Join our Data Science Community on Facebook. Every week, we talk about important data and analytics topics with top data scientists on Facebook Live.
In this #DataTalk, we talked with Dr. Sam Ransbotham, an associate professor of information systems at Boston College and the MIT Sloan Management Review Guest Editor for the Data and Analytics Big Idea Initiative.
Mike Delgado: Hello, and welcome to Experian’s Weekly Data Talk, a show featuring some of the smartest people working in data science today. Today we’re talking with Dr. Sam Ransbotham — I know I probably messed it up already. He’s an associate professor of information systems at Boston College. He’s also the guest editor for MIT Sloan Management Review. Sam earned his Ph.D. in information systems at Georgia Institute of Technology. He also got his Master of Science in management and his MBA from Georgia Institute of Technology, and he did his undergraduate work in chemical engineering. Fascinating background. Sam, thank you so much for being our guest today.
Sam Ransbotham: I love to be here. It’s fun stuff.
Mike Delgado: Could you take us through your academic journey, from your interest in data science to now being a thought leader and helping companies, as well as helping your students, understand how to use data and making it useful for business?
Sam Ransbotham: You’ve already intimidated me. You’ve used some of the smartest people and you said thought leader. No, I mean, I think this stuff is fun. You noted a couple of things in my background. I’ve got years of organic chemistry that I completely wasted. I don’t know what I was doing with that. What I found from that process is that what I really liked was the use of technologies to understand how things were happening. Whether it’s simulation in a chemical plant, which I really got more into than the actual chemistry, to understanding that we can model all sorts of things. And that’s kind of the journey to where I am at this particular point.
When we talk about “how did I get to this point?” — my actual Ph.D., my dissertation was using security data. With security data, you think about the security logs that are happening. They’re just millions, billions of records coming in all the time that people are monitoring.
And I got fascinated by trying to figure out what was going on in there. If you think about it as a giant haystack, there’s a whole lot of hay in that haystack. Lots of innocuous behavior, lots of normal good traffic. But buried in there are some really sharp needles. So I got pretty fascinated by, “Well, if I’ve got all this data, what am I going to do with it? How do I start to figure it out?” And that’s where I got started with, “Hey, some of these tools actually can be useful there and useful all over the place.” I got sucked in. It happens.
Mike Delgado: I like your illustration of big data as finding those needles in a haystack. I think it’s a really good one, because what’s really difficult is taking the amount of data that businesses collect and trying to find the right data that’s actually actionable. I heard an illustration last week … Instead of the haystack illustration, it’s like the Where’s Waldo books. Because you have all these different figures going on, and where is Waldo actually at? That’s the actionable stuff. It’ really funny hearing about your undergraduate work that led into your work with data science and information systems. You were working with a lot of security data that kind of sucked you in. What type of data were you working with back then?
Sam Ransbotham: It was intrusion detection logs. Basically, companies monitor pretty much all activity coming across their network. Within that, within those logs that accumulate massively, it’s just hard to appreciate the scale. They accumulate so quickly, but figuring out what’s going on in there can be hard. I’m going to push your Waldo analogy or the hay analogy a little further. One of the things that we keep finding as we do these annual studies about what’s happening with analytics, people … every year, 70 plus, 77 percent of people say they have more data than they had last year. Your real question here is, “If we talk about Waldos or figures, are we adding more extraneous figures? Or are we adding more Waldos?” In my analogy, are we adding more hay? The last thing you want to do is add more hay. If you’re trying to find needles, hay is not helping the situation.
That’s too glib, because we’re adding more needles too. So, the question is, “What’s our ratio of needles coming with that new hay we’re bringing in each year?” And I think that’s why we also see the difficulty, as you mentioned, actionable insights earlier and getting on top of that data. Everybody is struggling with that to some degree. It contrives them quickly.
Mike Delgado: Yeah, I think one of the big struggles is that, like you said, the amount of data that businesses are pulling in, it’s just more and more. There are more and more sensors. The new phones that come out, there’s just so much more data coming in. There’s a lot more hay now, and it takes somebody who’s really smart, who’s going to ask really good questions, to have a scientific mind to be able to sort through to find that actionable data.
Sam Ransbotham: What’s cool though, to push back on that, is that we have a lot more technologies to work with now, and those are so much more accessible to everyone. There’s probably a pitchfork analogy in here somewhere too. We all have a lot cooler [pitch]forks out there, and so many different people are working on new [pitch]forks. I haven’t really thought the whole pitchfork angle through.
Mike Delgado: I like that. I like that.
Sam Ransbotham: It’s there somewhere.
Mike Delgado: I like these metaphors. I think they work well. Make it tangible. So, you keynote, you speak, you write a lot about how businesses should be leveraging data. Today, we’re going to talk about AI, and I was curious about what sorts of questions you get a lot from business leaders when you speak about how businesses should be leveraging AI to turn data into insights.
Sam Ransbotham: So we recently did a study … It’s available, I’m sure there are show notes or something here that we can link to. We recently did a study about what’s happening with artificial intelligence, in particular in business. I think we have to be pretty careful, because it’s easy to get distracted by robots and shiny objects, but where is AI making a difference in business? And what we found was that about a quarter — actually less than a quarter — of the people involved in AI are very active in doing things with AI in business.
So, I think there are a lot of people wondering what to do. You ask, “What do people ask about?” Well, they’re asking, “How do I get started? What’s useful?” I hope that these reports offer a little bit of “Here are some people doing some good stuff, here’s some stuff that didn’t work out so well.” And so, we don’t all have to redo that process over and over again.
We can collectively make progress by not repeating mistakes. That’s what I hope happens for them. A lot of the people ask, “How do we get started? How do we get organized for this? How do I get people to believe in it? Should we invest in AI?” I think that we had a great quote from a guy at Airbus. He point-blank said, “We don’t invest in AI.” It turns out that a lot of our problems needed an AI solution, so we did AI there. But they didn’t go out there and say, “I’ve heard a lot about AI. Let me do some AI this week.” A very different perspective. Again, I think we’re getting too tool-heavy here, but the whole hammering the heel. Everything looking like a nail thing fits in here too.
Mike Delgado: You reference the study. I saw it quickly and just made a short URL for those listening to the podcast. This will work after the video show. The URL is ex.pn/mitstudy, and that will bring you to the latest research that Sam just worked on. Everyone can read it, download it, check it out. It’s very insightful. Again, the URL is ex.pn/mitstudy, and that will be functional after this broadcast. So, what were some of the findings you thought were insightful from this research?
Sam Ransbotham: We just released a different report yesterday, and one of the things that I really liked that we’re finding … I don’t know. I can get a little annoyed by people wanting quick fixes and short-term solutions, and so one of the things that I like about this study is that we found people getting much better results building off their prior investments. People don’t seem to be rushing in and getting magical gains from throwing magic beans in the ground and having beanstalks grow. What we’re seeing is payoff from people who have done what I think of as hard and sometimes not very fun work. We say things like governance, and we talk about a core index, like, how good are you at pulling in data? Ingesting data? Do you have the infrastructure and the systems built out for that?
Those things are not really headline-worthy. People don’t get excited about those sorts of things.
Mike Delgado: Right.
Sam Ransbotham: But they turn out to be really important, and what I like that kind of came through in this study is that people are getting payoff of that prior investment. They’re not dropping by saying, “Hey, let’s do some analytics, let’s do some AI” and having everything work out magically. That doesn’t mean you shouldn’t start, if you haven’t gotten further down that, but it says that there’s payoff from doing what I think of as some of the boring and less visible work. I like that story as an … I don’t know, shaking our tiny fist and then getting value from things that are work rather than flash.
Mike Delgado: Are there any use cases or things that you’ve seen recently where businesses have leveraged AI to make something that was very insightful for the business? I think it would be helpful for our audience to understand, especially those who are new to AI, not understanding how it’s being used by businesses. Can you share some use cases?
Sam Ransbotham: You’re going to get me a little bit on the soapbox here. We have a tendency to gravitate toward these shiny things. We like things that look like robots. We watch cartoons of robots growing up and we want to see those things. What I think are cooler examples, and the one that I’m going to espouse here, is something that’s completely invisible. One of the people that we talked to at Fidelity, they’re working on monitoring phone calls between the customer service people, and they’re trying to authenticate who you are. If you call into their systems and you start talking with them, the historical way that that used to work was, “Who are you? What’s your password? What are the last 17 addresses you lived at?”
Mike Delgado: That’s right.
Sam Ransbotham: Ironically, all the information that’s on Facebook. So, we gatekeep that. That customer relationship with the organization was based on this gauntlet type of approach. You come in the door. I am the guard, and I will protect you. And we want that. We want our data protected. We want our information protected, so there’s nothing wrong with that protection. But what they’ve done through AI is more transparent. They’ll have background processing about you and your voice to try to get authentication happening that way. So, that customer service person is seeing a little indicator of, “Hey, this is who’s checking out completely fine. Their phone call latency matches where they’re claiming that they’re calling from. They’re not pretending to call from one place and calling from someone else. There’s a red flag.”
All these things are happening in the background. The customer service person can actually be doing customer service. They’re not doing gauntlet. Very different approach. None of that visible to people, and all of it is cool. And they actually pointed out something that I hadn’t thought about: Even if you argue, you may not be acting as you. What if you’re acting under duress? You know, an extreme — someone is blackmailing you. They’re getting better and better at picking up on these clues that, if you think about it, would be very difficult for them organizationally to train all their people to do. We’ve gone years with the supermarket checkout carts checking your signature on your check. I’ve never believed that they had any ability to discern whether that was truly my signature.
Mike Delgado: That’s right.
Sam Ransbotham: And in this case that I’m mentioning here, we have a completely transparent use of AI that I think is going to add a lot of value, and that’s very different than the approach many people are taking of, hey, let’s put a chat bot out there and make people go through another gauntlet of for … Actually, I don’t want to call anybody out here, so I’m going to not give any examples here. But you may have noticed that I have a southern accent. It does not work well with the interactive voice response system sometimes, and if I’m to the point of having to call the company, I’ve already looked pretty hard on the website. I’m actually pretty good at that whole search engine stuff, so if you’re going to take me down this 97-level nested phone tree through a chatbot, by the time I do talk to someone, my fuse is going to be a little shorter.
Anyway, these are transparent ways and very different than what we think of as headline-worthy AI. All right, I’ll step down off the soapbox here.
Mike Delgado: I love that. That’s a great example because you’re so right. The headlines we see on Business Insider or Forbes that get a lot of attention are, like, the rogue robots taking over the world, because those are kind of scary. They get a lot of buzz, and of course, anything with robotics is cool, right? So we want to share that because we love robots —
Sam Ransbotham: There’s so much more —
Mike Delgado: Go ahead.
Sam Ransbotham: There’s so much more potential. We can’t let that stuff distract us. It’s cool, but we need to keep going on stuff that actually can be more value-creating.
Mike Delgado: Exactly. I mean, I grew up with the Jetsons. The robotic maid was great, right? And now we have the Rumba. I love your example of AI working in the background. It’s invisible, but it’s helping solve a really difficult issue, which is identifying who that person is on the other line, and AI working in the background so that, like you said, customer service can actually be customer service. Instead of having to ask all these other questions that you’re tired of answering, and like you said, a lot of the information, unfortunately, is already out there.
Sam Ransbotham: Yep. I think there’s a lot of potential, and we have to think more that way. I can’t help but put on the security hat, too, and think from the other side that … I mean, ever since the first person picked up a rock. Somebody used it to hammer something together, make something cool, and someone used it to bang on someone else’s head. This idea of espouse transparency, but there’s some downsides to transparency. There may be inherent biases that come in these algorithms that we don’t have good ways of checking. Again, that’s why these things are difficult problems that require work and are not going to be solved with a flash. That’s good. That’s good for us.
Mike Delgado: So, you’ve given a talk in the past around whether companies are collecting data or hoarding data, and I’ve never heard that distinction before. Can you talk about that? What are some of the problems?
Sam Ransbotham: It comes back to our haystack sort of thinking that we were just talking about. It is becoming incredibly cheap to generate data through sensors. I guess collect data through sensors. Storage has plummeted. I mean, storing it is practically nothing. There’s a tendency then to say, “Hey, let’s just keep it. It might be useful someday.” And I want you to think about your own personal life. I don’t know you very well, but I’m guessing — and I’m just going to use my psychic abilities here — that you have some stuff up in your attic that you’re saving that you might use someday, and there are a lot of analogies to what companies are doing as well with data. In the company context, and I framed it as hoarding versus collecting, what do hoarders do? A hoarder just says, “Hey, this might be useful. Let me collect it. Let me put it in a stack and later on, it will be really useful.”
Collectors are different. They look for things. They actively seek something that will add to their collection. They don’t get redundant things. They get things that will add difference. If you’re looking to understand customers, getting more data about something you already have a ton of data about, what value is that going to have? A collector mentality would be to say, “What would add to my insight about this customer?” And that’s a different philosophy. People push back and say, “Storage is cheap, what if you need it later?” I can see that point, but on the other hand — storage is cheap, but backing it up, data recovery, isn’t. Having humans go through and figure out what’s there isn’t. Sifting through all the hay to find the needle isn’t cheap. I’m not saying don’t collect any data, but I’m saying, back off this sort of, “Hey, let’s collect it. Maybe it’ll be useful someday.”
Nobody’s got time later on. By the time whoever collected that data, they’re going to be three jobs later. The governance associated with that is an overhead of, well, what is this data? How is it collected? What context was it collected in? These are things that require efforts to add to data, so if you’re not careful, you become a hoarder who’s got a bunch of stuff that you don’t really know what to do with. Can I say something that’s personal? I found a box of power bars upstairs recently when I was cleaning out a box, and they expired three years before I moved into Boston eight years ago. So, I moved three year old expired power bars because I didn’t have the moxie to go in and clean them up. You just have to be a little careful about that.
Mike Delgado: I’ve never heard hoarding versus collecting, and I love that analogy. I think it’s perfect. So, how do you counsel business leaders who maybe have that hoarding mindset? Because that’s so easy to get into, “Oh, we have this data. Might as well keep it.” How do you counsel someone who’s in that mind frame to go from being a hoarder to being very selective to collect the data that’s actually going to be useful?
Sam Ransbotham: Yeah, I mean, we don’t know what will happen in the future. That’s the whole thing. There’s some value in collecting them, but let’s say that you pilot collecting a little bit of information. Take it to a small test to see if you can get anything valuable out of it. Maybe you collect it and let it run for a bit, but you don’t need to just set up a process that continuously collects it. If you haven’t looked at it in three months, six months or whatever, put some sort of archive on that thing and say, “All right, we’re unlikely to get data out of that.” Because context changes. So, wouldn’t it be cool to have all this data? Well, that data was collected under an economy and under a competitive mindset that has likely changed rapidly. If anything is happening, we’re seeing more and more rapid change.
When it changes, you have to think about what that would do to your data. Let me use this as a technique to segue and build in another point that I think is important. We talked to CVS recently, and they have a ton of data about where they put stores, where they put locations. They’re pretty phenomenal at the models that they can build out about what each store’s traffic and retail sales will be. They made a strategic decision to drop tobacco products.
Mike Delgado: Right. I remember that.
Sam Ransbotham: Their historical data was all collected under the tobacco world, so if you want to go back and look at that data, you’d find out that the world has actually changed since that data was collected, and it changed in a way that fundamentally changes how useful that data is. The segue is that humans know about these exogenous changes. These changes that are happening out in the world, they knew that this tobacco thing was coming, in a way that models and existing data don’t. So, then we have to think about what are good roles for humans and what are good roles for data and for analytics. This is a great case that I think points to — hey, maybe courting all that data, that’s an example where it may not be useful, and it’s also an example where humans really kick in to understand a broader context. Anyway, I know that’s stuff that you were fired up about, so soapbox over.
Mike Delgado: I love all these examples. Do you have maybe one more example you can share about how businesses are doing it right? Maybe using AI right, or collecting data the right way?
Sam Ransbotham: Actually … I mean, I have dozens, but it’s like picking my favorite child. One of the examples that’s come out in a report yesterday, I’ll emphasize because we’ve got a webinar with this guy from WinField United, Teddy, who’s going to talk with us. WinField is part of Land of Lakes. They’re doing some phenomenal … They have relationships with their customers who buy their products, but they’re also doing a lot of test farms because those individual customers can’t test out hundreds of different seed varieties, but WinField can. They can test that out and provide that data to all their customers, or they can test different climates or different fertilizer treatments.
They can use sensors on the level of their test plot that individual farmers can’t do for their whole farm, and they’re doing some fascinating things about collecting data, doing experiments to figure out what they’ve learned from that data, but then also using that data to help their customers learn something that they can’t know and something that they can’t know individually.
And the net effect is that that link between WinField and their customers gets tighter because of that data sharing. It doesn’t get measured well in our GDP right now, because we just measure, you know, one seed is sold for this price and sent to this person. But that data flow is becoming increasingly important. And that’s what we see as actually a big thing going forward — this data sharing and these links that are going on. And I think that’s a fun example because it’s farming. I mean, that’s as old as people are.
Mike Delgado: Yeah. That’s a beautiful example of data sharing. I’m also curious about your perspective, or your insight for business leaders who are looking to make a cultural change in their organization to be more data-driven. To make decisions based on data instead of gut instinct. And I’m curious about your advice for the C-suite that is looking to make this cultural change. What tips would you share to help them make this move?
Sam Ransbotham: Yeah, that’s hard, right? What would be really cool — and would get a lot of video shares here — is if I came out with a magic thing that will solve this hugely hard problem, and it’s not coming. What we identified in our latest report is that there is a series of challenges that we’re just seeing over and over again each year. You mentioned leadership. Data quality is another one that keeps coming up over and over again. Security, governance, some of these things that we deal with, metrics, what you measure and what you do not measure keep coming up over and over again. The role of humans in leading this within organizations is big. You say, well, what advice do you have for leaders? We’ve seen plenty of examples where people go whole hog with these massive data undertakings, and I’ve never seen a massive IT project that didn’t fail massively.
The things that seem to work are people starting with something gradual and small, and this makes sense in retrospect. It doesn’t get the huge headlines, but you start with something small. If it works, you build it out a little bit more and a little more. You show some people that it works. One of our interviews this time was with Art Hu of Lenovo, and he talked about, “I don’t walk up to people and say, ‘Hey, I want to solve this internal problem.’ I say, ‘Our customers don’t like providing their serial number to everybody in the organization, so how can we make this one change?’” When it’s customer-focused, everybody can get behind it. You start to say, “Well, I’m going to share data between marketing and operations.” You might get into silos, you might get into turf wars, you might get into, who owns this data?
When you frame it small — and you frame it in terms of “How does this help our customer?” — people are much more likely to get behind it. People need examples of it working well. We don’t like to take giant leaps of faith, and there are good reasons for that. A small pilot study, things that prove somebody worked. When something does work, make a show of it. When something doesn’t work, don’t get out the chopping block and say, “What were you thinking for trying something like that?” I feel like lots of things end up being Goldilocks. Not too much, not too little. Just right.
We want to find the balance of trying something new without just trying something completely crazy that overinvest, and we have to find that balance between doing nothing and going crazy. And I think that’s where the human managerial role is. Knowing what’s appropriate for your organization and for the people in your organization. Those are tough managerial jobs.
Mike Delgado: I want to remind everyone listening to the podcast or watching the video that you can get this latest research from MIT, from Sam, over at ex.pn/mitstudy. It’s simply a redirect to bring you over to the MIT Sloan website, where you can download the report and read all the fascinating research. We’re basically out of time, but I do have one last question for you, Sam. It’s a quote from the study that I actually loved. I’ll read it here and I just wanted to kind of get your thoughts on it. The quote is, “As new developments in IT, AI, augmented reality and even neuroscience are introduced, one frontier lies in the marrying of datacentric and consumercentric lenses.” I was wondering if you can share your thoughts on that, because I thought it was a beautiful line.
Sam Ransbotham: The context for that was something that we did with Nielson and Carl Marci, and what I thought that really blended well was this sort of … We’re not John Henry, the steel driving man, that needs to ignore all data and do it as a human. We’re also not complete automatons that need to go pure data, and that really summed up a lot of the balance that I think made sense. We’re not at the point of turning over most operations to autonomous activity. On the other hand, we cannot take the whole, “You know, I think I’ll make a widget today” approach. So, we’ve got to get that blend. The people who do both are going to beat people who do one or the other.
Mike Delgado: Well, Sam, I want to thank you so much for sharing your insights with our data science community.
Sam Ransbotham: You bet.
Mike Delgado: It’s been awesome having you as our guest. I want to remind everybody to check out the study. Again, the short URL is ex.pn/mitstudy. Sam, where can everyone learn more about you and your work?
Sam Ransbotham: Well, actually, my name is Ransbotham, so with a name like Ransbotham, it’s hard to hide. I have a website out there that has links to all of my research, to all of our studies, a lot of the blog posts that I do for various places. That’s at samransbotham.com, and oh, yes, I mean, it’s exciting. You like data. I like data. I’ll be happy to talk forever. I like data.
Mike Delgado: I’ll put up another short URL: ex.pn/samr.
Sam Ransbotham: There you go. Much easier than Ransbotham.
Mike Delgado: And that will redirect over to your website, so people can read more.
Sam Ransbotham: I appreciate it.
Mike Delgado: And in the comments of this YouTube video and Facebook video, we’ll also put a link so everybody can check out your website, and we will put the links in the podcast show notes. So, I want to thank you again.
Sam Ransbotham: Tell the people listening I love this stuff. I’m sorry that we can’t hear from more people besides you and me. I mean, I enjoy talking to you, but I’d love to hear what other people think, so feel free to contact me.
Mike Delgado: Great. Again, thanks so much for sharing your insights. For everyone who’s watching, we do these data talks every single week. If you’d like to see past broadcasts and upcoming episodes, you can search for data talk on Google, or the short URL is ex.pn/datatalk. Sam, thank you again.
Sam Ransbotham: Thank you. I appreciate it.
Sam Ransbotham is an associate professor of information systems at Boston College and the MIT Sloan Management Review Guest Editor for the Data and Analytics Big Idea Initiative. In 2014, he was awarded an NSF CAREER Award for his analytics-based research in information security.
Sam was awarded one of eleven inaugural Google and WPP Marketing Awards to support research into how online media influences consumer behavior, attitudes, and decision-making. He received his PhD, MSM, and BChE degrees from the Georgia Institute of Technology.
Check out our upcoming live video big data discussions.