 This is Dave Vellante of Wikibon.org, I'm here with my co-host, Jeff Kelly. You can check out his work at Wikibon.org slash Big Data. He is the primary principal Big Data analyst at Wikibon. He's written a lot about this topic, a lot about visualization, and has just put out a new study on the whole SQL, Hadoop, NoSQL rather, marketplace, so check that out. We're here at the Tableau Customer Conference, we're at the nation's capital. Phillip Kim is here, he's the marketing operations leader at GE in the measurement and control side of the business. Phillip, welcome to the Cube, thanks very much for coming on. Thank you for having me on. Yeah, so tell us a little bit about what you do at GE, obviously a big complex company, we're talking off camera, we were covering the industrial internet announcement. You guys are an enormous conglomerate, so where do you fit into the whole thing? Right. We're in a division of General Electric that specializes in what we call measurement control or the industrial healthcare, so a lot of the customers' assets that we monitor are the assets that power the economy, so power transmission lines, oil and gas pipelines, things that really are vital to the infrastructure of the United States or the countries around the world. We provide equipment and services and software that helps monitor that. Okay, so where does the data piece fit in and what's your role in all of this? So we collect a lot of data from a variety of different sources, including as you mentioned before the industrial internet where we're trying to figure out how do we get productivity into the intelligence of our software and systems. The group that my team is a part of is really around the commercial intelligence. How do we identify growth areas using the data sets that we have available? How do we identify where the cost savings can come from a customer's perspective? And we work with collaborate with a large cross-functional group to take very large disparate data sets and boil it down to a single visualization that might help answer a very important question. So Amit, am I correct? You're looking for both market opportunities and cost-saving opportunities? Yes, absolutely. And as well as productivity. So how do we make things a little fast, a little easier for the customer? How do we simplify things? So we will work with quality organizations within our own business. We'll work with customer sales and marketing representatives. We'll work with anyone basically who has a compelling question that data could help answer. So you've been working with data for quite some time, I would imagine. So this big data meme must be kind of tongue-in-cheek to you, like big data, big deal. I've been working with data for a long time. But what's different now about this whole data discussion? What has changed? Is it the volume of data? Is it the power of data? Is it the economic value of data? I don't think there's any one answer, but I think that's a great example of how many things can contribute to why it's the right time for big data. I do believe that the democratization of this accessibility to data is a major driving factor. Tools like Tableau are a fundamental piece of making it easy enough for a user to interact with a very large data set. And I think also because the business has started to recognize that there's a lot of value that can be retrieved if you know how to get it right in the right structure. So Stephen, you mentioned being able to bring together a lot of different disparate data sets and distill it down to one visualization and really start to make sense of that data. Now, if you take that in a more traditional scenario, that would take months to put together a data warehouse. You would have IT build maybe an app, dashboarding application. Eventually roll that out to your end users. Talk a little bit about, as you just said, tools like Tableau democratizing data, making that process a little bit easier so that you can actually get value from that data while it's still fresh, while the insights will actually allow you to take some action and monetize that data or drive other value. That's a great question. I think the first thing I would say is IT is actually part of our extended broad team, so we're partnered with them on this type of effort. What we can do though is we can kind of be on the leading edge, the prototyping innovation investigation stage. And when it's ready, the IT team actually has a built-in template to do a much more comprehensive rollout. But what allows us to do is we don't have to go to the IT team and bother them with very, very quick iterative prototyping steps that may not pan out to be quite honest. And so what we can then do is use really our innovation to find those patterns, find those distinctive, I would say applications of that. And then once they've been proven out, we can then apply that back. And actually our IT department uses Tableau to productionize it and put it onto the Tableau server and make it available to the rest of the organization as a whole. So we're really more of the prototyping agile development stage. And then we transfer that ownership back over to the IT department as well. So it works both ways. Well that's interesting because I have been to this conference before and a few years back there was some, there's what you heard from IT was a bit of pushback from what Tableau was doing because maybe a few years back they didn't have as many of the enterprise-grade controls around governance and security. It seems like Tableau has come a long way in terms of adding those types of capabilities and IT is actually starting to embrace the tool. And in fact as you mentioned actually GE actually using it themselves. Well I think that's another great example of the adoption curve that you typically see with high technology products. I think it was initially designed for people like myself and my team. But then over time obviously as the product has matured it's become much more accessible to a larger group of folks. And I think on top of that the business need has been really championed by executives who've seen the value of doing this in this way. So by partnering with IT we can get more done. We can scale. And in an organization like General Electric we have to be able to scale to have some sort of value. So I wonder if you could talk about some of the organizational issues that you've faced. You mentioned that IT now is embracing tools like Tableau. But initially there might have been some tension. So I wonder if you could talk about that dynamic and as well from the users that you serve us. How did you drive adoption? What about training? Things of that nature. Start with the organizational tension and how you got through all that. That's a great question. It may not necessarily be as the people do us. We actually had a tremendous buy-in from the IT organization. So we didn't actually have any pushback of this tool per se. It was a love-love relationship. It was a love-love relationship. It was much more about the fact that we had a blank slate. We were starting literally from ground zero. And so we sat down together as a team and said what is the good use of this amazing technology that would potentially drive significant value for our organization. And so what we wanted to do is focus on what would we want to do with this technology. What we want to do with this relationship. And we said well we want to help grow obviously the sales of our company which is really driven by adding value to the customer. We want to make sure that we save valuable time because obviously everyone's time is limited. And so we're very focused on how do we make a day-to-day life for a person using Tableau much better. And how do we make their day a little bit less and truth a little less difficult, less challenging. And then we also wanted to figure out okay well there's a lot of stuff that we don't understand yet. We're going to be very quickly prototyping and iterating. What environment lets us do that? And so we sat down with IT as a partner and we figured out this is actually a pretty good solution set for what we're looking to do with it. So that the small team that we have within my group and IT was actually a love-love. We focused very heavily on the core things that were important to us. And then what we ended up doing was setting up these core pillars of our strategy and then communicating across general electric was really the incremental step that you might see in terms of organizational change of behavior. Phil can you talk about the different textures of data, the diversity of data, the data sources. Did you have to think through that from an architecture standpoint? Do you have a data architecture? I wonder if you could talk about that a little bit. I think everyone's data architecture you might have alluded to it I think or Jeff you might have alluded to it is largely based on Excel. There is no real data architecture that I've seen even with big data that hasn't been fragmented. And I think what we are conscious about that is we're never going to be able to stop fragmentation of data sets. What we have to be able to do is figure out how do we intelligent process what's important to us? Data warehousing, BI will continue to evolve. We'll never stay ahead of that curve if we want to deliver value. So what we want to do is focus on instead of the data first we'll focus on a question. And so our approach is very simple. Before we do any kind of analysis we ask what's the question you're trying to answer? What's the answer look like relatively speaking? And what's the benefit? That is if I do a question I give you the answer what are you going to do with this that's going to make a difference? So our question answer benefit structure is really one of our fundamental pieces that we use whenever we get into any kind of project. And then what we might find out very quickly is that there is no data, there is no architecture and therefore we might not be ready to try to answer this question yet but there are other ones that might be much more valuable that it may be a little orthogonal. So that's interesting, so I have to ask you. So a lot of data practitioners will tell us that a lot of times you don't know what questions you want to ask until you see the data. So what's your philosophy or maybe even recommendation for fellow practitioners? When you don't have all the questions and you look at the data and you say oh wow it'd be great to know this or that or that. Should you go after that or should you rather focus on the corpus of data that you have and extract value from that? That's a great question. I don't think there's any one way that works for everyone. What we've done successfully is use expert round table opinion. What we'll do is we'll have a brainstorming session and we'll say hey, before we get into the actual tactics of deciding to go one path or the other, let's first figure out, give us everything you think would potentially be good questions, good answers, wish list and just walk it down. And when we finish then after about five, 10 minutes you start realizing the run out of ideas. So you've got a quick snapshot. You take these ideas and you basically figure out what's the impact and what's the feasibility. Which I think is a correlation to how good your data sets might be, whether you can investigate and so forth. And what we'll do is we'll structure that to figure out the top two or three that we can answer in a very short time from two to three weeks. And using that we'll get a pipeline of ideas that we know we can go after and we might have a very strategic question to answer that is very hard to answer and we might just have a spike where we'll say figure something out. But we might have other very specific deliverable to say give me this view, because this view will translate to this type of financial forecast or this type of campaign strategy. So it's a very pragmatic and practical prioritization. That's some alliteration for you, Jeff. Exercise that you go through, pick off the ones that are easier to do and can drive value. And then if you've got something that's more strategic you got to make a business case, maybe get an investment, maybe find new data sources. Absolutely, and we try to keep that discussion as minimized as possible. The reason why is we don't want to spend too much time talking amongst yourselves. In GE, meetings can spill over into other meetings, which I think is true of any large organization. So what we want to do is in a core stakeholder room decide what those top 18, 20 might be. Figure the top three or four that will make a difference in a very short timeframe and then focus that on. And what we'll do is we'll bring in data, we'll bring in the BI folks, we'll bring in the IT folks, we'll bring in Tableau and we'll see if we can make sense of that. But we'll find our sweet spot and we focus on those. So let's talk a little bit about the Tableau piece, Jeff. Yeah, so we hear a lot from Tableau that their goal really is to unlock the power of data for the average business user. For somebody who's not a data scientist, doesn't have a lot of training and statistics and other kind of disciplines around manipulating data. So just from your experience with your team, how easy is it? Is it as easy as dragging some data sources into a field and while you've got some visualizations or does it require some level of training despite the self-service monitor? Well, I think I can best answer a question by giving the example that what we've done in our team is help lead a larger organization within the business that isn't trained in statistics and they're actually becoming the power users that are actually driving a lot of the value that we're seeing in some of the projects. So the project that we've talked about in my organization, fully 50, 60, 70% or more are now being driven by non-scientists, non-statisticians, people who are not in that big data space. So I think that answers your question. But certainly, it's an evolutionary process. I think they always are trying to get better at it but it makes it easy enough where they can get to that level. So take us back to when you brought in Tableau. What were the drivers? What did the environment look like before and after? So the reason why we brought in a Tableau and used it as a platform for us to do our analytics was that we were spending too much time with the final end product which what the users cared about. We were spending far too much time on the data side which didn't translate enough in benefit. We would spend so much time on a data mining project that we couldn't have a product at the end of the day. And what we wanted to do was have a template, a factory that we could use to drive data mining. And so Tableau fit that very well for us. And without a lot of training and a lot of the factory, sorry, the templates that were existing in Tableau today, we could stitch together different data sets, do the blending in or outside of Tableau, but ultimately end up with a fairly well finished product without a lot of the interactivity built into it very quickly. So am I to understand you were just spending too much time cleaning data or interpreting data? I think getting a handle on it which was I think the first step. What do we have? What do you have? That's always the first challenge of any data mining project is you're confronted with something entirely new. It's not something that has already been cleaned up in stage, it's a new problem, new question statement. So now the question is well, how do you get your hands around it? And Tableau is wonderful for taking in large, disparate, complex data set and then you could just start playing around until you start seeing something that might be interesting. I think that's the first step to exploring the data set. But we typically have that question with that exploration kind of dovetailed in. But it's a really intuitive environment for doing data mining. Did you have to set up any kind of particular, again the organizational question, any particular training regimen for your users or how did you handle all the knowledge transfer there? Actually Tableau came in and held our first training session for us. We had some really great help from our account rep and on top of that, I think there's a tremendous community of Tableau users. I think that's one of the selling points is that if you have a question and you don't quite know how to do something, you can actually go online and find it in one of the boards pretty easily. Philip, kind of even taking a step back from that. Talk a little bit about your employees or your team and the idea of being data driven. That kind of is beyond just the tool you're using. It's a mindset, right? How do you kind of instill that kind of mindset? Do you look for certain characteristics when you're hiring? We hear in some organizations they're starting to adopt business intelligence and data visualization technologies that it's like pulling teeth to get people to do things in a new way. Because it's scary, change can be scary. And doing things when you've done it kind of intuitively for so many years. Sometimes when you're looking at data driven methods, it can be a little bit of a challenge to get people to adopt that style of decision making. Did you have that challenge and how do you go about kind of instilling that mindset in your team? Well, I think we have one advantage in general electric is that we are a very data driven company and we have Six Sigma as the backbone of a lot of the core of all the executives, all the leaders have been fully well trained on it, including all the green belts, the people who just have a very solid understanding of data. And I think what we wanted to do was, how do we take that to the next step? How do we actually translate this core skill set into something even more powerful? And initially that transition point required leadership. It took the buy-in of our CEO and president, Brian Palmer, our chief marketing officer, as well as a number of folks who saw the potential of it. And I think they gave us the room to experiment, to try out this idea with some key stakeholders. And then I think success breeds success in that we had been successful with a few projects and then suddenly more and more people realized, wow, this is a very quick way, it's a non bureaucratic way, it's a very organizationally simply way to get something done that they might not be able to do in the past or might be a little faster, getting more data. People I think welcome good data to make their decision. And so as long as you get to that level, I think that's what you're shooting for. Philip, any big surprises from your experience in bringing in Tableau and this whole journey that you've been on? What surprised you the most? I think the hardest thing for anyone to get into in this field is you got to have, I think as Jeff alluded to, you got to have a passion for it, so we like surprises. We like to aim for surprise. And if we're not surprised then we're probably not pushing ourselves too hard. So every day for us is a little bit of a surprise where the data says something that we didn't entirely anticipate. We might have found certain key indicators that are a good predictive performance of our economy and of our business. We might find that there are hidden opportunities where things that are driving up costs for us are driving up costs for our customers and we find those things and raise those. And do people, they didn't believe you at first, I would imagine, right? You had to win them over, right? I mean, Nate Silveron later, of course, you know the story, there's a lot of controversy, right? People feel like he was biased or whatever it is, but the data ultimately never lies if you can interpret correctly. So did you get a lot of initial skepticism? Like, well, maybe there's a bias in the data or for somebody that maybe the business decision wouldn't favor their agenda, it would try to discredit the methodology or did you get any of that or how did you deal with it? Absolutely, and I think that's true of any organization where the first thing that people ask is, well, how did you get to this conclusion? What's your basis for this? And our business is fairly complicated and so what we wanted to do very closely with our partners across our organization is help validate that what we did made sense. So we worked very closely with our internal stakeholders. We have tremendous people, tremendous experts in the business and in the industry and they could help validate what that was. So that was the first hurdle, that we wanted to get the experts in the industry, folks that we have within our own business saying that kind of makes sense, they help write us the stories that we would want to satisfy. That was a big lever that we had to use to get adoption. The second I think is, you know, how do you give people to act on it? And that's the benefit side, that when we structured this project, we asked for the question, we asked for the answer and then we said, what would you do with this? What's the benefit if we do this? So we kind of had a foregone conclusion that if we did one, two and then we'd get three. And so that was one of our little tricks that we had was we said, before we get started on this, what's the really, what if I do this? What will be the difference? You're having fun with data, aren't you? Oh, we have a lot of fun with data, we have a great team, we have a great team. Last question from me, Phillip, is any advice that you would give to fellow practitioners or things that you would do differently if you had to do it over again? I think simple is underrated. I've seen very complicated models that blow me away from their design and complexity and just the scientific acumen needed to build that model. And I've seen a presenter in front of the people who will have to make a difference with this model and they don't understand it. And so that effort, all that passion, all that intellectual capital wasn't used. It just lay on the shelf and over time it'll probably unfortunately disappear. And I think simple is you need to do this differently. And so that question, the answer, that structure of approaching every problem as if it were that simple question and answer is really the key thing that I would recommend to a lot of scientists. Yeah, so if they don't understand it, they don't trust it, they don't trust that they can't act on it. Absolutely. All right, Phillip Kim, thanks very much for coming on theCUBE, it was great to have you. I appreciate it, pleasure being here. All right, keep it right there, everybody. We have Simon Zang coming up here. If you want to know the secret behind LinkedIn's sales growth, pay attention to this next segment. Simon is a data scientist at LinkedIn, former neurosurgeon turned data scientist, a pretty interesting segment coming up. This is theCUBE. I'm Dave Vellante with Jeff Kelly. We'll be right back after this message.