 In Cambridge, Massachusetts, it's theCUBE, covering the MIT Chief Data Officer and the Information Quality Symposium. Now, here are your hosts, Stu Miniman and George Gilbert. Hi, you're watching the Cube SiliconANGLE Media's flagship program. We go out to lots of technology shows and symposiums like this one here, help extract the signal from the noise. I'm Stu Miniman, joined by George Gilbert from the Wikibon research team and really thrilled to have on the program the keynote speaker from this MIT event, Tom Davenport, who's professor at Babson, author of some books, including a new one that just came out, and thank you so much for joining us. My pleasure, great to be here. All right, so, you know, so many things in your morning keynote that I know George and I wanna dig into. I guess I'll start with, you talk about the, you know, four eras of, you called it data today. It used to be four. Information, I call it yeah. Information, sorry. But you said you started with, when it was three eras of analytics and now you've changed information. So I'm just curious, you know, we get caught up sometimes on semantics, but is there a reason why you switch from, you know, analytics to information now? Well, I'm not sure it's a permanent switch. I just did it for this occasion, but you know, I think that it's important for even people who aren't, who don't have as their job, doing something with analytics to realize that analytics or how we turn data into information. So kind of on a whim, I change it from four eras of analytics to four eras of information to kind of broaden it out in a sense and make people realize that the whole world is changing. It's not just about analytics. Yeah, no, it resonated with me because, you know, in the tech industry so much we get caught up on the latest tool. George will be talking about how Hadoop is moving to Spark and, you know, right, if we step back and look from a longitudinal view, you know, data is something that's been around for a long time, but as you said from Peter Drucker's quote, when we endow that with relevance and purpose, you know, that's when we get information. So it kind of changes. And that's why I got interested in analytics 20 years ago or so. It was because we weren't thinking enough about how we endowed data with relevance and purpose. Turning it into knowledge and knowledge management was one of those ways, and I did that for a long time, but the people who were doing stuff with analytics weren't really thinking about any of the human mechanisms for adding value to data. So that moved me in an analytics direction. Okay, so, Tom, you've been at this event before, you know, you've taught and written and, you know, written books about this whole space. Are you saying I'm old? No, no, no, you've got a great perspective. Okay, so bring us, what's exciting you these days? What are some of our big challenges and big opportunities that we're facing as kind of humanity in an industry? Yeah, well, I think for me, the most exciting thing is there are all these areas where there's just too much data and too much analysis for humans to do it anymore. You know, when I first started working with analytics, the idea was some human analysts would have a hypothesis about what's going on in the data and you'd gather some data and test that hypothesis and so on. It could take weeks if not months and now, you know, we need to make decisions in milliseconds on way too much data for a human to absorb. Even in areas like healthcare, we have 400 different types of cancer, hundreds of genes that might be related to cancer, hundreds of drugs to administer, you know, we have, these decisions have to be made by technology now and so very interesting to think about what's the remaining human role? How do we make sure those decisions are good? How do we review them and understand them? All sorts of fascinating new issues, I think. Along those lines, Tom, you know, you know, at a primitive level in the big data realm, the tools are kind of still emerging and we wanna keep track of every time someone's touched it or transformed it. But when you talk about something as serious as cancer and let's say we're modeling how we decide or how we get to a diagnosis, do we need a similar mechanism so that it's not either or either the doctor or, you know, some sort of machine learning model or cognitive model, some way for the model to say, here's how I arrived at that conclusion and then for the doctor to say, you know, to the patient, here's my thinking along those lines. Yeah, I mean, I think one can like her, just like Watson, it was just being used for a lot of these, I mean, Watson's being used for a lot of these oncology oriented projects and the good thing about Watson in that context is it does kind of presume a human asking a question in the first place and then a human deciding whether to take the answer. The answers in most cases still have confidence intervals, you know, confidence levels associated with them. So, and in healthcare, it's great that we have this electronic medical record where the physician's decision or the clinician's decision about how to treat that patient is recorded. In a lot of other areas of business, we don't really have that kind of system of record to say, you know, what decision did we make and why did we make it and so on. So in a way, I think healthcare, despite being very backward in a lot of areas is kind of better off than a lot of areas of business. The other thing I often say about healthcare is if they're treating you badly and you die, at least there will be a meeting about it in a healthcare institution. In business, you know, we screw up a decision, we push it onto the rug, nobody ever considered it. What about 30 years ago, I think it was with Porter's second book, you know, and the concept of the value chain and sort of remaking the understanding of strategy. And you're talking about, you know, the API economy and the data flows within that. Can you help tie your concept, you know, the data flows, the data value chain and the APIs that connect them with Porter's value chain across companies? Well, that's an interesting idea. I think, you know, companies are just starting to realize that we are in this API economy. You don't have to do it all yourself. The smart ones have, without kind of modeling it in any systematic way, like the Porter value chain have said, you know, we need to have other people linking to our information through APIs. Google is fairly smart, I think in saying, we'll even allow that for free for a while and if it looks like there's money to be made, then we'll start charging for access of those APIs. So, you know, building the access and then thinking about the revenue from it is one of the new principles of this approach. But I haven't seen, I think it would be a great idea for a paper to say, how do we translate the sort of value chain ideas of Michael Porter, which were, I don't know, 30 years ago into something for the API oriented world that we live in today. Would you think that might be appropriate for the sort of platform economics model of thinking that's emerging? That's an interesting question. I mean, the platform people are quite interested in our organizational connections. I don't hear them as talking as much about, you know, the new rules of the API economy. It's more about how to two-sided and multi-sided platforms work and so on. Michael Porter was a sort of industrial economist. A lot of those platform people are economists, so from that sense, it's the same kind of overall thinking, but lots of opportunity there to exploit, I think. All right, so Tom, I want to bring it back to kind of the chief data officer, one of the main themes of the symposium here. I really liked you talked about kind of, there needs to be a balance of offense and defense, because so much, at least in the last couple of years that we've been covering this, governance seems to be kind of a central piece of it. But it's such an exciting subject. Yeah, it's an exciting subject, but you put that purely in defense, and we get excited, the companies that are building new products, either saving or making more money with data. Can you talk a little bit about kind of as you see how this chief data officer needs to be, how that fits into your kind of four eras? Yeah, yeah, well, I don't know if I mentioned it in my talk, but I went back and confirmed my suspicion that Usama Fayad was the world's first chief data officer at Yahoo, and I looked at what Usama did at Yahoo, and it was very much data product and offense-oriented. He established Yahoo research labs, you know, not everything worked out well at Yahoo, in retrospect, but I think they were going in the direction of what interesting data products can we create, and so I think we saw a lot of kind of what I call 2.0 companies in the big data area in Silicon Valley, saying it's not just about internal decisions from data, it's what can we provide to customers in terms of data, not just access, but things that really provide value. That means data plus analytics, so at LinkedIn, they attribute about half of their membership to the people you may know data product, and everybody else has a people you may know now as well. These companies haven't been that systematic about how you build them and how do you know which ones to actually take the market and so on, but I think now more and more companies, even big industrial companies, are realizing that this is a distinct possibility, and we ought to look externally with our data for opportunities as much as supporting internal decisions. And I guess for you, you talk to companies like Yahoo, some of the big web companies, the whole big data meme has been about allowing tools and processes to get to a broader piece of the economy, counterbalance that a little bit, large public clouds and services. How much can a broad spectrum of companies out there get the skill set and really take advantage of these tools versus is it going to be something that I'm going to still going to need to go to some outside source or some of this? Well, I think it's all being democratized fairly rapidly. I read yesterday the first time the quote, nobody ever got fired for choosing Amazon web services. That's a lot cheaper than the previous company in that role, which was IBM, where you had to build up all these internal capabilities. So I think the human side is being democratized. There are over 100 universities now in the US alone that have analytics oriented degree programs. So I think there's plenty of opportunity for existing companies to do this. It's just a matter of awareness on the part of the management team. I think that's what's lacking in most cases. They're not watching your shows, I guess, enough. Along the lines of the, you know, going back 30 years, we had a preference, actually a precedent where the PC software sort of just exploded onto the scene. And it was, I want control over my information, not just spreadsheets, you know, creating my documents. But then at the same time, IT did not have those guardrails to, you know, help people from falling off, you know, their bikes and getting injured. What are the, what tools and technologies do we have for both audiences today so that we don't repeat that mistake? Yeah, I know it's a very interesting question. I think, you know, spreadsheets were great, you know, the ultimate democratization tool, but depending on which study you believe, 20 to 80% of them had errors in them and there were some pretty bad decisions that were made sometimes with them. So we now have the tools so that we could tell people, you know, that spreadsheet is not going to calculate the right value or you should not be using a pie chart for that visual display. I think vendors need to start building in those guardrails as you put it to say, here's how you use this product effectively in addition to just accomplishing your basic task. But you wouldn't see those guardrails extending all the way back to the data that's being provisioned for the users. Well, I think ultimately if we got to the point of having better control over our data to saying you should not be using that data element, it's not, you know, the right one for representing, you know, customer address or something along those lines. We're not there yet in the vast majority of companies. I've seen a few that have kind of experimented with data watermarks or something to say, yes, this is the one that you're allowed to use has been certified as the right one for that purpose, but we need to do a lot more in that regard. All right, so Tom, you've got a new book that came out earlier this year. Only humans need to apply, winners and losers in the age of smart machines. So I'll ask you the same question we asked Eric Brunyolton and Andy McAfee when they wrote the second machine age, you know, are we all out of jobs soon? Well, I think they and I have become a little more optimistic as we look in some depth at the data. I mean, one, there are a lot of jobs involving working with these technologies and, you know, it's just somebody was telling me the other day that, I was doing a radio interview for my book and the guy I was talking to said, you know, I've made a big transition into podcasting. He said, but the vast majority of people in radio have not been able to make that transition. So if you're willing to kind of go with the flow, learn about new technologies, how they work, I think there are plenty of opportunities. The other thing to think about is that these transitions tend to be rather slow. I mean, we had about in the United States in 1980 about half a million bank tellers. Since then we've had ATMs, online banking, et cetera. Guess how many bank tellers we have in 2016? About half a million. It's rather shocking. I think I don't know exactly what they're all doing, but we're pretty slow in making these transitions. So I think those of us sitting here today or even watching are probably okay. We'll see some job loss on the margins, but anybody who's willing to keep up with new technologies and add value to the smart machines that come into the workplace, I think is likely to be okay. Okay, do you have any advice for people that either are looking at becoming, you know, chief data officers? Well, yeah, as you said, balance offense and defense. Defense is a very tricky area to inhabit as a CDO because if you succeed and you prevent, you know, breaches and privacy problems and security issues and so on, nobody gives you necessarily any credit for it or even knows that it's because of your work that you were successful. And if you fail, it's obviously very visible and bad for your career too. So I think you need to supplement defense with offense, activities around analytics, adding value to information, digitization, data products, et cetera. And then I think it's very important that you make nice with all the other data-oriented C-level executives, you know, you may not wanna report to a CIO or if you have a chief analytics officer or a chief information security officer or a chief digitization officer or a chief digital officer, you gotta present a united front to your organization and figure out what's the division of labor? Who's gonna do what? In too many of these organizations, some of these people aren't even talking to each other and it's crazy, really. I'm very confusing to the rest of the organization about who's doing what. Do you see the CDO role five years from now being a standalone piece in the organization and any guidance on where that should sit structurally compared to, say, the CIO? Yeah, I've said that ideally you'd have a CIO or somebody who all of these things reported to who could kind of represent all these different interests of the rest of the organization. That doesn't mean that a CDO shouldn't engage with the rest of the business. I think the CIO should be very engaged with the rest of the business. But I think this uncontrolled proliferation has not been a good thing. It does mean that information and data are really important to organizations so we need multiple people to address it but they need to be coordinated somehow and a smart CEO would say, you guys get your act together and figure out sort of who does what and tell me a structure. I think multiple different things can work. You can have it inside of IT, outside of IT but you gotta at least be collaborating. Okay, last question I've got is you talked about these errors and they're not one dies and the next one comes and you talked about we know how slow people especially are to change so what happens to the company that are still sitting in the 1.0 or 2.0 era as we see more 3.0 and 4.0 companies come? Yeah, well it's not a good place to be in general and I think what we've seen is in many industries the sophisticated companies with regard to IT are the ones that get more and more market share, the late adopters end up ultimately going out of business. How many you think about in retail who's still around? Walmart was the most aggressive company in terms of technology. Walmart is the world's largest company in moving packages around the world. Fanix was initially very aggressive with IT, UPS said we better get busy and they did it too. Not too much left of anybody else sending packages around the world so I think in every industry ultimately the ones that embrace these ideas tend to be the ones who prosper. All right, well time to have important really appreciate this morning's keynote and sharing with our audience everything that's happening in this space. We'll be back with lots more coverage here from the MIT CDO IQ Symposium. You're watching theCUBE. Hi, this is Chris.