 Welcome back everyone, we're here live in Las Vegas for IBM Impact. This is SiliconANGLE's theCUBE, our flagship program. We go out to the events and extract the ceiling from the noise. I'm John Furrier, the founder of SiliconANGLE. I'm Joe Mike, co-host Paul Gillan, my co-host. Our next guest is James Taylor, CEO, principal consultant, decision management solutions author of one of the best selling New York Times, best seller books, decision management systems, a practical guide to business rules and predictive analytics from IBM Press. Welcome to theCUBE. Thanks very much. Don't you wish it was, does someone die at the end? I mean, come on, New York Times best, I'm only kidding, obviously, but welcome to theCUBE. Thank you. It is one of those riveting books. I mean, IT right now is pretty riveting. I mean, I don't, I say that, you know, seriously, it's one of the most dynamic times in IT where there's top line business focus on revenue and business growth, not just bottom line cost reduction. There's some significant converging trends happening that are really making IT different, you know, faster, agile, more productive. Now with instrumentation and big data and unlimited compute and cloud and power system, I mean, literally everything's instrumentable now. So that's changing the game. For sure. What's your take on that? I mean, the book is a couple of years old, but now still the big data trend was in play there. Yeah, I think what's really changed is the acceptance of analytics. When I wrote the book, predictive analytics was still a little bit, you know, I'm not sure about predictive analytics. And there were ways to use small data that weren't predictive analytics. But once you start getting into vast amounts of data that arrives really quickly, if you don't process it in some fashion, if you don't turn it into some kind of usable prediction, then it's very hard to consume it. I mean, you can put it on someone's screen, but then you just drown them in more and more noise and less and less signal, right? So you have to extract some kind of useful predictions out of your data. And so predictive analytics has gone from being a minority occupation to being like front and center, accessible inside, everyone wants it. So that's really changed since the book came out. You know, when I was in my youth doing my internships in college, there was departments called DSS, decision support systems. You know, computers have always been helping people make decisions. Like what scale and what pace and what data was a whole different discussion. So, but let's talk about today, right? I mean, you know, what is the criteria now for businesses to make these changes? I think it's pretty clear to most CEOs and C-level executives that, you know, getting more data about their customers, their business and instrumenting their value chains is critical. Right. So what was the big change? I think what's really changed is if you look at the, there's a great Peter Drucker quote. He once said, there's a tendency to believe that only senior executives make decisions or and the only executive decisions matter. And this he said was a dangerous mistake. And I think what's happened in the past, all these DSS systems, they were aimed at individual managers, knowledge workers, executives so that they could make big important decisions. I think what's coming up is particularly with this sort of move to real time and mobile apps. You know, I have a 25 year old, when he gets car insurance, he expects to be able to give you his data and get a quote. If you say, well, you know what the quote is, he's like, I'm going to go talk to somebody else. You know, I'm just not interested in waiting for you to decide what my rate is. I want to know now. So you have to respond in real time to what is a relatively complex question. You've got to use analytics to predict how risky he is. He's 25, so is he really a typical 25 year old male? High risk or not? So you've got all this stuff to do, like that. And so you have to refocus. That's why it's decision management systems or decision support systems. You've got to refocus on a different kind of decision and on how you can use all this data that's streaming in in real time to make an immediate decision and then embed that into your operational processes, into your mobile apps, into your website. Because if you don't, then you're missing the point. Your book outlines four principles of decision management. And the first one is begin with the decision in mind. And that struck me as kind of obvious. Isn't that how you would go about this? But obviously there's a reason why you said that. Did you find that people typically don't? Yeah, what I found, you know, the reason for misquoting the late Stephen Covey there is really two-fold. The first is that when you look at decision support systems, there's obviously a long history there. People are often very unclear what decision it is, in fact, someone's going to make with the data. So they're like, well, you're making kind of decisions about supply chain. So we'll put a bunch of data about supply chain in and we'll put some graphs and some widgets and let you drill down. But we haven't really ever thought about what decision it is you're making. And we don't need to because you know what decision you're making and you're a smart, intelligent, well-educated, experienced person with time to make the decision. But what happens when I try and make a decision about what offer to make a customer who's on the phone to the call center right now? Well, they've got seven seconds and they were hired yesterday and they got three hours of training. So they're not in a position to know what decision they should be making. So if you don't know what decision you're trying to embed the analytics into, you can't do a good job with the analytics. You know, so I put it in because there was this sense that people were very laxadaisical about what the decisions were they were supporting with decision support systems. And then if you wanted to build, you wanted to embed that decision making then you had to really begin by saying, well, what exactly is my decision and how do I want to make it? What are the moving parts? So what's the balance between policy and analytics and add add? And that was a big shift for people. Another big shift is another one of your principles which is be reactive, excuse me, be predictive and not reactive. I think you're right that the reactive decision using data to support decisions reactively is more intuitive. What is the mind shift that's involved in moving toward predictive analytics? Yeah, well part of it is being much clear about what the decision is because it turns out to be one of those things where it's very easy to build things that are predictive but if they don't change people's decision making behavior, they don't help. And if you talk to anyone who does predictive analytics they'll tell stories of building highly predictive models that didn't change the business. And so you have to be clear how it's going to affect the decision you're trying to make before you can build the predictive. But the keeping is this sense of you're moving from absolutes to probabilities. If I'm measuring last month's results I can give you an absolute number. I can tell you exactly what you sold last month. If I'm trying to give you demand for next month I can't do that. I can give you a probability or a range. And so you have to start dealing with a little bit of uncertainty. And that's why it's important to wrap through some rules around these predictions. So you can say well if given this prediction and how accurate or not it might be here's what I'm going to do with that prediction. So you have to move away from this absolutes this mindset to this much more what's likely what's unlikely, how probable is something. And that's a bit of a shift for people. They get uncomfortable running their business on probabilities. But in fact it's what we do subconsciously all the time. But we use experience. We use experience. And we're just using replacing that with data and with compute power so that we can make that same patterns spotting that same ability to recognize a pattern that's happening and deliver it to people who don't have that experience and don't have that time. And to systems, mobile apps, websites but I never going to have that experience because they're computers. There's a sluice startup getting funded around business intelligence and data warehouse that market's shifting. We're seeing a lot of real time, actual insights, a zillion dollar funding, valuations in Silicon Valley. But in particular, the role of folksonomies. Now you mentioned taxonomies in your book as obviously key to success and database roles deal with taxonomies very, very well. Structured data. When you deal with unstructured data how does this new loose data or data that's more a folksonomy approach. How does that affect some of the opportunities? Because certainly it gives you with no SQL databases you can now store a lot of data that officially had no schemas. Yeah, and so one of the key things is the ability to store data before you figure out how you might use it. And that's a key advantage, right? You schema on read, right? I can stuff it away and then when I want to come back to it I can apply a schema to it and extract some data. What I'm seeing right now I did some surveys recently on predictive analytics in the cloud for instance and we asked about some of these big data sources. And what we found is that the people who were getting value from big data and from these unusual sources were people who had some experience with more advanced kinds of analytics on structured data. And then what they were doing is they were going and saying, okay, we're already getting this level of accuracy out of our structured data. We're maybe segmenting our customers into a dozen customer segments and it's pretty good. But what if we could bring in the stuff they tell us in emails and in texts and in unstructured data? Could we make a more fine grain model? Could we be a bit more accurate? And so it's ability to refine existing predictions is really strong. And right now that's the biggest use case I see in customers that they've already got some analytics working on their structured data. And they're starting to say, can we get better at this by applying these things? You know, against that, you have some other folks who in theory could do what they need to do with their structured data. But in practice, their structured data is a little bit iffy. I mean, I like to pick on Zappos in this one. Zappos always asks you, is the shoe true to size, too big or too small, right? Well, but they already know that in principle because if you send, if you always buy a size seven and then you send this size seven back and buy a size eight and three or four people do that, then you know that shoe is small for size because you know from the returns, right? The structured data tells you that. Well, it turns out that wasn't data that they had in a format that let them access that. And so it was easier to ask you to tell us. A lot of people I talked to say they want, they'd look at business intelligence in this modern era for real time to be more like a Google search engine that after a few iterations of querying, then you find your question. So like, there's no magic question. Now, if you know stuff in advance, that's good, but if that's the user experience that's coming around the corner, how can people prepare for that? Yeah, I mean, I think that is. The real time pressure is pushing analytics into these two very bifurcated environments. There's the, I want to have like a Google experience. I want to just write my question and have you come back with something sensible? And the other extreme is I need a real time response and there's no human involved at all. I need the mobile app to respond. I need the website to respond. I need the call center rep to say the right thing. And that means I need to give them a script that was driven by analytics. Well, that requires embedding predictive analytics models into very highly automated decision-making systems. The other one is a much more complicated problem because you've got to balance this fact that you don't know what the questions are with some ability to learn what kinds of questions you might want to structure for. And that one, I think we're still figuring that out. I think search has a lot to go on there, but understanding what a question is and how questions might be broken down. And what's the role of data science in all of this? Honestly, data scientists is a hard trend. Will data analysts be data scientists or data scientists be regulated to the Python script or the geek math job? Oh yeah. I think there's going to be, there's a lot of answers to that question, right? I think, first of all, data scientists will not be, there won't be 100,000 person shortage of data scientists any more than there was 100,000 person shortage of programmers or lift operators or telephone exchange operators before, right? Because we will automate enough to fix this. I think we're going to see a range of people from business people who need to do analytics, business analysts and data analysts who need to do more advanced analytics, data scientists doing very high-end, complex analytics. I think we're going to see a lot of machine learning, making all of them more able to do more advanced analytics more quickly. And I think we're going to see far more analytics embedded in the system so that people with no analytic skills are able to consume analytic insight, even if they can't build it for themselves. So I think you have to be planning on a broad range of analytic capabilities. You know, yes, there's a maturity curve, but as you move up it, you don't stop needing the things that will lower down the curve. You still need, you're going to need a broad landscape of analytic capabilities. That's where it's going to end up. IBM has a machine that's really an unstructured data machine called Watson. And it sort of does what you were just talking about. It helps you to decide what questions to ask. So as a data management or decision management expert, when you look at Watson, what potential do you see there? You know, I think Watson has some tremendous potential. One of the big things in decision management systems is this idea that they're adaptive, but they learn what works and what doesn't work. And I think, you know, one of the challenges comes in a lot of systems is that the, how you tell whether you made a good decision or not is often a fairly qualitative thing. And I think, so one key thing is being able to plug the results of that, decision making back into something like Watson when the results are not described in numbers and be able to say, well, what did work? You know, if we apply everything we learned about how well we did, what's working, what's not, that's one clear use case. I think it's also true that Watson has some very strong use cases where there's, where we've been making human decisions a long time, like medicine. Because there's lots of stuff written down for human consumption. And computers don't consume it very well and we don't have the sort of structured data equivalent of it. We only have the written version. And so Watson's ability to consume all that written text, munch it and make analytical decisions is tremendously valuable. So I think it's got some very strong use cases but I think we're going to see them converge with more traditional machine learning and structured data analytics. I think there's, you know, the interesting one for me is how we use them together. And that's I think still an emerging space but it's going to be an exciting one. We're seeing in the marketing profession which is something I know more than many people about, I guess. We're seeing panic right now over the need for, it's becoming a very data analytics driven profession. I think we'll see that in a lot of other fields as well. We're just talking about HR in an earlier interview on theCUBE about the data analytics is applied to people. What skills do you think that people in going into college right now should master as they prepare to move into these traditionally more qualitative or softer professions? I think one of the key things is what you might call a feel for statistics. Not an ability necessarily to do the statistics or to do the math but a sense of how one does statistics. It was a great illustration of this the other day. Someone surveyed a bunch of medical professionals and said there's a disease. One in a thousand people get the disease. You have a 95% accurate test. Test comes back. You have a positive result. How likely are you to have the disease? Most of them said 95%. And the answer is 2% because 5% error rate are 999 people who don't have the disease. It's 2%. It's 98% of the positive results, right? And so in fact when you get a positive result that's not super bad news, right? It just means you've got a 2% chance instead of a 1 tenth of a percent chance, right? So it's good data but far too many people don't have that kind of basic feel for how statistics work, how probabilities work. And without it you're not going to be able to work with these huge data sets. That's one example. They're just going to overwhelm. Jay, talk about the dynamic going on in the market with data because some companies are swimming in data. Some companies have been swimming in data but kind of park it out with a data warehouse and not really aren't swimming anymore. They throw it away. They put it in the, you know, in the hinterlands of a spenced organization. Then you've got companies who sell stuff, have no data experience. They sell on basically software. They're infrastructure. So you have people making solutions that have no data experience. They're not swimming in data. So is it just the evolution of the marketplace or will you see new solutions come from people who are used to dealing data? For instance, Facebook is swimming in data. So hence they did a lot of innovation around that. Yeah, I think you'll see both. I think you will see companies that are good at working with data try and expand it. Much as you saw Amazon get good at infrastructure and start selling infrastructure services, I think companies that get good at data will start selling those data services. I think they'll see that as an opportunity for a new line of business. But also, you know, I have a thing about sort of data. You know, I often hear people say, oh, you need to be out there looking at big data because maybe you could be the next LinkedIn or the next Facebook or whatever it is. And I'm like, you know what? I remember years ago having this conversation with a guy and he worked for a big paper goods company. And he said, son, I was young at the time, said we're a paper goods company. That means we turn trees into toilet paper. I'm not going to buy your software unless it shows me how I got to get better at turning trees into toilet paper. And I go to most big companies, whatever you do today in 10 years time, that's still going to be what you're going to be doing. If you're a paper goods company today in 10 years time, you'll still be a paper goods company. So figure out what it takes to be a good paper goods company. And then go look- Well unless they're in the publishing business and they won't be using that paper anymore because that's a whole nother discussion. But, and then go say, what might I be able to find in these data sources that I can buy that I've got stashed away? That would help me do that. But begin with the decision in mind, don't just go, you know, hope that the data is good. So you're saying, stay within your core competency and use the data to differentiate. Exactly, right. Say, what would, you know, and one of the tricks I use with clients when I'm interviewing them, I say, right, you know, answer the sort of if only question. Well if only we knew which clients were actively considering a competitor, then we would behave differently to those clients. Okay, well maybe we could go look at big data. Look at some of these unstructured data sources. Look on social media. Maybe we could find some signals out there that would say, these guys are looking at competitors, these guys are not. Yeah, so you ask that if only question, but do it for a decision that matters to your business. But if you don't know how it would change your behavior, you don't know what decision you would make if you knew it, then it's probably not going to help you to know it. Okay, final, you know, you're going to be done. Final question for you. Here at IBM Impact, actually, they got all the messaging right online. They're looking good. Cloud mobile, social, actionable insights. But they are really rolling out some stuff. So what would you share with the folks out there about why IBM is in a good position right now or not in a good position? What do they work on? And what is the role of this new infrastructure and how do they enable to use that if only, answer that if only question? So there's a couple of things. I mean, IBM has a broad range of analytic capabilities. It's got Cognos, it's got SPSS, it's got InfoSphere Streams, it's got Watson. It's really built out a robust analytics portfolio. And I think that's important because I think as organizations, you've got to be realistic. That you've got people who need very straightforward help making decisions right up to real-time streaming data that needs to be acted on. So working with someone who can sort of blur those things, that's important. I think, you know, my experience is that you've got to embed these things as services. You've really got to manage them as services. So the whole SOA, SOA infrastructure piece, that's important. Again, IBM is a strong player there. And they're being very aggressive about moving that to the cloud. And for me, in decision management and analytics, decisions as a service, analytics as a service, those things are great ways to make analytics pervasive. Because the big problem is, it's no good if it's only working in one system or in one little silo, it's got to be everywhere. And the cloud is a great way to say, here's a way to segment our customers. Okay, now we're going to use that absolutely everywhere. Every time we interact with our customers, the first thing we're going to do is segment them accurately using this analytic model. And so cloud is a great infrastructure for that. So I think IBM's got a lot of the pieces. They see how they fit together. And I think they're trying to help their customers move. And they're used to dealing with big, slow companies. And many of us work at big, slow companies. We need someone who's going to be patient and work us through the steps. And I'd be well. James Taylor here inside the cube, final word. What's the next book? What are you working on now? What's around the corner? What's got your attention? What's got my attention? Right now, it's the role of building this kind of decision-making into the back end to improve mobile. Because mobile's hot, everyone's trying to do mobile. A mobile with your old current back end systems? It's not pretty. You need smarter back end systems in order to make smarter decisions using mobile devices. So deliver them to mobile devices. So deliver to mobile devices. Because your mobile device can't just be your enterprise app. Screen small, the personalization, this is low latency. Where are you? In fact, I'm speaking on this very topic tomorrow here at Impact at the end of the day. It's got all your attention. There we go. You'll be having full attention on stage. We are here. We are the backend for the data here at IBM Impact. This is the cube, our flagship program. We'll be right back with our next guest. Thanks, Phil.