter in the exact same name into the tool when they expensed it?"
Machine learning can help sort that, but it isn't perfect, particularly in the beginning. It can pull apart names, parse data and match up similar occurrences based on probability, but IBM's users served as fact checkers to take the machine learning and AI to the next level. From there, the development team added data into the platform from different categories and vendors and systems and geographies, validating all of it along the way.
IBM's human-centered approach offers the true power behind the platform, according to Busby, but that human knowledge is captured permanently in the machine. "Once we teach the machine those connections, it's learned [and] we can apply it to any future transactions," he said. "Once you have the data right, what can you do with it, what insights are uncovered through it and how can you build analytics around it?"
Busby said the platform expanded what he thought was possible with travel analytics. "It opened up the aperture in my mind to think what I could do if I had this other little bit of information or if I could bring in context through external [data] sources and be able to understand or predict future patterns. [That's where] technology, and AI specifically, can take you when the data is right."
IBM Travel Manager Now & in the Future
From the combined efforts of the developers and the users came IBM Travel Manager, a dashboard that offers end-to-end visibility of travel program spend, integrating both internal and external data, including from expense, card, hotel, air, rental car and agency. The company rolled out IBM Travel Manager to business managers internally in 2017 and this year struck a joint go-to-market deal with longtime innovation collaborator Travelport. They'll sell the cloud-based tool with all the machine learning and cognitive capability working beneath the user interface.
The data integration and normalization allows IBM Travel Manager to serve as a sort of full travel program audit, Busby said. Users can easily monitor traveler compliance, see whether IBM is getting its contracted rates and services, and conduct gap analyses in price paid versus market rates for hotel and air.
That visibility translates into savings, said Busby. "We're finding traveler and supplier behaviors that we need to fix," he said. "When we do fix them, our rates and discounts become more available. That realizes savings, but it actually helps traveler satisfaction, too."
He said the team hopes to bring in additional data to enrich the platform. For instance, IBM already pulls in benchmarking and fair marketshare data from Travelport. In the future, IBM may bring in Travelport shopping data to understand what wasn't booked. "The agency data comes in and you're also looking at [shopping data] in real time," Busby said. In that scenario, IBM Travel Manager could conduct a real-time gap analysis of what is booked versus what is available.
Busby said the team is considering specific dashboards around things like value-added tax reclaim and travel management company performance. IBM Travel Manager could identify how often the agency is booking IBM's corporate contracted rates versus published rates versus their own rates for both hotel and air. And potentially, why, by looking into capacity or oversold situations.
Changing Perspectives on Procurement
Busby admitted that procurement isn't a favored department within most organizations. Traditionally, it's been "a place that everybody hates to go to because we add processes and make things more difficult," he said. With visibility and actionable insights now at their fingertips, IBM's procurement leaders are being seen in a new light. "Historically, [our internal clients] came to us and said, ‘I need you to reduce your cost to me because I'm challenged in my budget,''' Busby explained. Now, when those same internal clients approach procurement with budgeting difficulties, the conversation can be very different. "We can say, ‘Hey, let us help you help yourself,'" Busby said. Procurement leads can make specific recommendations for ways business units can alter their buying behaviors to find significant savings within their own travel budgets. "We have fundamentally shifted the view of what procurement is," Busby said.