 So my name is Michael Kauffman. I'm CEO of Big Nimble. I was very lucky to get that name dot com. Data consultancy that helps organizations get the most out of big data. And thank you for coming here today. And regardless of how much you know about big data, even if it's very little, you at least know this. Big data is a lot like teen sex. Everybody's talking about it. Everybody thinks everyone else is doing it. So everyone claims they're doing it too. Now, like many of the most insightful points about something, it comes from someone who's on the outside. That came from Dan Ariel. He's an economist at Duke University. And this is the most popular quote if you see a lot of talks on big data. And they usually open with this, and then they move on to what their presentation is. And I'm not going to do that. Because when I first saw this, I asked myself, well, why is it funny? And unfortunately, a big part of the reason it's funny is because it's so true, which is a real damning indictment of the whole big data industry. I mean, this is an industry that generates... We're reaching into the billions per year now, and they really only have one task, which is to help you understand and use big data. And yet most people struggle to answer the question, yeah, but just... Can someone just tell me what it is, please? And so what I'd like to do today is both talk about what it is and dip slightly into the different ways that organizations can start creeping up and making the first initial steps with using big data the right way. Now, I don't know about you, but I can't get anything done without a to-do list. So here's our to-do list. So what is big data anyway? Now, when I give this presentation, it's usually an hour, and I don't have that much time. So unfortunately, we're going to have to strike the second part. But please feel free to come and ask me any questions if you have. But we're definitely going to define what big data is and give you a better understanding of what it is. Now, I want to start by sharing with you the breakthrough moment that led to what helps us give the most value to our clients. And it's starting... Now, I've been programming since I was nine. I rose to the ranks, became director of IT. I was CTO for a number of organizations. But the person that helped me with the insight that I'm going to share with you today, a very special person, he got started working with data at a very young age. He was actually an unpublished photo of him. This is my son Finnegan. He's one here. Now, he was not one at the time that he managed to give me the breakthrough insights that I'm going to share with you today. We had to wait until he was two for that. You see, what I wanted to do was I wanted to introduce Finnegan to the classics. I'm talking The Beatles and Zeppelin and Paul Simon. And so I made a CD with two or three songs of each one of these artists on it. Four or five CDs. But I didn't want to be prejudiced against the modern classics like Bieber or Perry. I don't even know what. But I made a few CDs of those, too. And I gave them to him and he has a CD player and he knows how to put the CDs in. Nonetheless, he took the first CD and he just chucked it. And then he took the next CD and he just chucked that one. And I got really frustrated and I was like... And so I googled, when do kids learn that things have a purpose? When do kids, the brain, start to understand that there's something for which you should use something else? And I found out that my son was right and I was wrong. Who here has heard of the candle problem made famous by Dan Pink at TED? Yeah, there's a few of you. It's actually published by a guy named Dunker in 1945 and it goes like this. People are walked into an empty room. There's nothing in the room except a table with these items on it. You have a candle. You have some tax. And you have matches. And you're told that what you're supposed to do is you're supposed to light the candle and mount it, not on the table. You have to somehow elevate that candle off the table. Well, I'm not going to brain-tuse you. I'm going to tell you what the answer is. The answer is there's not three things. The way you do it is by recognizing that there are four things. That box that's there is in fact integral to making it happen. It is not ancillary to just holding the tax and keeping them from rolling off onto the floor. And indeed we know this because if you run the same experiment and you position things like this in the lower left, everyone gets it and everyone gets it in about a third of the time. And the only way that you are able to solve the problem is by overcoming something that we call functional rigidity, which is the way you use something as the way it was presented to you. And what my son helped me realize after I've been working for data for 30 years was that functional rigidity doesn't just apply to physical things, which we've known for about 40 or 50 years, but it also absolutely applies to data. People tend to use data for the reason for which it was originally gathered or presented. And I'm going to go ahead and give you an example of what I mean about using things for different reasons. There's a commercial real estate company out there, I think they're in Australia, very smart. What they decided to do was use elevator logs to increase rents and cut down on default payments the number of tenants that couldn't pay. Now I was also an attorney for a little while and I'll tell you a little bit about the way these commercial leases are written. The way it works is if you're not doing so well and your bank accounts draining, there's a clause in there that says if you don't have enough money for more than six months, you have to post a six-month bond. Well, that's a good way to just drive them right into the ground because they don't have the money. And so once they're in bankruptcy, it's really painful if you're the landlord because they get to stay there but they don't have to pay. So what they did was they looked at elevator logs and elevator logs used for safety reasons, track how many trips there are, up and down, what floors you go to, what buttons you hit, and because you're not allowed to run the elevator if a certain weight is exceeded with a quick small amount of math, you can figure out how many people are going to which floors and you can figure out which companies are starting to lose employees because they can't make payroll. Because the first thing that happens is if things are not going well in your business, you know the rule, tell no one. That's step one. Step two, you lay off a few people. And step three, you enter into bankruptcy and you're just shut down. So the commercial landlord would reach out to those clients and say, I know you rented 10,000 square feet. It was a 10-year lease. You only made it two and a half years. How about this? Maybe you're just a smaller organization. We'll move you into 3,000 square feet. And now they get to turn around and rent out the 10,000 square foot and they don't have to go through bankruptcy and they don't have to do any of that. This is a perfect example of using data, giving it a second or third or fourth life. You already pay to store it, manage it. Why not use it for more than one reason? So that's a big part of the insight about what I help clients do is look at data in different ways. We're going to go into some of the big data types of data that there are and we're going to look at the very creative ways that people are using things in different ways. Before I do that, I want to take a look at what other people are saying what big data is. And I don't know if you've done this, but if you type in Berkeley University and 40 big data definitions, you will discover that someone has compiled a list of competing big data definitions. Even the experts have no idea what it is. Or if they do, they all think the other is wrong. I've gathered together four here that I think are the most, they're reasonable, they're good ones, and we'll see what we're dealing with here. So the most famous one here is high volume, high velocity, high variety data. Raise your hand if you've heard that one. That's probably the most famous one. That's from Doug Lanny, 2001, got bought by Gartner. He got upset that people weren't giving enough credit. He republished the paper, go Doug. So he writes this, that's from him. This one, big data is the new and massive data types that have appeared over the last decade or so. This one comes from Thomas Davenport in his book on big data, probably the best one out of, in 2013. He's a grandfather of data decision science, sort of the precursors to big data reaching back 40 or 50 years. Big data is about the analysis of data that's really messy, or where you don't know the right questions to ask, where you look for patterns or anomalies. A lot of big data scientists really like this one. This describes the process of big data. He's the curator at data.gov, one of the largest repositories of public information in the world, and a backdrop to a large number of big data projects where you have to connect with populations, incomes, race, geographic distributions. And the last one here is big data is harnessing information to produce useful insights and goods of services and significant value. This comes from Kenneth Kukie, now only one person got to be the person to go to ted.com and introduce all those Tedsters to big data. That's Ken. They asked him to do this one. You look at this and you're like, yeah, okay, that's great. But I don't even think you realize how different these are. I mean, we're talking about a definition here. This guy says, the first one, he says, how do I know if it's big data? Well, I'm going to look at the characteristics of it. Is it like moving really fast? Is it like a lot of it? Wait, is it like really different? Oh, that's totally big data. And this guy says, I don't care about the data. Is it like from 1999? Because that's totally not big data. But in 2012, yeah, that's totally big data because it appeared over the last decade or so. This guy says, I don't even care what the data looks like. Did you think about it in a new way? Did you stare at it differently? And this guy, this is my favorite one. This guy says, you may have thought it was a big data project, but if it didn't produce useful insights of significant value, it turns out it wasn't. That thing you've been doing for the last year, it's not a big data project if it turns out not to produce a real home run. Now, I have my own definition of big data which I'm going to share with you at the end. I think you wouldn't like the one I gave you right now. It's a little bit contradictory, and it wouldn't make sense given what we've just talked about. So let's dive into how I view big data. But before we do that, one last thing I want to do is confront the concept that it comes from big. And to do that, we have to look at numbers. Now, spoiler alert here, we're going to build up to Amazon. So, air passenger bookings, 1.3 billion. This is all per year data. Visa transactions, 3 billion a year. This is sizable numbers. UPS track deliveries. Not all deliveries, just the ones for which there's a digital fingerprint. It's being tracked. FedEx, sorry, the federal check clearing house. I don't know if you know this, but when you write a check at Bank of America, and then you deposit it at Chase, those two banks have to connect with each other, and the way they do it is through a clearing house that operates at the center spoke. The total quantity of checks in the United States finally, after 2009, is finally decreasing. We finally turned the corner on moving to digital money movement. But it was 36 billion checks. There are different clearing house centers. The largest one, 15 billion went through that. And AT&T calls 60 billion. Now, I said we're going to build up to Amazon. Here's the weird thing. Amazon is smaller than all of them and bigger than none of them. But it's even worse than that, because all these numbers, except for Amazon, come from 1997. Here's the reason Amazon is... 1995, excuse me. Here's the reason Amazon's not on that. That's what Amazon looked like in 1995. So people are associating big data with big, and yet you have to be able to explain why you had volumes of data, orders of magnitude larger 20 years ago before one of the companies that's a poster child for big data. I mean, there's others, but certainly Amazon would qualify. So this is the right way, I think, or this is the way I figured out how to think about big data. All right. Yeah, 2013. I think the key to understanding big data is to recognize that it's in fact five separate data technologies that have all occurred simultaneously, and then to look at how those cross-pollinate to see how they create the various solutions that you guys are used to. I'm going to go ahead and go through and give an example of each one of these, and then we'll put it all together. Language data. Language data, the definition here is getting computers to understand language, English, Hindi, whatever it is, I'm going to use English, the way we do. About 22 years ago, this young enterprising guy, Amsterdam finds a Shakespeare play in his attic. And he claims it's Shakespeare. No, it wasn't. He wrote it, but he just wanted the publicity associated with it, and how did we find out that it wasn't that way? Because we ran heuristics on it, we figured out that his sentence structure and his cadence didn't match Shakespeare's. That statistical approach to language. That's not what I'm talking about. I'm talking about you can look at language now, and you can think about it. Whether they're sad, whether they're serious, whether or not they're having an informal conversation or a formal conversation, opinion-mining and sentiment analysis. That's the big data we talk about for language and NLP. I'm going to go ahead and give you two examples. There's a hotel, it's a subsidiary of Marriott. They used to think that it took 80 things done right in order to be able to make their customers happy. But what they did was they used a bunch of tools to scour the Internet, blog posts, and they were wrong. Then instead of 80 things done right, only five mattered. And again, here's what they found. They used the data in a different way to ask a different question. It turns out that the first five minutes was paramount. If you treated people right for the first five minutes of their stay, it positively colored the entire rest of their stay even if you screwed things up. The same screw-up that occurred later in this day had a huge effect if it occurred in the first five minutes, or the next one is litigation. I know something about this. As you know when you litigate, you're required to disclose a whole bunch of documents to the other side. Boxes and boxes, millions of pages. Those used to have to be done by hand. Those can now be ripped through by natural language processors, trained either unsupervised or supervised, and figured out what is it that you're looking for. And again, they found something that they didn't expect. They thought that they would look for answers in the documents. But what they ended up discovering was the documents actually didn't contain what they wanted, but it told them where to look for the documents. And here's how it went. The email exchange would start off like this. Hey, how's it going? What happened to blah, blah, blah, the Ruby project? And someone else would be like, all of a sudden you'd ask a question? Like, is Bob still in charge of that? And the answer would come back far more formal. At this point, we're still in juncture and we're not really sure what's going on, why can't you tell me? And then he says, the last line is always call me. And that's where the real data is. And we had missed that for the last 30 or 40 years of doing massive litigation discoveries. We were so busy looking for facts in the document, we forgot to look for the indications for who really knew the information that wasn't in the documents. All possible because of natural language processing developed over the last 20 to 10 to 15 years. The next one's media data. Media data is like language data in the sense that you're getting the computer to understand data the way humans do. Now, if you've worked with Photoshop and you've had color balance, ink presses, then you know that you could use Photoshop's histogram to tell you how many blues and reds and yellows and CMYK. Again, that's not what we're talking about. We're moving to the next floor. So rather than 15 years ago, Photoshop could tell you you have a lot of greens and a lot of blues and a little bit of white. Today, we're talking about the image of the way humans do. Now, this also goes for audio. Your visual voicemail is a perfect example of that. So audio, it's listening, it's understanding it the way humans do, and it's transcribing, it's writing it out. It's not a series of sine waves, it's now language. I'm going to go ahead and give some examples here. Retail has been big in this space. This right here is a screenshot from Retail Next. They're a service provider of intelligence software that will track humans as they move through a space and see which displays are most popular. Now, I knew someone who early, early on was in charge of providing this sort of consulting service to Fortune 100 companies, and they used to hire college kids to sit in the corner and draw little schematics with their pencil and trace people around the displays. If you do it over time, you end up with heat maps like this. Drive through. This one's really clever. This one actually comes from Doug Lanier. There's Drive Through. They had security cameras that was trained at the driveway coming up, and they trained the computer to figure out if there's more than two cars, if the line is long, you should change the menu items that are offered at the window to be much shorter and faster so that the line can pass through quicker. But when no one's there, go ahead and put in the high profit, slow to cook offerings. Very clever. Again, taking data that had been originally recorded and repurposing it, thinking about it differently. Sensor data, you all know what a sensor is, a thermometer. You think it's F in cold? It will tell you. It'll put a number on it. It'll say that it's negative 15. What you might not be used to is a sheer volume of sensors that are out there. Modern cars today ship with over 300 sensors. If it's a luxury car, it'll ship with over 500 sensors. And what you're also not used to is a lot of that data is going back to the mother ship, and it's revealing all sorts of insights in the manufacturing space that previously weren't possible in R&D. It's you're co-opting every one of your users to be part of your R&D, not just in the digital world, but in the physical world. Now, I'm going to give you two examples here when we've already been over. Parkinson's disease. It used to take 30 minutes in person visit, the standard protocol, 30 minutes in person visit with a physician to diagnose Parkinson's disease. Now, you can do it in three minutes over the phone and it's free and it's more accurate. Now, I chose this one for a good reason. This one here is listening to the human voice, but it is not listening for the content of what's being said. When you call and ask you a series of questions, designed to get you to say a series of phonems, different sound forms that, and here's the insight, it turns out that your larynx degrades in the same way your muscles do. And if you've seen people with Parkinson's, you know they have tremors. It turns out your larynx will develop a tremor as well. You can't hear it, but the computer can, watch the sine wave on top of your voice. Now, notice the difference. Language data, we care about what's being said, we want to hear it as a human hears it. Are they joking? Are they not joking? Are they being sarcastic? What are the antecedents of this, that, her, it? Sensor data, Parkinson's, we don't care what they're saying, even though they have to pass through a number of phonems to be able to do it. This has been revolutionary because we're discovering that diet, exercise, and sleep has a dramatic impact on Parkinson's because you can call in as long as you want. You can call in every day or every week. You can begin to self-diagnose which activities have an impact on dementia. And it's free. Elevator when we already did, so I'm not going to go over that. Geographic data, MapQuest in the Old Days, Google Maps Today, I'm sure you're familiar with it. What you might not be familiar with it is some of the corporate uses for this. My favorite, which actually isn't on here, is Staples. Staples.com will charge you a different amount depending on where it thinks your IP address is located and whether or not you're within 20 miles of a competitor. So I encourage you to search for that stapler at home and at work, and the price will change. This has given rise to a countervailing technology called geospoofing where you can decide to be wherever you want to be in order to compile a list of all the prices so that you can just choose to be from here and not there. Careful, they might want to ship it there. Twitter since 2009 is geolocated every tweet. Now you can turn it off, almost no one does. Now we've long known using seismographs where earthquakes actually occur. But what we haven't known is the human perception epicenter of it, where people think it occurs. And this may sound somewhat trivial until you recognize, and I was astonished when I found this out, that the Richter scale is based on observation. The difference between, say, a 7 and a 9 is 9, I think, says no buildings left standing. You might have thought it had some magnitude on the little things going back for it. It doesn't. It's a descriptor. And so knowing how humans perceive the earthquake enables us to continue and refine a scale that we've used for a long, long time. I have no idea, by the way, how they distinguish between 8 and 8.1. I don't, like, it's like it fell a lot or I don't understand how that works. Another one is popular times. This comes from courtesy of Google. This here at the bottom, you use Google Maps in order to go different places, but now Google has taken that data and looked at it through a different lens. Instead of looking at one person moving to lots of locations, they've decided to focus on just one location and look at a lot of people there. This right here is being used by stores to figure out how to adjust inventory levels and schedule staff. Because not all of us have access to big expensive, but we love it when it's provided to us. This one's the hardest to explain to a non-technical audience. You've got no SQL and MongoDB and graph databases and document stores. Suffice to say, and I'll just give one example with Amazon, who here is familiar with item-to-item collaborative filtering? They published a paper about this a long time ago. Item-to-item collaborative filter. Prior to... Pattern matching has always been possible, but what wasn't possible was pattern matching in 2008. I think they applied for the pattern in 1998. It was awarded in 2001. They decided to write a paper about it in 2003. What it basically said is we have been able to figure out... You know when you go to Amazon and you buy some products and then it suggests another product that you should buy? A big component of the way it's doing that is if you've bought four products, it looks for everyone in the world that bought that fourth product and then it looks for either the most popular next product bought regardless of when it was bought. If you're going to get a fifth one, this is it. We've been able to do that for a while, but the volume that we could do it at was limited. Amazon listed as there was a movie preference selector. They had 40,000 users and about 3,500 items, 3,500 movies. Using these more modern technologies that allowed them to push past SQL, they were able to get it up to a million items and 250 million users. Orders of magnitude increased. This right here is the five different big data types that exist, but where it really gets interesting is when you start looking at how these cross-pollinate to understand the solutions that exist out there. Let's go ahead and take a look at that. Language data and media data. Let's connect those two. Who here has been abroad and used this app? This right here uses two technologies that allow you to understand the solutions that exist. First, you have to get the computer to understand, not in terms of how much red or how much blue I'm going to print, but you have to get the computer to understand what is it that you're looking at? What's a dog and what's a sign and where are the letters? Find the letters and please pass them on. What are you going to pass them on to? A natural language processor that can understand what's being said in a human sense. Let's do this to you guys, some of those screenshots. Let's do another one. Sensor data and geographic data. Another retail example. I'm not going to ask who's been into American Apparel, but you've all heard of American Apparel. American Apparel has what they view as a competitive advantage, which is they have figured out how to sell cheaply in urban spaces where rent is a premium. And what they've done is they've absorbed the high cost of managing their inventory to a ridiculous level at the point where they can now occupy smaller spaces in their competitors. They like to call themselves a boutique shop and a component of that is they maintain that they only have one item of every size, color, and style on the floor at any one time, which of course means that the moment that someone comes and picks that off the shelf, you can no longer sell that again unless you replenish. Average replenish time at American Apparel is 45 seconds. And the way they do it is they are as you pull it off, they know you pull it off, as you go through checkout, they know you go through checkout, and in the back there's little lights and they pull down the items that have just been indicated as being marked and then you run them by a little scanner to indicate that they're in process and then you hang them out. Now, that costs, but it apparently costs less than the high cost of rent in the areas that they want to be in. So they are tracking the location, i.e. the geography of their items in order to be able to make leverage This one's everyone's most interesting favorite. This right here, sports metrics, this right here comes from Sportsview. This is a new revenue stream for the NBA. The NBA owns and installs and runs in all 29 arenas because I think the Lakers and the Clippers both play in Los Angeles. 25 cameras, it tracks all 10 players, all three refs in the ball, 25 times a second, and if you watch closely you can see they can tell when they're getting ready to shoot or if they're going to put it in their left hand or their right hand, and you can almost see which way they're gleaning. They sell that now. And now, using the pattern matching technology that I described for Amazon, you're able to ask questions like of the 11,000 times that LeBron James has come down the court and he happens to have two teammates with him and he's against three defenders, does he dribble left or right? You can ask that now. Now, there's different ways to do different technologies. So, for example, this sport track tricks I've done, geographic data and pattern matching, that particular one also could have a line going to media data because it was done with cameras. FIFA went in the opposite direction. Their fields are so large that they decided to put RFID tags on the back of their shirts. That's the way they monitor location. So when you start pulling big data onto its constituent pieces, you can start seeing the choices you have about sculpting the particular solution that you're interested in and different solutions apply in different contexts. Sensor data pattern matching. This is being used for predictive maintenance. Something happened about four or five years ago. GE, they may coal-fired plant generators and locomotives. Really industrial stuff. Here's what's crazy. They now make more of their money from service and Jack Imelt, their CEO, says, we no longer sell locomotives. We sell locomotion. They won't sell you the item anymore. What they'll do is they'll rent it to you. What's really being sold is some quantity of thrust power going from north to south on some railway that's going to pull this much coal or this much freight. And GE is now responsible for just making sure that engine works. The origin of this was when GE first sold the engines and then other people came in and ate GE's lunch by providing service contracts. So GE started doing service contracts. Then a bunch of enterprising young data scientists in New York started plugging their little laptops into the data maintenance ports for their locomotives downloading all the data and started looking for the patterns for when locomotives tend to die. Those locomotives produce about a gigabyte of data every hour. And if you look for all the patterns when you can most accurately change out parts or ask for a maintenance stop. So GE stepped into that game. Now, as you imagine, big data is going to improve GE's ability to do this over time and they didn't want those lower cost to be passed on to the consumer. They wanted to capture that themselves. So they no longer sell their locomotives. They'll rent them to you and the lower cost of ownership over time is going to accrue to GE. And it's working because they're now more of a service company than a manufacturing company. All right. I just wanted to put this one up here. A lot of people, they get confused about the relationship between big data and whatever small data might be, structured data, there's different names for it, legacy data, whatever you have. Amazon's a good example. Remember, the way in which Amazon figured out what suggestions to make for you. Where did that data come from? Well, that was transactional data that was sitting in SQL servers. I mean, there was a product and then you took their credit card. All of that was sitting in standard typical databases since 1971 at IBM when Cobb invented SQL server. But you had to move it into a new form to be able to see patterns in a new way that SQL server wouldn't let you do it. That SQL wouldn't let you do it. Face analysis. This woman here has PTSD. This is a project funded by the Department of Defense. And we know that because her muscles don't smile as long as people who don't suffer from that disease. We're now up to about 16,000 data points on the face in order to be able to do facial recognition. And the University of New Zealand study, they did something kind of curious. They wanted to see whether or not they could fool the face recognition authentication systems in various computers. And so they got 16 pairs of twins and they gave them the same hairstyle and the same shirt and they put them in identical twins. And the computer nailed it. Every one of them could tell the difference. That's all possible because we're beginning to understand and being able to see human faces the way humans do and recognize what a smile is. What's a frown? All right, so let's go back to this here for a second. Now, what I want to say is that I think the best way to understand big data is understand that it is in fact a collection of five different technologies and there's some benefits about in a second. And based on that and using that, I'm now going to give my definition of big data, which of course I think is correct. First thing we want to do is that we want to put this top one down at the bottom because that's the worst definition. So let's go ahead and start at the top. Thomas Davenport said big data is the new and massive data types that have appeared over the last decade or so. Yeah, as annoying and disappointing and ambiguous as that is, he is absolutely right. That in fact is the best definition of big data, but it is so ambiguous What in fact you should say is big data is the new and massive data types that appeared over the last decade or so, comma specifically language data, media data, geographic data, sensor data, and pattern matching or associated data. There's different ways that you can think about it. Now, I totally understand why marketers have stepped in and co-opted this term to basically mean it's the product we sold last year, but it costs more now because it has big data attached to it. I totally understand it because that is a very terrifying answer and it's way too long and it almost requires that you have someone like me stand up and talk about it for 20 minutes before you're able to understand what's really going. But in fact that is what big data is. This second one here I think is exactly right. It is about the analysis of data where you don't know the questions to ask and you look for patterns or numbers and that's what happened with a whole range of things, whether it was the elevator or whether it was Parkinson's or whether it was being able to use Google to understand what's happening in your small Shake Shack, all of these are different ways of looking at data. And I tend to agree with Ken Kukia, yes, you will see dramatic results as a result of this. And this last one, I just hate a lot so we're just going to get rid of it and I'm glad it's sort of dying and I'm actually very pleased that very few people, not as many people as usual, raise their hands in response to this one. So that's what I like to say is that the keys to understanding big data are two-fold. One, at a technical level, and also at a skill level. Gone, if they ever existed, were the days where you could get a data scientist, just put them in a corner and do amazing things. You cannot ask them as a PhD and NLP to start running GIS systems, GIS systems, nor can you have someone who's good at time series analysis for low level sensor data and then ask them to start working on image processing. Those are different skills and organizations that recognize that we want to specialize in certain particular types of big data technology and we can also benefit from recognizing that there are certain skill sets that go with that and those are the ones that you actually want to hire for. And the second part of big data is that you want to start, you want to co-opt from big data, the creative thinking that a lot of big data players are using and you can even use it on the data that you already have. For example, again, the elevator. They had that data, they were sitting out. People don't even realize the full volume of data that they may have just because they don't consider it under management but that data is there and it is available. So usually I talk about the ways in which organizations can start using big data but we're out of time. So I'm just going to say thank you and I'll be available for questions.