 Live from Las Vegas, extracting the signal from the noise, it's the Cube! Covering IBM Insight 2015, brought to you by IBM. Now your host, Dave Vellante and Paul Gillan. Welcome back to IBM Insight everybody, this is the Cube. The Cube goes out to the events, we extract the signal from the noise. Check out ibmgo.com for the full digital experience of IBM Insight. David Kenny is here, he's the chairman and CEO of the weather company. A lot of talk, David, about your company at this event. Welcome to the Cube, thanks for coming on. Glad to be with you, thanks for having me. So it's really interesting to note the IBM partnerships. We go to a lot of these events and a lot of times you say, hey we're going to make SAP run faster, we're going to make Oracle more reliable, whatever it is. The weather company, you see Twitter, you see Box, really interesting partnership. So congratulations on this transformation. How did that all come about with IBM? Well I think IBM saw us at a number of shared clients. We've done a lot of work on our weather forecast in the last few years to make it mobile centric. So if you start with weather on a mobile device, which is how most people use it today, you have to have every location on earth, so it's about two billion named locations, and you have to be live. We have to update our model every 15 minutes. Historically models were updated four times a day, and they did about two million locations. So it was about an 11-fold increase in volume. And we solved that as a technology problem, we solved it as a platform. IBM were interested in how we did that, so we've licensed our platform to them, which they use with other data sources, and they've helped us take these great weather forecasts into insurance, energy, retail, all the places they're serving where weather makes a difference. So you talked about this in your keynote, and you said you generate essentially 40 terabytes a day of data every day. Well that's how much we consume. We need to read that much to feed our algorithms to predict. So those are data sources that you're ingesting and blending. Exactly. And then you set it off camera, whether it's just one big algorithm. Well it is. Ultimately it's physics, right? So we map the whole earth, and you have to map it exactly at the same time, and then you predict where it's going to be in 15 minutes and 15 days in three months. And so you need those models to work. So when you talk about Internet of Things, I mean you guys are way out front in processing real-time data. Most businesses just getting started with this. What are some of the complexities they should think about when they get into large amounts of sensor data? Well I think the sensor data, it comes from different sources, so there's no standard to it. We get barometric pressure from 40 million smartphones. We get data from 140,000 personal weather stations that people put in the yards. We're trying to get data from windshield wipers and when they go on and off. We get data from the wings of airplanes. We get data from satellites. So it all comes in a different form. I think you've got to find an efficient way to bring it in, and then you've got to have a talk to each other. So that's a lot of what our algorithms do. Part of what we're excited about with Watson is then being able to actually read that in a cognitive way to be able to connect things that are very unstructured to be even more useful to people. Is there an example of that? What kind of connections are you discovering? So one example is what we do for airline pilots. So we've always been able to do the weather. They need to know the weather to fly. But some big issues for airlines which were related is which route to take, which runways are going to be open or closed at the time they land, and what's going on with turbulence. So we took all three of those and said, how can our weather data improve those decisions? And in fact, because we predict the weather, we predict the closure of runways before they're actually announced. So that's allowed them to be ahead of it. That's also allowed them to consume less fuel. On turbulence, we've been able to measure the wave oscillation in the middle of the atmosphere by how much the tablets are moving because they're reading the weather on a tablet. That in turn allows us to build a turbulence model for all the other planes. Kind of like the Waze model for traffic, but for airline pilots. So how is your business evolving then? You're seeing IBM talking about all the different ways the weather can be applied to different businesses, insurance and logistics. How are you evolving your business to identify these new areas where you can find new customers? So fundamentally, I think what we've shifted is from focusing on weather data to focusing on weather decisions. So whether it be the hundreds of millions of people who use this every day as individuals, or whether it be the individuals in businesses, we're trying to help people make the right decisions. Some people it's about sports, some people it's about their life outdoors, some people it's about travel, insurance adjusters, it's about what happened in that place airline pilots have covered. So our business has really become a decision business more than a data business. So talk about your story, how sort of the weather company has evolved. You've got private equity, you've got some relationships with, I think, NBC. How did you get here? Well, you know, those are investors. So they largely hired me to do the job and leave me alone. I would say, you know, when I came, it was a cable channel that had built a digital property. Coming from Akamai, it was clear I was going to take a different view, which is fundamentally how do we disrupt weather before it gets überized. And my view was it's going to change completely when your primary way of consuming it is in a mobile device that enables individuals and businesses to make better decisions. So we rebuilt our entire infrastructure to be able to do this on mobile. In order to do that, we had to move the whole thing to the cloud. We had to close a dozen data centers. We had to build a different kind of platform. A lot more Internet of Things data. Our forecast is far more precise because it's for each individual location, not counties. And it's real time. And that's really changed the way people use it. And you've overseen that transformation. Yeah, we largely did that since I got here four years ago. And how do you make money? So we make money in a number of ways. First, we have subscription businesses. So most of our B2B businesses pay a subscription. Airlines per tail, insurance as a percentage claim. So how they're using it. And then on the consumer side, some subscribe to specialty forecasts. And the rest of it's an ad model. Similar to data-driven advertising like Google. It's not display advertising. It really isn't. It's really about giving people ads in the right context. So your model is dependent upon your weather forecast, your algorithms being very good. And also your distribution model. Maybe talk about that a little bit. So what makes your algorithm so good? What's the DNA of the company? Well, I think the reason the algorithms are the best in the field, and they really are, there's independent people who measure that, is that they're very precise. That we rebuild all the algorithms for every lat-long combination down to four decimal points on Earth. Because to make a decision like the Red Cross talked about yesterday, where to send resources or to tell a plane where to avoid turbulence, you've got to be really precise about that location. And so that was a fundamental change. And there's just so many new things we can do with that. Now that precision only mattered when your forecast was on a mobile device so that it's exactly where the decision maker is as well. So that's where we are. Listen, we continue to get better at it, mostly by getting more data. And that's where the Internet of Things has been so important, and where this partnership started. I think most people are fascinated by weather and how you predict the weather. What are some of the toughest prediction problems that you still wrestle with? So I think that the challenge with the observations is the middle of the atmosphere. We know the outside of the atmosphere is about 100 kilometers thick. We know the outside pretty well from satellite data, and everybody has had governments run those. We've got lots of observations on the ground. Smartphones, weather stations, windshield wiper data, radar. The middle's hard. We do a lot with airplanes. But for example, tornadoes and lightning happen between the clouds first and then come down. So how do we better understand what's happening in the middle of the atmosphere? And we're continuing to find new ways to do that. Part of what I want to challenge Watson on is to do ways to read visual data, to read photos of the sky, to read videos, so that we can find different ways to understand what's in the middle. That's the difference between today, which is a very precise forecast and a perfect forecast, is understanding the physics in the middle. So is that the human today, the human factor actually closes that gap, that sort of tribal knowledge that either a pilot has or some other... Well, we read the data from planes. No, I think it's just a gap in the map. So we just have to interpolate what's happening based on what we know around it versus knowing precise. And that's partially why it's not 100% accurate. So there's no perfect way to know this yet. But it's a problem to be solved. Describe in your view the advancements in weather predictions over the last 10 years have been pretty... They're awesome. So the five-day forecast today is as accurate as the one-day forecast a couple of decades ago. So we continue to make progress on that. I would say our added contribution is to be really precise about the location because accuracy has been historically measured by county or state. It's pretty broad areas. And I think that's less relevant. That's not useful enough for a decision as being really precise. You mentioned the uberization of your market earlier. It was a really interesting comment. You see the potential for disruption to happen at the local level. Could your business be disrupted by micro-forecasters who are working just within a specific city or even neighborhood? No, I don't think so. Because the atmosphere is above the earth. It doesn't know geography. So the weather over Las Vegas today was probably the weather over Los Angeles yesterday. Hawaii a few days ago, it started in China. So understanding the patterns of the atmosphere is what's important. So you actually do have to map the whole earth to be accurate. You're not going to crowdsource the weather is what you're saying. Well, we do crowdsource contributions. You do crowdsource it. We do with personal weather stations, with phones. We absolutely crowdsource observations. Or yes, okay. I don't know. I mean, I think that there are some humans who can give us added knowledge to fit the model. But I don't think there's anyone who could just focus on an individual location and do anything as close to what we do with a global model. So you are the uber of weather essentially. But if we didn't do it, somebody else would have. That was the question. There are big companies that have done a lot in mobile and data. And my belief was we needed to run faster than them because they would get to it just as they've come to maps, right? So they changed the map world and made it right for a mobile device. And my view was the next project is to map the atmosphere. We have to do that before one of these big players who have done mobile maps comes to it. What is the culture of science in your industry? I'm curious because, I mean, AccuWeather is a competitor, right? On some level. I mean, they don't do anything as precise as we do or the B2B part. They have a consumer forecast for sure. Right. I think in the community security area, there's lots of competitors but they all sort of collaborate at some level. Is that the dynamic in your industry as well? Well, I think what we've done at the weather company is changed the definition of the industry. So historically, there's been a weather enterprise, a weather community with a bunch of small companies who were involved with weather and were closely with NOAA and its counterparts around the world. I think the challenge with that is it was just the atmospheric science and it wasn't advancing the algorithm. So part of what I've been focused on for four years is we've got to bring in the IBMs. We've got to bring in the big cloud platforms. We've got to bring in folks who understand data. We're not going to make further advances just in atmospheric science. We actually need the intersection of atmospheric science, computer science and analytics. And that's why this partnership was important to me because it totally changed the game of what we are. It is the overization of weather, which is to bring a whole new set of skills into a historic business. What kind of weight does historical data have in forecasting the weather? Is it a little bit important, hugely important? Can you describe that? Historical data is important for macro-translight climate change and sort of our climatic conditions changing. It's only semi-useful. Weather never repeats itself. I can't have an algorithm that says there's a 10-year cycle and this will happen again. I can't tell you this El Nino this year is anything like the others. So it's somewhat helpful to improve the algorithms but what really matters is the data right now. And that's the other reason I think updating our model every 15 minutes, 96 times a day, has been key to improving the precision. When you work with Watson, have you seen any immediate results to your business in working with Watson? And if not, what are the one- or two-year forecasts, if you will, for what you expect Watson to do for you? Well, the multiple of us. In the forecast itself, because our forecast is through my algorithms, we still have something we call human over the loop. So sometimes the computer will generate a numeric forecast which will say heavy rain with high wind. What it should say is category 5 hurricane. And so we have to overrule that or blizzards sometimes take different directions. So we have humans change the language to make sure it's understood. I think what Watson can help us understand is how will people understand that so that even that human interaction can begin to be automated. Because I can't employ enough meteorologists to put an override on every forecast. So I think this is going to change it. I also think different people interpret forecasts differently. It's a tragedy to me that people still die in weather-related accidents, that people get stuck on ice storms in the south because they don't know what to do. I think us being more clear, what are people hearing? Are different people hearing different things? How do we communicate it so people actually take the right action? We'll save lives and property. And I'm excited to solve that problem. So really it comes down to a human dimension. It comes down to language, to translating all these algorithms into something that's meaningful to the audience you're trying to reach. That is the reason people still watch The Weatherman on TV and why we have a lot of text around our forecast. Not everybody is skilled at reading the data. And there's a lot of data in the weather. It's not just precipitation and temperature. So are helping people understand that adding a human voice is key. Part of what I think Watson does well is to humanize numerical prediction. You said that you develop your own platform for gathering all of the sensor data and for ingesting it and interpreting it. Presumably you did that before all of these open source big data platforms came about. You had to invent that? We work with Spark and we work with Hadoop and we work with Cassandra. So we absolutely stay current and use new platforms. It's great when something like IBM gets behind them because they can scale. And we think in a very open source way. So we continue to use new technologies as it's developed. You mentioned that's part of the reason why people still watch The Weatherman and The Weatherperson on TV. It used to be the case I want to watch Channel 5 because that weather person has a better forecast. Eventually you're saying they're all operating now in the same weather company data, right? For the most television stations in the United States and about half around the world use our data. They license our data in our graphics to tell the story. I was going to ask you about the non-U.S. component of your business and how prominent you are there and what the competitive landscape looks like overseas. So on their consumer side, after iOS 8 we're preloaded on every iOS device as a widget which connects to our app. So obviously iPhones are everywhere. In Google we're the default. So if you search for anything weather related you'll get a one box with a 5-day forecast from weather.com and a link to our app or our mobile website if you don't have the app. So our usage is pretty global. Our monetization tends to be a little more domestic and we're going to work on that. And there are other apps that are competitive to the iOS app but you've powered those as well? Absolutely. We've long believed in being open ourselves so our API is used by about 40,000 other publishers some of which are for consumers and some of which are for businesses. So where do you want to take the company? IPO in your future? Are you happy being privately held? I've run public companies. I've run private companies. The capital structure needs to match the strategy of the company. And maybe it should be combined with somebody. So we're going to look at all those options over time. The real focus has been making sure that we save lives and property, give people a perfect view of what's going to happen to them next. And then the rest will take care of itself. It always does. David Kennedy, thanks very much for coming on theCUBE. Really great story and appreciate your time. Delighted to be with you. Thank you. You're welcome. Alright, keep right there everybody. We'll be back with our next guest right after this is theCUBE. We're live from IBM Insight 2015. Right back.