 Live from Las Vegas, extracting the signal from the noise. It's the Cube, covering IBM Insight 2015, brought to you by IBM. Thank you for having me, thanks for remembering. That means IBM. Welcome back to IBM Insight everybody. This is the Cube, we go out to the events, we extract the signal from the noise. A lot of buzz these days about internet of things. I love this example because we have an example of a thing that is a human, Dave Hasse is back, he's joined by Doug Barton who's the Director of Product Marketing for IBM Business Analytics. Dave Hasse is an ultra-cyclist. We first met you gentlemen at Vision. Welcome back to the Cube, it's great to see you again. It's great to be here Dave. Great to be here, thanks. So for those of you who don't know, Dave cycled across the country, who was at 3,000 plus miles last spring, and then IBM wired them and captured analytics and heart rate and monitored the system and I presume caloric intake and helped you fine-tune your performance. So did I get it right, Doug? Yeah, well look, Dave had a lot of experience at this. When we met him, he was the three-time top American finisher of this race across America and it is a non-stop bike race, right? From the Pacific, Oceanside, California, all the way to Annapolis, Maryland. And when the gun goes off, you race. If you sleep, you lose, you know, people pass you. But with that said, we saw an opportunity to help Dave, I guess as we put it, race is perfect race, right? Be ready for the conditions, be ready for the deserts that he would cross, the mountains he would climb, the windy plains of Kansas. And also put some foresight to work. Our partnership with the weather company has given us an ability to look down the road and actually tell him when he should rest and when he should race. So those combination of things helped Dave achieve a really good outcome against this here. So you were set up at Vision, you had a little stationary bike going, you're going to do the same thing here, you're going to be wired up, feeding all the data. We actually had a channel on it. If you go to ibmvisiongo.com, you'll actually see an ultra-cyclist channel with Dave pedaling away, which was awesome. So how'd the race go? Tell us about it, take us through the experience. So the race for me, the goal was to race the perfect race. So we started in San Diego, took off racing pretty fast and it was a great race overall. We hit temperatures in the desert of 120 degrees, went through mountain passes. It's 3,000 miles, 170,000 feet of climbing. And for me, I finished the goal to win the race. I finished second, eight days, 20 hours. My previous fastest time was nine days in 21 hours. So we were a full day faster than I had been the previous race, which was in 2008. So seven years later, I'm racing a full day faster with the ability to monitor myself and keep myself safe and put out an effort that was very controlled and consistent the whole race. So honestly, the race was as hard as it was, it was somewhat easy because we were so maintaining the heat and keeping my core body temperature where it needed to be and just watching all of the other factors in the race. That's phenomenal. I came to second, congratulations. I mean, maybe it didn't meet your goal, but that's incredible. Yeah, we wanted first, right? That was more first. But the one interesting fact that I think we ought to share is how that is really a spectacular outcome just to stay in the race. Half the racers that started this race would DNF did not finish this race. There was at the same time that the race across America started on the Pacific Ocean, a shorter race, which was merely, I think, 900 miles to Durango, 900, right, nonstop. 60% of those racers didn't make it. The demands of that first Sonoran Desert, the 120 degree heat, took people out. They efforted above a threshold for the human body to operate a metabolism and they keep them fit and able to continue. So when 60% of the racers on a shorter race succumbed to that, it shows you the impact of our ability to monitor in real time Dave's core temp and intervene smartly, cool him, make sure he kept his effort below a threshold. You got to be in the race to win it and we met that first challenge with that technology. You got to stay in it to win it, right? So how many people actually entered the race? So there were 41 solo racers that entered and 18 finished the race. And so the next closest person, third place to me was nine days and 21 hours. So I finished, so I was kind of. So similar to your previous personal best. Yes, and I was 12 hours away from the first place finish. And I think the IBM analytics that we were using certainly played a big role in my effort and improving upon that performance. So we had a plan and we used the information we were getting from the analytics to change our plan along the way and constantly change that as we went when I would sleep and when I would rest. But really it helped us make good decisions at very real times when they were needed to be made during the race. And that was a huge role in the finished time in the effort that I made. So take us through the race, you start in San Diego and then immediately you're thrown into the fire, literally, you get into the desert pretty quickly. Yeah, the first hundred miles of the race, it's all uphill from the ocean and then you drop into the desert and at that point it goes from, I think we started at about 70 or 80 degrees and we were at 110 degrees within the first hundred miles of the race and then from that period of time for the next 300 miles you're in that type of temperature and at night as well, but it was still 100 degrees and monitoring that core body temperature became really important for me. So the crew that I have along following, watching our dashboard in the car of all that data and all the things we were capturing was able to tell me maybe when to effort and the biggest part was keeping my body cool by using ice various ways to take my core temperature down so when it would hit above 100 degrees we'd cool down with ice or pour water on my head or to keep that performance at a very consistent level and continue to ride into Arizona and Monument Valley in Utah was extremely hot during the day and so we were able to control that the whole time with that information. The feedback, I mean obviously the data was continuous was the feedback to you continuous? How did you get the information and how did you respond to it? So I wear a headset that my crew gives me information through so they give me the route, take a right turn here but they also, I can tell that my core body temperature would get a little hot. I could feel my effort kind of struggle a little bit and I'd be like, what's my core body temperature and they'd be 102 degrees and it's like, well get me some ice and a cold bottle of water and I'd pour that on my head and my temperature would instantly go down on my core body and my effort. You stop to do that or you just do it as you're driving? They'd pull up next to me and hand me a water bottle with ice or periodically would stop and put ice on my body and strangely enough the technology we used down the road, we were using ice packs and different things and went to using a sports bra packed full of ice to keep my body cool. We found that at the time of racing the best way to really keep me cool was when the crew reacted to what was going on and the conditions and changed what they were doing to improve the performance. So a little ice bucket challenge where you're on the way. It's about what it was, yes. Okay, so day one you're in the desert at first 100 miles right away. When do you first stop? How did you decide to stop and eat or rest or sleep? Yep, great question. So the food is coming from my crew nonstop so they're passing me water bottles with nutrition in it. We had said in our mind that we'd ride 30 to 40 hours and the first day and sleep somewhere so that carries into the second day of the race. And we were approaching Flagstaff and the next city was four hours away and within that dashboard window that we had we had analytics that said it makes the most sense to sleep within this window based on the weather conditions are gonna be facing and our plan kind of fit to that so it gave us a very confident plan and idea that this is the time we should sleep. So we slept in Flagstaff the first day. It was a city, it made sense. The next city up the road, there was nothing really in between was Tuba City along the course and so that was our first day of rest. Slept for an hour and a half and then got back on the bike and rode another. The next day we actually pushed to 26 hours because the weather and analytics was saying go a little further than what you originally had planned and we slept at 26 hours and then even day three we used that information to maybe sleep just at 22 hours in and so we did change our kind of game plan a little bit along the way each time based on the weather conditions. So you had the weather channel guys feeding in from their API into your system. Absolutely, if I can just give you a little bit of picture of what this looks like. It's a prescribed route. You don't get to pick your trip across America, right? It's a prescribed route. So we had the 25,000 geo location waypoints that describe the route and we knew where Dave was. We built a model to predict and estimate his progress down the road and we would intersect then or join the geo location with the time Dave would be there and we would know the direction of the wind from the weather company relative to the road and we would estimate how much help he would get or how much hindrance he would get from the wind direction relative to the road. So basically we could peer into the future and know what Dave would face and therefore what yield he would get, like how far down the road he could get and therefore we could draw a curve and allow him to know the best time to rest so he would awake to the best conditions. When I say awake, remember we're only resting 90 minutes or so or 120 minutes. So and when he was sleeping, was he instrumented or? It's a good question. I hope you took all those things off. No, we were charging most of those. So they're all battery powered, you know and so they were being charged when I was sleeping. So we'd be, you know and then I'd wake up, we'd quick throw them on and take off again. So did you have trouble getting to sleep or no problem? No problem. So you typically have no problem sleeping? Generally I'm pretty tired when it's time to go to bed. So, you know, but so I'd pull into where we were going to sleep, whether it was a motel, the back of our van or an RV and go instantly in, take off all my devices. My crew would throw them on chargers, someone would take my bike and then would sleep for an hour and 30 to 40 minutes. Our goal was always to be off the bike no more than two hours. So anytime that was spent, you know, taking off clothes or devices or whatever, that was part of that two hours. So generally I would get about an hour and a half of sleep each night. So in the eight days and 20 hours I had 14 hours of sleep over that time period. And is that typical for the, let's say with half that finished? The guys that are going to compete to win that race, they're not sleeping very much at all. My competition I think went 42 hours, maybe the first day before they took their first rest and gained an advantage there. And we were kind of chasing that advantage the whole time. And as far as, is it always sleeping or do you sometimes just rest for 15, 20 minutes or a half hour? We generally weren't off the bike at all. Unless you were sleeping. Unless we were sleeping. That's mistakes I've made in the past is, you know, you get off your bike and then you're there kind of, you know, wasting time, so to speak. So it was generally riding the bike or we were going to the bathroom or sleeping. And you had to be awakened at these intervals or did you just kind of wake up with an interval? Most of the time, strangely enough, I would wake up about two minutes before the alarm was set for me to get off and I'd wake up and, you know, throw on my clothes, do whatever we were doing and then get back on the bike and take off again. And to the advantage to pretty good conditions usually. I wasn't fighting a headwind or anything like that. It was usually waking up to favorable conditions and moving down the road as fast as I could because it takes a little bit to get your body back moving after you've been, you know, riding and then sleeping and riding again. So weather predictions sometimes can be unpredictable. How did you find your ability to predict the weather were you better than 50% accurate, 100% accurate? You know, if the real tail of the tape as it were would be our ability to predict where Dave would be, I can say we were spot on. The combination of things, I have to tell you, we did learn a few things along the journey. One of the things that where we deviated from race plan and I know Dave wasn't keeping secrets, we worked on this hard, but he did have a plan to kind of met out his effort evenly. 220 watts, watts is the measure of a biker's power they're putting into the pedals, right? He had him and his coach had tuned in and dialed in the 220 watt line to stay at threshold. And so we had modeled 220, but we found that our models were a little bit off. We kept checking the error rates and we were still off by 10% on where we predicted he'd be. So we found that we actually had to predict Dave's behavior as he went up some of the steeper slopes, especially as we got into Colorado, he was putting out a lot of watts. And so our models had him too slow on the road. So we inserted predictive analytics to actually calculate and predict what Dave would do when faced with the slope conditions and our estimates got much better. So the combination of predicting his behavior, again he wasn't keeping secrets from us, I think his plan was to ride evenly, but the conditions frankly wouldn't allow him to ride at 220. So we had to kind of predict his behavior, add weather to that and we were spot on. I mean, we could predict exactly where you'd be on the road. But the conditions wouldn't allow you to ride at 220, but you were riding faster than 220 if I just inferred. Yeah, so at certain points in the race, you're putting out more power because you're climbing a hill and that effort goes higher and then maybe on the descent you're coasting, so you're really not putting any power out or you're pedaling slowly. So, but I think what we learned was that the effort and my ability to recover after sleep was maybe better than we were thinking it would be. So that's the gut feel human factor. That's the gut feel, how you felt. You said, I feel good climbing. I'm going to, it's not speed, I said you speed, but it's really output. It's effort, yeah. Right, right. Okay, and you said, okay, I feel good. I'm going to really crank up this hill, exceed maybe 220 to sort of violate that threshold that you guys set for yourselves and then you were able to adjust to that pattern over the course of the race. I mean, this is sort of, in some sense, this is exactly what we describe in cognitive computing, the ability to learn and adapt. And so our model's actually improved. We had to build a learning system because I don't know if you remember this, but when you interviewed Dave and Jean-Francois Pouget who's our distinguished engineer on this, he said there was a race to prepare for the race. Well, sure enough, our project had to get done on time. You know, Dave was starting with or without us. So we put in place a model that we thought was a good first approximation, but we knew we could tune it. And by the time we got to Colorado, it was kind of like, we know where Dave is, you know, we'll stick out our arm with a water bottle and sure enough, Dave would be there. And how did, how did your competitors react to this? Are they, do they want to do it? Are they saying, oh, they're crying foul? Or what's the story? So there's two stories to that, yes. Several racers were like, well, how do we use this? One actual IBM employee was doing the race as well and she was like, I'd like to use this, but didn't have any of the equipment we use. She's not into using power or a computer or anything. And then one of the competitors in an interview was making fun of, you know, he said, well, I'll just swallow Advil when I'm in pain. And I was swallowing a core temperature pill to maintain my core temperature. He did not finish the race. So, you know, I think what we used was effective and worked and to be honest, it made the race really, I hate to use the word simple, but it was ride your bike. We've got all this information. Just keep going at this rate. There was one point in the race where it was like, hey, can you pick up the pace? Because we were getting close to the first place racer and they were doing some analytics on his speed and pulling up information for him saying, you know, in the next, maybe but with 80 miles to go in this race, based on your effort and his effort, you're gonna catch him. So we need to effort just a little bit more. So that gave me a little bit of confidence in my ability saying, okay, let's pick up this pace just a little bit. You know, then we hit some steep mountains and the East coast and, you know, that changed a little bit. But there was, it kept me very confident in the racing that we were doing and the decisions that we were making, it also added. So from just the standpoint of, you know, learning, it was like, I feel confident. They're telling us to sleep. I feel like I should be sleeping. So this is a good decision to do it. So all of the decisions we were making in racing were backed by information we had that also seemed very, you know, very normal in how we should react. The invisible hand that became visible and guided you. That's right. There is a lesson here, I think, you know, as we use analytics in our businesses to make smarter decisions in the moment is that we can make him with more confidence. We don't spend as much time kind of let data and analytics inform our judgment, right? It doesn't necessarily over swamp intuition, but it can be there to guide. And frankly, you know, any effort that Dave would waste mentally thinking about, am I doing the right thing? Would have been something that would sap his power, right? And take away from his ability to effort. So, you know, save those calories for something else. So, did you take any pain killer? I did, no, no. Honestly, we were so perfectly, my crew was so good at hydrating me and keeping my body accurate, using information on the dashboard, they could see what I was doing, they could see my efforts and hand me the right amount of calories and everything. So it was, there were points in the race where I just felt, you know, absolutely perfect riding through Kansas was a dream we were just ahead of the wind, you know? So there's Kansas, previous races has just been a nasty crosswind. And we had kind of a side crosswind, which is almost ideal, you know? It was perfect. We were just ahead of everything, never faced any bad weather conditions. And it seemed like we were always just right almost on the perfect weather conditions the whole time. I never faced anything other than heat, you know? But from a wind standpoint, which, if you talk to any cyclist, their biggest complaint is, well, I'm not going to go ride the wind today, you know, the wind, the wind. And we had, you know, we didn't necessarily have the greatest tailwinds, but we never had a headwind either. So it was very good all the way across. See, I've smarted the friction of the headwinds. What about solid food? No solid food? The only time I ate solid food was when, I shouldn't say completely, that when I would sleep, so right before I would sleep, we would eat some protein as solid food. And then fruit was our main source of, I guess if you'd call that solid food, we ate a lot of fruit for other calories. And then the rest was liquid calories. All right, Doug, and can you describe the tech behind this in some detail? So a lot of tech in Dave and on Dave, as you might imagine. I guess I would start with the core temp sensing pill. So I think you ingest at eight of those during the duration. They stay in his digestive tract and give us information on the core temp. That was actually, would send a radio frequency to a module. Module would go to the Android phone, Android phone. Actually, let me pause here, because that Android device actually consolidated a lot of the data feeds from the power on his pedals to his speed. His bio harness, we had a bio harness on him that would give it ECG breathing rate, heart rate. And those were the primary things that we would consolidate under the Android phone and then send to our cloud. That technology is called the Internet of Things Foundation. So securely using the cellular to cloud data transfer. We would use the weather company API to bring in weather data about the geolocation points. We used kind of our analytic technology to build the predictive models and then they calculate the what ifs. Remember that decision optimization really said, it kept looking out to say, where should Dave rest given the conditions we can predict for him when he gets there. And then finally we would deliver it all back to a dashboard using our Blue Mix cloud instance. So people had a dashboard on their iPhone and their iPad that would always show Dave's location and these biometric and other optimization recommendations. And the people that were analyzing the state of the human factor were what, data scientists? They were mathematicians? We had one data scientist, Jean-François Pouget who you've met before and by the way, I'm sitting in this chair. There's a lot of great IBMers and other people that contributed to this. I'm so humbled by the team we put together. Really proud of that team. But Jean-François and I were probably on 24 hours. Well, maybe not 24. We were on 18 hours a day. We would sleep for six and we'd get up and find out where Dave is and we would do a little strategizing in our war room. But yeah, we were just in the cloud, as it were, right? For Dave. But obviously humans were part of this, clearly with Dave, but in terms of helping optimize, what percent was the sort of human last mile interaction? Well, there was, well certainly, Dave got all his instructions through his earpiece. So we don't put a dashboard in front of Dave, we let Dave ride his bike, but he would get from his crew the information. I would be in constant contact through text messaging and other things with his crew, though, phone calls. We cooked up that plan to catch Sevi. Dave, you didn't let us down. I should say really quick, as Dave kind of characterized that time when Sevi's in Pennsylvania, Dave's coming out of Ohio, we picture a weather storm moving right between them and I'm looking at the pattern of the winds and Sevi's got a tailwind and there's Dave fighting a crosswind and my heart sank a little bit in that moment. So that's exactly what you don't want, right? The leader's getting the benefit of something that's essentially free and helpful and Dave's suffering more. But aside from that little moment of sobering circumstance, it was a terrific partnership, people working on it, but the systems really held up for us, right? The models that we built were really good at predicting and giving us insight on the go. And where did it end? Where'd the race end? Annapolis, Maryland. Okay. So we finished there and a lot of the things that happened. So like the dashboard, I've seen the dashboard, but I've never seen it physically working because I was riding my bike and so, you know, we finished second, looked at all of that information and learned that Sevi, who finished first, had a 1.46 mile per hour wind advantage across the country. So he had a slight wind advantage. We figure we saved 12 hours using the analytics. There's other devices we can use and so I think we're looking forward to trying to improve upon what we called the perfect race maybe and do it again. So you'll do it again. We're working on that process. So June of 2016, the next races. Okay. And so we're working on how to use a few other devices and how to make the system easier for the crew that I have along to use with less devices, but get the same information. So if you had to do it again, you wouldn't have slept the first night. You would have gone the whole 40 hours to get that extra mile and a half. If I could, yeah, I think that's a big key, I think, is how far you can ride that first day as long as you can recover the next days. And one of the racers, the elite racer who was supposed to win pushed himself and he ended up dropping out in, I believe, Kansas. So, yeah. Can't win if you don't finish. Right, right. So Doug, how about you guys? What are you going to do differently? Well, I think there are two things that we think there's another 5, 10% yield to get out of these super athletes. One is, and there's been a lot of science put into this and sports science and physiology, so we're really proud to, we'll have an announcement here soon, but there's a Watson ecosystem partner that's put the power of Watson together with their sports science understanding to assess the trade-off in when, how to get athletes ready to perform on race day. Now, Dave's got a lot of experience training and getting ready, but the challenge with someone like Dave is, how much do I train so that I have the big engine for race, Dave, but I get it fully recovered and ready to go so it's optimally ready to perform. We have a partner in the Watson ecosystem that has put Coach Watson together. So you've heard of Chef Watson, this is Coach Watson, and we're going to put Coach Watson on Dave's side this year. The second thing we're going to do is another real-time monitoring of what's called muscle oxygenation. So think of this as part of the physiology of efforting, especially endurance athletes, is keeping track of where you are relative to your threshold. You want to be aerobic exercise, not anaerobic. Anaerobic creates a lot of byproducts your body can't deal with over long distances like lactic acid. But if we keep Dave right up to the threshold without ever going over, we think we can get a little bit more effort out of him without doing harm to his body. So imagine this is just creating a bigger engine just by having a little insight. And what he wears is a, it's an optical sensor on his calf to give us that real-time data feed. So those are two things that we're going to do that's going to get the next 10% of it. Yeah, instrumenting the burn. So, and this is for me, is that right? Yeah, that's right. So that's a photo general look here of the race. So I'm really thankful that you guys came on the cube again, I can't wait to hear what happens next time. So Dave, congratulations and Doug, you too. We really appreciate the story and the time. So, well done. Thanks for having us. Thanks for having us. All right, you're welcome. All right, keep right there, buddy. We'll be back with our next guest. We're live from IBM Insight. This is the cube right back.