This article is part of an Essential Guide, our editor-selected collection of our best articles, videos and other content on this topic. Explore more in this guide:
2. - BI trends for a mobile future: Read more in this section
- How collaborative and mobile BI can work together
- U.S. Xpress trucks more data with big data analytics and mobile BI
- What next-generation BI will mean for CIOs
Explore other sections in this guide:
- 1. - Making the business case for mobile BI
- 3. - The benefits and pitfalls of mobile BI
- 4. - Must-know terms before diving into mobile BI
'Big data' analytics, mobile BI apps help U.S. Xpress truck more dataDate: May 23, 2012
For U.S. Xpress Inc., a trucking company based in Chattanooga, Tenn., the motivation to move to “big data” analytics and real-time business intelligence (BI) reporting was rooted in wanting to get more out of the large volumes of sensor data being collected from the company’s trucks. U.S. Xpress was looking to use the data to enable its fleet managers to “answer very specific, detailed questions” about trucking operations, according to Timothy Leonard, its chief technology officer and vice president of information management.
Leonard says U.S. Xpress collects data from trucks in near real time on everything from how fast they’re being driven to how many times drivers engage in hard braking. From there, the data is sent to fleet managers equipped with iPads, iPhones and other mobile technology so they can make on-the-go decisions in an effort to optimize operations. The trucking company is also able to do geospatial analysis with the help of its information management team.
The key to working with big data analytics software and deploying real-time BI capabilities, Leonard says, is to know the business -- at a deep level. “Become the business,” he advises. “Actually get so involved with the business that you almost learn more than they do on the successes you can bring to the table.”
In this video interview recorded at SearchBusinessAnalytics.com’s recent “Delivering Deeper Insights with Big Data and Real-Time Analytics” seminar, Leonard spoke with Editorial Director Hannah Smalltree about U.S. Xpress’s big data analytics, real-time BI and mobile BI applications and how they all work together to support the company's business operations.
Viewers of the video will learn:
- How the trucking company is dealing with the large volumes of data it’s collecting
- About the three-tier system that U.S. Xpress uses to support its big data architecture, from the process of collecting the sensor data to the reporting stage on mobile devices
- How U.S. Xpress has integrated sensor information from trucks into its analytics program
- The role that mobile devices play in getting real-time BI reports to business users and how U.S. Xpress decided to implement mobile BI applications
- Tips on how to make a big data and real-time analytics project successful
Read the full transcript from this video below: 'Big data' analytics, mobile BI apps help U.S. Xpress truck more data
Hannah: Hello, and welcome. I'm Hannah Smalltree, editorial
director for SearchBusinessAnalytics.com and related sites. I'm here at our seminar on delivering
deeper insights with big data and real-time technologies. Today I'm talking with Tim Leonard. He's
the CTO and VP of U.S. Xpress. Thank you so much for talking with me today, Tim.
Tim: Thank you for having me here.
Hannah: Tim, you lead a unique project to improve your company's fleet operations, which involved big data, real-time, and mobile technologies. Can you give us a high level overview of that project?
Tim: Absolutely. One of the unique characteristics of a trucking company is that there is so much large data that's being collected as well as real-time data, and I think when you look at some of the projects, particularly the large project that we implemented, the real problem that we're trying to do is ensuring that we can take those large volumes of data, and just not calling it volume, but actually answering very specific detailed questions out of it.
What was so unique about this particular project is not only did we take the large data that we're collecting through analytics, but we did geo-spacing of this information to collaborate it with the operations team, and we collapsed the whole concept of transactional information into analytics of information.
Now you don't have to go through transactional and the data warehouse, you actually went to a one stop shopping interface that collected the data from both and pulled it into it. So, we had big volumes of data, but it answered very specific detailed questions.
Hannah: This data is about where trucks are, or what kinds of problems are you running into?
Tim: Actually, that was the unique characteristic of this project. Not only did it geo-fence the data of where it was at by collecting sensor data of the coordinates of the information, but also what was so unique is that we have an interface through DriverTech into the engine data. We collect about 900 to 970 data elements of the engine data and actually bring it into our environment itself.
Yes, information about where the trucks were at, but more importantly, we actually could tell you when you hard brake, how fast you're going. We actually got sensored information about the trucks themselves.
Hannah: A big part of this was integrating that machine data you referred to earlier, that sensor data, into your programs. Can you give us a sense of that big data? How big is it, and where exactly was it stored? Can you tell us a little bit more about that data?
Tim: Well, in the past, we were able to answer very few questions about the mobilization of our fleet itself. Now what we've done is we've turned that around. For example, one of the pointed questions about 17 data elements of those 900 data elements was around idled fuel information.
Most normally, a truck, when it's running, you'll stop at a rest area and it'll idle on its engine to do air conditioner and heating. Before, we only had an estimate about what the fleet actually idling on, which was really in the high 80s. Of course, that's not good to idle that big of a percentage on. So, the information was being dropped on the table, per se, or dropped on the floor.
With this concept, what we did is we collected that data real-time, every 10 to 15 minutes. We brought it into an operational data store. Not only did we encapsulate the data so we could answer those questions about who was the idling the most – gas guzzlers, gas versus gas misers -- but we also created a big semantic layer that pushed it out to iPads, to the fleet managers and their iPhones, and to the Droids. So, they can quickly see which particular trucks were idling the most, so we could help council or do corrective action to get them down.
That was just one small piece of that whole scenario around that big data. Seventeen data elements that were used out of the 900 to solve a very multi-million dollar problem. In the first year we saved well over $20 million just on controlling the amount of fuel being burned.
Hannah: Then you were integrating that sensor data in with your other transactional data. Can you give us a sense of the architectural and technology approaches you use to do that?
Tim: Well, let's start with the first one. When you're looking at the amount of data being collected, if you can think of thousands and thousands of cabs -- the trucks that are out there – passing through 900 data elements that are pretty, with and in of itself, coming into a database environment, we're collecting close to that one to two terabytes of information just off the operational detail.
Every single truck that's on the road passing that much data, every certain amount of five to 10 minutes into our near real-time messaging bus populating up into our ODS, that data is going to grow extremely fast. Now, the good news is, we filter the necessary information that are used for operational reporting, and we filter for analytics that are being used.
Then, the next step was to actually collapse the two together, and that's where you come up with your brand new system called XPM.
Hannah: At a high level, you're integrating all of this sensor data with some of your transactional data. Could you talk a little bit about your architectural approach? You mentioned ESBs earlier. How you're actually doing this, how you're bringing that data together, and then delivering it back in near real-time to people.
Tim: Well, it's a good point you raised. The transactional
systems, when I talked about the 900 data elements, were very specific to the machine data that you
referred to earlier. We also started collapsing our driver information, the number of orders coming
in. Now, you did take that transactional data, merge it with these 900 data elements into that
operational detail storage. But as we started normalizing the data itself, we started building out
the, "How do you deliver this information?"
The architect encapsulated really three tiers to it. The first tier was the near real-time messaging bus. So, we built a bus system that enabled us to get the data, put it into a SOA architects, demand it out or publish it out to the various systems. One data element, think of a driver name, for example, that would come in, that driver name would be spread across three systems at one time.
Instead of inserting it into the transactional, it flows into the ODS, which flows maybe to the warehouse. It would come into the SOA architect and actually be published at the same within milliseconds into three systems at once. So, what that enabled us to do was to collapse the transactional in a warehouse into the near real-time capability. That's just an example of what we did.
Hannah: A really interesting piece of this is how you're pushing this information out to devices in real-time. Can you talk a little bit about the mobile aspect of this project?
Tim: Yeah, it was a really unique story. It almost defaulted to happening that way. I wish I could sit here and say, through the trials and tribulations of how we wanted to push it out, I had that on my thought process. But what happened was, we started dealing a lot with the fleet managers. Let's say there are 1,000 trucks in your company, a fleet manager would own certain parts of that. So, maybe the northeast manager would own 50 trucks, the south [would control] 200 trucks.
What we're hearing back from the business is that, "You're getting me information, but it's a day or two old. It's not, not only information on demand, but I don't have it at my fingertips." Then a light bulb went off, and this was about two years ago.
About two years ago, we started thinking through, "Well, jeez, an iPad's coming out. Maybe if we started designing these very thin data sets for the fleet managers, they can actually make decisions that day." Based on default, we got into the role of doing designs around iPads and Droids, because of the business saying, "We're getting data way too late."
So, we grew from there, and we actually became a very large subscriber of iPad, the iPhone, the Droid. Today we've got several apps that we actually even sell on the iTunes. We own our own private store internally, where can drivers can download our Droids and [iPhone] apps that we built specifically for the company itself.
The importance of that is the concept of the information pass, what we call the life cycle of on-demand, and ventured into the information at your fingertips. That whole ecosystem of data is ever-progressing. The next one is evolving into the voice of information itself with complex event processing.
Hannah: Now, I'm curious to know how long this all took. How did you deliver this project? How long did it take?
Tim: Well, I come from a background where the business, certainly when they talk to IT, the trust factor sometimes is not there. The first concept was to establish a beachhead, establish a landing zone, per se, that we gain the credibility of the business. The very first project, which was the idle project by the way, we looked at delivering something in that four to six week window that had a very intrinsic business value for a transportation company.
Remember, in a transportation company, there are really three aspects that influence the company itself: the maintenance, fuel, and the drivers. We looked at the fuel area and said, "What can we do right away with this big data concept, with this near real-time analytics, that would impact something that the business would pay attention to us and really consider us a serious enabler to help them support it?" That's where you got idle project that came in as a result of that.
We did something very quick. We did something very fast. We got it and we made a difference for the business to see that we will add value to them. So, that was our approach. Once we establish that "landing zone", we came up with a 36-month information management strategy that had over 13 strategic projects, and we're at our very last project today, as of right now. We've implemented the rest of those already.
We took the first successes, and then we built on the rest of the successes with those strategic projects.
Hannah: What were the biggest surprises or unexpected things you found during this project?
Tim: It's actually a good surprise, actually, to me. I had worked with a lot of various companies, from the U.S. Army, to the Dell's, to a lot of big corporations, and the adoption rate was extremely painful. But for some reason, the more that we showed them the capability here, the more they adopted it.
I don't know if it was because they just didn't really have an IT team that actually successfully listened to them, and actually worked with them to solve the problems, but it was the adoption. It was a pleasant surprise.
It was such a surprise that I didn't expect that we would get that many strategic projects to have to go implement. I think that was probably something that I was unprepared for. I thought, "Well, we deliver the win, they will give us a few more projects. We'll unfold the strategy a little bit at a time," but like you said, you mentioned it's only been three years.
They looked at it and said, "Well, how fast can you go?" So, I think that was something where we're guilty based on our own successes of implementing something that we had to force to go fast on.
Hannah: Finally, what advice can you offer to others who are contemplating a project of this scale, perhaps involving big data, real-time, or mobile technologies?
Tim: I used this analogy with my team initially, and I've stuck with it for years now, and I've changed my philosophy on it. You have to become the business. I know it's a cliche-ish term with IT departments. But when you're looking at and working with a business, if you truly don't understand what problems they're going to try to solve with your information itself, you're always going to be delivering things that you may not know will be successful.
So, know the business, become the business. Actually get so involved with the business that you almost learn more than what they do with what successes that you can bring to the table.
That's what I would tell people, is that if you really are trying to enable or build something that's going to make the company successful, really get involved with the business, and know what jobs that they take to become successful. If you don't, and you're just delivering them whatever they ask you to do, you'll have no knowledge on what this truly means to the company itself, and that's what I did.
I didn't know trucking at all. I came from a different background, but I made it a point to learn every aspect around the company in what I was trying to solve from a solutions aspect.
Hannah: Thank you so much for talking with me today. Tim Leonard, CTO and VP at U.S. Xpress.
Tim: Thank you.
Hannah: Thank you for joining me today here on the SearchBusinessAnalytics.com video interview. My name is Hannah Smalltree, Editorial Director for SearchBusinessAnalytics.com. Remember, you can find more articles, videos, and other resources right here on SearchBusinessAnalytics.com. We look forward to seeing you again. Thank you for joining us and have a great day.