United Airlines received the U.S. National Safety Council’s 2016 Green Cross for Safety Innovation for its data visualization project, which relies on data mapping to discover patterns and reduce injuries. Michael Quiello, United Airlines’ vice president of corporate safety, talked with Lee Ann Shay about how it turned volumes of data into easy-to-use, actionable information with the help of the University of New Haven.
What gave you the idea to pursue the visualization projection and contact the University of New Haven?
I wanted to understand the data—I’m a data guy. When I came to United (from Delta Air Lines), I found that it had reams of data. But how do you aggregate and analyze that data? I was talking with J.J. DeGiovanni [managing director of ground safety] and said, “Let’s see if we can develop a program where we can take reams of data, map it and visualize it so that anybody from the CEO to the front-line person can look at it and say, ‘that’s what our problem is.’” It also needed to be multilingual and interactive. That was the genesis. J.J. and I have been working on this about 2.5 years.
I’m on the board of the University of New Haven in Connecticut, and I was touring the School of Criminal Justice. The school had just gotten a Department of Justice grant to do open-source data collection and analytics—and geo-map it to the world to understand where hot spots are. They were looking at things such as gang violence. They showed me a program where they take a city and add a sliding scale on the bottom—so as you move it, you can see where crime moves. It’s a very robust data program, and I asked if we could use this for safety. They said, “This is a criminal justice program,” and I said, “This is a data program.” You can put any data on top of this and get a view of the world.
So we started the program to geo-map the United system and then put safety data on top. We take four University of New Haven interns per semester to help. When I was in school, I had an engineering professor who said, “Look at the map and anytime there’s an accident, put a pin there. Color-code your pins and when you see clusters, there’s something wrong, whether bad design or the wrong speed limit.” That always stuck in the mind, so when J.J. started developing this program, he put electronic pins on the map, and all of a sudden we started seeing clusters. Once we were giving Jeff Smisek (former CEO) an overview of Chicago O’Hare, and we were looking at on-time by gate. He looked at the map and said, “What’s going on at B24?” It showed 25% on-time at that one gate.
B24 was a regional gate used for close-in routings, such as to Minneapolis, so when we get weather here in Chicago, it affects them, too. Once we understood the problem, we re-timed those flights.
How did you start?
The challenge we had was deciding whether we needed to build an enterprise-wide database. Once you do that, it’s a huge project for the company. J.J. came up with a process where you just structure the data and then import it in a structured format so you don’t need an enterprise database. We brought multiple databases from across the enterprise, put them in a data warehouse and mined it across. We took legacy databases and cross-pollinated them with what’s a variable. We put everything from aircraft damage data, injuries, ops analysis, LOSA (line operation station assessments), workers’ comp and audit data history into a data warehouse, use statistical trend analysis and put a data visualization overlay on top. Then you can break it down to stations or divisions. We’re now working on how we become more predictive. We just hired an organizational behaviorist and a master statistician.
How are you going to make it more predictive?
From a ground safety standpoint, we’d like to get to the point where a station manager can type in all of the variables for the day—bag loads, passengers load factors, number of passengers, gates out of service, meteorological data, . . . hit enter and see the threats. So, it’s a threat error-management tool. If the wind is 30 kt., put your high-wind plan into effect. We want to give the station that has a snow event every five years the benefit of our experience in Chicago, which has a snow event every five days in the winter. How do you electronically move which information around so that everybody learns from others’ experiences? We’re in a variable business—70% is fixed and 30% of the influences on the operation are variable: How do you take the variability out of your business? That’s the key right now.
What are the visualization project’s impact on safety?
There has been a significant reduction in injury rates, which we weren’t focused on before. Our charge is to send people home the same way they came to work, save lost work days—and to bring them back to health when they do get hurt. Ground safety workers comp and corporate medical all report up to me, so I have the charge of preventing the injury, managing the injury and returning the employee back to good health.
What additional real-time data sources are you adding?
We’re going to start overlaying a lot of the variability of the operation, such as meteorological data. So when you look at predictability and you throw in meteorological data or ops tempo or arrival rates at Chicago O’Hare—any variable that affects our operation—how is that driving behavior? Are people rushing because there are thunderstorms? We want to be able to say, when this and this happens, you can expect this. Just like when the weatherman predicts an 80% chance of rain, what they’re saying is, when all of these elements are lined up, 80% of the time it will rain. You always want to put the risk model on top. Risk for us is how you get from reactive to predictive.
People have to take ownership of data. Is it a cultural shift?
We’re trying to push accountability and responsibility as far down to where it happens—to the front line—as possible. Being able to portray things in the granular level that we do by gate and by shift, that accountability doesn’t get upwardly delegated—it gets pushed to the front line, to where the lead who is responsible for his or her people can manage it—and not just say “That’s the cost of business.” We’re managing to zero—zero injuries, zero lost days, zero damages, zero lost resources. How we get there is through programs like this.
Are all of United’s divisions on this?
All of our operational divisions have adopted this, and we’re helping them put their data on top of the geo-mapping. We also have non-operational divisions, like marketing, that are looking at this and figuring out how to put your high-value customers and where they’re flying on top of gate delays, for example, and understanding how you can be proactive in preventing that. We’ve been working with divisions—we’ll help them start their first project—and then they create their own team, manage their charts and mine their own data.
There’s one common data lake that each division can use?
Exactly. We’ve been the initial catalyst to get them going. We seed it, and then they take ownership. We’re trying to build a model that is intuitive and actionable. IT supports us in the background, but they don’t build this—their interaction with us is negligible.
This is a multidisciplinary process. Is it leading to more collaboration?
Absolutely. We didn’t know what we didn’t know. When you get into the granularity of things, you understand what’s really driving the operation. If we were on a production line and we had this much variability, you would say, “What’s going on?” We have the same thing. Why do we have so much variability with like processes?
Do you have advice for people who want to integrate data from multiple sources and visualize it?
First have a corporate mandate to do it and understand the value of it—from the people, operations and financial side. Once you have that commitment, bring in as many people as you can across disciplines to say, “This is what I need.” We ask people what they need—we don’t tell them. Make it as simple as you can. Don’t get bogged down in complexity. As complex as this looks, it’s a simple program from an IT perspective.
What’s the cost to build and maintain this?
The cost to build it was negligible— the resources we already had plus the interns. Find a good university partner like the University of New Haven that is willing to work with you. This was off-the-shelf software.
What is the cost of poor safety?
You can quantify safety in a lot of ways. You can quantify it in workers comp days. I can tell you the cost of an aircraft’s damage or its out-of-service days. I can then connect the out-of-service aircraft to missed connections, and that creates so much downstream work for our employees. But the most important thing is that our people are not injured; that is our highest priority.
United on June 21 said it expects to save $300 million from improved reliability. How is this effort contributing to that?
If you look at the number of injuries that are down in airline operations, that’s significant. On the reliability side, the things that we’re driving from this department are fewer aircraft damages, and so this program has a direct effect on aircraft reliability.