 from San Jose, in the heart of Silicon Valley. It's theCUBE, covering Big Data SV 2016. Now your hosts, John Furrier and Peter Burris. Okay, welcome back everyone. We are here live in Silicon Valley for Big Data SV, Big Data Week as part of Strata Hadoop. This is Silicon Angles theCUBE. This is our flagship program. We go out to the events and extract the signal from the noise. I'm John Furrier, my host, Peter Burris, head of research at Wikibon Silicon Angle Media. Next guest, Dan Graham, general manager of R&D of Enterprise Group of Teradata and Stephanie McReynel's VP of marketing, Nalation. Welcome to theCUBE, good to see you again. Thank you, good to see you. Happy anniversary, your 15-year wedding anniversary today. It is today, today. And I was playing it with you. I've got to take the hard path. Hold my hand. Congratulations, it's a big accomplishment. You're five years older than Hadoop on the wedding side. I know, actually, our first son was born at the same year as Hadoop, so that's a special connection. I won't go there, but that's a good, we're gonna come back to that later after hours. So Dan, so at Teradata, obviously we were having a great conversation earlier today and all throughout the week, how big data and certainly data warehouses is changing. We had Jerry Heldon, Berkeley guys, done those databases and went through at Oracle. And he's on a lot of boards and his view is the transitions we're in. So I want to get your thoughts on the R&D view of the transitions because the word retrofit has been kicked around, not rip and replace, that's kind of the old word we knew from the 80s and 90s, but retrofitting is what people are thinking about because data warehousing is not going away, but it's being scaled differently. What's your thoughts on that? Because this is an area that's been getting a lot of discussion. Well, I think the data warehouse, particularly the technology, has started to grow into some of the things that the open source community started us and sort of poked you about and said, we need this and we need to change. So putting things like unstructured data in the data warehouse has been going on for quite a few years. Teradata's got it, IBM's got it, Oracle's got it, I can't speak for everybody, but we're not dumb, we're sitting there saying, okay, what's the trend, what do we get going on? So it's easier for us to put in a little feature than it is for others to build an entire ecosystem and reliability and performance and all the security. So it's not really a struggle for us, and so we've got quite a few customers doing weblog analysis, doing Internet of Things sensor data inside the data warehouse. I think it's a mystery to some, but I do agree with you that this is really a coexistence or an augmentation relationship. Teradata, my company, was very instrumental in saying, look, the data warehouse is gonna be here, Hadoop's gonna be over here, Discovery Zone's gonna be over there. The data warehouse was the only game in town for a long time, and we were not ashamed of that, but the reality is it needs to break out and has broken out into these other camps which all should be working together to help the customer. The customer has to have choices for a whole bunch of reasons. There's a lot more choices than just costs associated with this stuff. Yeah, you make a good point about having Hadoop kind of be on, so Cloudera's shift from, with Impala, someone was saying, oh, they're trying to be, they're trying to be teradata, they're trying to be a data warehouse, and let's just not playing out, and we had Rishi Yadav on earlier, he says, structured data's winning. And although unstructured data's certainly out there, everyone's using it, but everyone ends up coming back to some structured data, so is that how the cases we're gonna have this coexistence, and what does the coexistence look like, because now we can store our data, not only in a data warehouse or data lake or whatever we could call it, it's other places, so the diversity of storage becomes big, but the need for structured data. Let me give you a good example. Sensor data is about to flood the world and make big data look like really small stuff. It's gonna dwarf big data as we know it, so sensor data's gonna be huge, but because of the nature of the data, you wouldn't wanna put that in your data warehouse. It's kind of difficult to work with at first. It's very repetitive, and it's expensive if you put it in a data warehouse. So you put this in Hadoop. You put the less than ideal data in Hadoop, this data that's raw, needs a little bit of cleaning, needs a little bit of work, and you're gonna throw it away in three to six weeks. So I go to all the trouble to model it and put it in the data warehouse and store it at a high cost. If really all you're going to do is use it for a short period, have Hadoop grind it up and refine it, hand that refined information to the data warehouse where it can be put together with the other parts of the organizational model. So now I can see the sensor data in the context of labor costs. I can see it in the context of what am I doing on the production line from suppliers, from channels, from sales. So I can connect all the dots in the data warehouse, but Hadoop did the heavy lifting. So these two begin to work together. We're seeing this everywhere we go. It's not really a problem. We're working together. Great comment. I want to highlight a tweet that Ash Perk sent me. He's from Informatica, another company because I threw out the chum in the water on Twitter. Hey, you know, data warehouses is failing because certainly if you try to do something that's not meant for, it's going to fail. So I'd like to think everything's failing or if you're out in the media business, everything's dead and that gets instant reaction on Twitter. So I threw it out there and I said, you know, the data warehouse market delivered value but not where we wanted it, says Peter Burrus, or Peter was talking to a previous guest. And highlighting your point Dan, which is there's a lot of stuff going on where there's a roll for data and Ash Perk said, isn't it rather about the right tool for the right job for data warehousing? Hadoop and Data Lakes have a job to do too. That's your point. That's what you're basically saying. It'd be terrible if Hadoop became a data warehouse. It's got so many other important things to do. It's just a big, big mistake to try to make it into a data warehouse. I mean, it's not what you should do. You're right, best fit engineering is what you should be doing. Take the right tool for the right job, fit these things together. And you know, that includes price points. That includes a whole bunch of technologies. What's exciting about this is that it brings up the whole composability kind of DevOps theme. And Stephanie, I want to get your thoughts because you have a lens into the analytic world. And one of the recurring themes has certainly come up on theCUBE certainly the past year or more. And then last night, twice at our event, a customer raises his hand and says, how do I get actionable insights? Or how do I get the insights? And how do you make them actionable? Is this fantasy or is this real? So if you don't have the right architecture or the composed backend for data warehousing and do data lakes, the insights, the analytics become a problem. So what's your thoughts on that? Is it not actionable? Is it a function of getting the right insights? Is that having the right data? What's your thoughts about this? Getting the actionable insights surfaced and delivered? Well, there's the data, the diversity of that data when we restore it. And then there's an equal perspective, which is what is the diversity of the users that are going to access that data? And when you look at the user base now with all these self-service BI initiatives, we're getting access to everyone. We're saying, open the floodgates, let everyone access the data, let them figure it out, have a very iterative model for analyzing. The challenge of that, of course, is you need a little bit of structure and a little bit of governance to do that right and keep people kind of in their swim lanes and make sure they have a positive experience. So we're very much focused on, let's make sure that folks have access to the data, but they also understand the nuances of the underlying data and how it's structured and how they can apply it to the analytics. And when we can get that, we can get the whole organization ideating around analytics and how to apply them in the business and then really getting ROI from these great systems that DevOps has stood up and made available to the organization. So we've seen an explosion in data volumes, and you mentioned, Dan, that sensor data is gonna even eclipse all the data volume that other types of systems have created. We've not seen a comparable explosion in the number of users to access that. We had someone yesterday talk about the idea that maybe we're at peak saturation point or peak need for the old BI model because the number of users doesn't seem to be growing at the same rate. Now, software can fix that, can improve that. But from your perspective, how do you anticipate the user community is going to evolve to make use of a lot of the technologies that we're talking about? And I think it really depends on who you're classifying as the user community, right? Well, that's what I want you to do. How would you classify them? Well, so you have the traditional BI players and their model was deliver a dashboard, deliver metrics that we're gonna monitor the business by. And that's a very rigid model. What you see happening now with the advent of data science and the addition of a lot of data science teams to almost every business is you have individuals who are really trying to hypothesis test against the data. And while a data scientist has a PhD to do that and a lot of statistical techniques typically behind them, what we're seeing now with Tableau and Click and the more modern approaches to data visualization is a larger group of users can be given access to that same type of hypothesis testing approach. Now, that only works if you have some sort of central point of reference to what that data means. So there's a lot of education that needs to happen around data literacy and how to interpret data points and how to apply the right algorithms if you're gonna broaden that user base. And to be honest, until quite recently, we haven't as a community focused on developing and delivering tools and technology to help out there. We're so focused on getting the infrastructure down and making sure we can handle sensor data and high speed streams. We're now at a point where we're not gonna hit the return on investment unless we solve those problems of human interaction with the data, human interaction with one another so that we can have a language around analytics that everyone from the manager down to the IT or the DevOps guy really understands. Are we gonna change the people though or can we do things in software to make it more available and obvious to them? Yeah, I'm a firm believer that we should be doing things in software and technology. And I think if you look at human-computer interaction design and some of the breakthroughs that we have there about predictive transformations, predictive interfaces that can guide people with machine learning and AI to the right path, I think there's a big opportunity for innovation there. And we're doing a lot of that at Elation, but we're not the only vendor in the market. There are other companies in the big data space and self-service data prep in the visualization layer that are really pushing the envelope of how do we think about this differently? How do we think about crowdsourcing some of this information? Let's move away from a totally manual process of only the IT guy can define this stuff to opening it up to a broader audience and having the right tooling in place so it becomes very social and almost like the consumer apps that we're used to interacting with now. How do we get that experience more into the enterprise and use that to really create knowledge around the data in the organization and how to apply it? So to go back to your question, you were asking, are we gonna hire 500 or 600 more people because of this data explosion? The answer is no, based on what you just said, you're gonna make them more productive. You're gonna make the people 20, 30, 40% more productive and that's pretty much what your company does. I think there's also this notion that the size of the data has some relationship to the number of people, it doesn't. That's the computer's problem. Let the software and the computer go deal with scale and bring it down to the masses, but we are hitting a wall. We are hitting a wall in the sense that we're not going to get a lot of budget to expand our data science team, to expand our business intelligence team. So we're gonna just have to be smarter down there in those groups. Now, that doesn't mean there's not a lot more interest in this and there's some new budget but you're not gonna get 50% more budget or 100% more budget. And that's where the relationship comes in. Well, if the top line increases though, this is where the interesting budget is, that the budget is a percentage of revenue, then that could be, that's where interesting opportunity. I don't see CEOs thinking that way. But let's talk about that. And by the way, Dan, maybe offline, we could talk about the relationship between adding new sources and adding people because I actually think there's gonna be a little bit closer relationship than you're suggesting, but we'll talk about that some other time. But on this notion of the CEO, we are talking here about the idea of thinking about data as capital. Today, someone says, opportunity, how much is it gonna cost? I think we need to say, opportunity, what kind of data are we gonna need? What kind of information are we gonna need? So when you think about this relationship between people or the action, the people that need to take the action and the information they're gonna need to act properly, that's, I think, where we're going. That's more than just looking at a visualization tool. It's something deeper than that. So, Stephanie, what do you think about that? What, I think the interesting thing is it's not just about the data. The data isn't the only capital. I think you're right. We have to think about data as capital. But when I look at the landscape, I really think about the algorithm as capital. So the intellectual property is in the algorithm. It's not in the data. We need the data to feed the algorithm. And so it's important to store it, but you need the algorithm to make decisions. So to be clear, I totally agree. But most algorithms today are still stored as data. But you're absolutely right. Right. But it's interesting because when you look at the space and you look at what we obsess about and what we optimize, it's not always optimizing the storage of the algorithm. It's optimizing the storage of the data. And the algorithm is where there's a lot of confusion too in organizations. So if you're gonna change the behavior of a business manager, business managers, there's been a number of different studies and somewhere between 50 and 70% of business managers are still making decisions based on gut feel, even when they have a report or a dashboard or visualization right in front of them. And the reason usually is because- They don't trust the data. They don't trust the data. And because the analysts can't, you can send two of your best analysts out on the same inquiry and they come back with two different answers because they've transformed the data in a different way and they're both gonna be accurate. And so until you understand the transformation, the algorithm, you don't really understand how to apply that insight. And so if this is all gonna work, we gotta figure out how to be more descriptive about the algorithm and how it can be applied. Last night, I called that management by astrology. You have a model, which is the gut feel, but you negate the data which may deny that gut feel. What's about the operationalizing? It's one of the themes in our opening. Top three topic here was the operationalizing of it. And this really gets down to some of these concepts. And the people conversations is interesting. I throw out if the revenue increases, maybe you're getting more budget. But that brings up the people issues. So one of the things that I'm looking at is the indirect people component, which is the people who run around BI and the data, that's maybe that number's fixed, or maybe reduced with software. So there's some inefficiencies there for that. But the people who are using the data, the actual casual user might be an indirect user because they're also now tied to the data, whether it's gesture data or if it's machines and sensor data, that now that is where the actionable insights at the edge of the network might be. So that's where you get to the algorithm. So this comes back down to the operationalizing it. What is your guys' view on that? Because that is on everyone's CXO's agenda because they have some back room operations building up the infrastructure, the data warehouse to do data lake. Now the business owner says, okay, I want to put it in practice. So how do I operationalize an investment? Which talks about the people and the users. So that kind of thing. So your thoughts on operationalizing it. I mean, it's interesting. If you look at some of the technology organizations that have really built out their data team, there are the data scientists and there's data engineers. And the data engineers think about how do we operationalize and scale this? So you have your experimental group and then you have your execution group. And it's all about optimizing the skill sets and the flow of that data in the organization. And so I think we have to get much smarter about what are the machines really good at doing right now and what can you automate? And what are the different skill sets of users in the organization really good at? And how do you increase their productivity as much as you can? And that's different between different classes of users. And so what we're very focused on with Teradata is how can we make a Teradata platform and an architecture that maybe crosses Hadoop and a traditional Teradata database and asterisks to get ready to platform? And how do we make sure that the insights from all of that data that's been stored and managed there are available to the end users and clear what the role is of that data possibly in the organization and what role can it play? So how do we increase analyst productivity 20% to 50% is what we're seeing in our production implementations that we can get there by having the machine make suggestions and the analysts then finish the final mile. Dan, talk about the Teradata, because you're in R&D, which is a balance between trying to invent the future and match what the current market is doing. So the current market is saying, OK, here's how I store it. Here's how I process it. Here's how I operationalize. Here's the ROI calculations. And a lot of it's stuff around the corner they may or may not see. You have to see around the corner for Teradata. What's your thoughts on what you look for and what are the dots that we can connect here? Well, to connect this dot with the prior question as well, we have something called the sentient enterprise. It's a vision. It's a roadmap. It's a North Star for people to navigate by. And to connect those dots, I think the fastest way to be productive is not do the work. If you don't do it, you can do something else. And so as Stephanie was saying, we've got a situation here where we just can't hire enough kids out of college and put them in front of dashboards. It's not going to happen. No matter what we can pay them, I can't have 500 young people staring at dashboards falling asleep because it's relatively boring data at some point. You have to automate it. You have to take that automation step. And you have to take the algorithms, the neural nets, the Bayseans, the crazy algorithms that people are still inventing, and embed them in the software so that those decisions not only are made automatically, but the only thing that comes to the human is the anomaly. This one's out of bounds. This one's past a threshold. This one's worth looking at. I don't really want you to stare at these little needles going by all the time and wish that you had a better job. That's not going to work out very well. It's exciting data. It's the little outliers. It's the exploration, right? It's the problem areas that you need to go and dig into and solve that really have an effect on the company. And it's not just alerts and panics because if our alarm went off somewhere, it could be a long-term analysis that says, this is trending towards a bad bottom line in three months. That may not be a big deal today, but in three months, you'll get the phone call. So you need to automate things that are not just simple alerts and panics and anomalies, but you need to look at even deeper analysis that looks out a few months and saves your CFO from that rather testy boardroom meeting. I think sometimes too, it's a piece of data that might spark an idea or a thought. As we get more subject matter experts using data and looking for things in the data, they're gonna be thinking about, how do we optimize the business? It's kind of a change in mindset. You're going from a static metric and like, oh my God, we're under the metric and we can't make it. And you're opening up a world to, wow, how can we think differently about this business process? How can we think differently about how we approach the market? What's a trend that's happening that we might be able to jump on? And so it really, you build a tighter closed loop to how you're making decisions in the business and you start to see benefits from those small experiments. They give you confidence to make the larger that. Stephanie and Dan, thanks so much for sharing the insight here on theCUBE. I'll give you guys the last word. We have a few seconds left. What's the bumper sticker for the show this year? Here in San Jose, Silicon Valley, in San Jose for a big day to SV and Strata had to do. What's the bottom line summary of the event if you had to put on a bumper sticker? One word bumper sticker. You can do a sentence, just do it in 10 seconds. What's the vibe of the show? What's the, is it up trending up, kind of ho-hum? What's the vibe, what's going on? What's the theme? What are you seeing? I mean, I've been pretty laser-like focused on data collaboration and how we get humans and machines working together. So I've heard a lot about ML and AI and what's the next, what's the next frontier, how we apply all of that. I think Dan, you tried to capture that with the notion of the sentient enterprise. Right. Yeah, I would say it's best-fit engineering, the right tool for the right job. I think all the hype is starting to settle down and we're looking at the facts that not all of the tools are doing what they were originally promised, but they have good value in what they do. So let's put- Stand the tool set, don't be the hammer and everything's a nail, be everything. That's right, let's get all the tools. Just tools. All right, just tools. If you're a hammer, everything's a nail. Expand your toolbox, AI, machine learning, a lot of good technologies, business outcomes. This is the cue. We'll be doubling next week for Ireland, for Hadoop Summit. Check us out there. Check us out on Facebook for all the photos. Go to Twitter and search hashtag cube gems for all the short highlights from this interview and all the others. And if you want to see pictures, go to cube cards, hashtag cube cards on Twitter. They're all up there right now. This is the cue. We'll be right back with more after this short break.