20 A. Charles Thomas, CDO, General Motors

[MUSIC] Welcome, welcome, welcome everyone to this episode of Tech Cars Machines. My name is Ali Tabibian. As always, there are lots of good information and good links in the episode notes, including a link to where you can find a transcript for this episode. A, Charles Thomas, Chief Data and Analytics Officer of General Motors. What a pleasure and honor it is to have an executive of Charles' seniority and experience, spend time with us. I know you are listeners, we'll appreciate his time as much as we did. Charles' entire career, as you'll hear, has been about analytics, and has experienced spans industries from energy to technology companies such as HP and financial services. As an example, before coming to General Motors, he was the Chief Data Officer and head of enterprise data and analytics at Wells Fargo, which is, you know, is one of the largest financial institutions in the world. Before that, he served in a very similar capacity at the large insurer, USAA. I find it interesting and important to point out for you that Charles launched his career off a PhD from Yale, focused on organizational behavior. Here's why I find this fact notable. You've noted in our episodes, let's say, with the executives from General Electric, from the Tier 1 Auto Suppliers Magna and ZF, that the mandate of those executives isn't just about selecting a particular technology, or partners, or making the right investments in third parties. It's also about their ability to affect change in a large complex organization, which rightfully typically have proud traditions and historically tried and tested workflows. In those environments, how do you influence people and processes such that they seek the change rather than feel that those changes are being forced on them? When you survey business school graduates a couple decades after graduation, organizational behavior tends to be the course that they point to as the course that's been most useful to them in their careers. That's one reason why to me, Charles's educational background was just as I catching as his accomplishments. And I invite you to listen carefully as he describes his objectives and approach. Let's get to it. Tech, cars, machines, subscribe here or at gtkpartners.com. Hello listeners. We're here in Los Angeles today, downtown Los Angeles on a beautiful day. It's a bright blue sky with the whips of cloud. And I was walking over here to our interview location, just the absolute perfect weather. And if I add to it that it's almost December and the weather is perfect, that kind of information will let you understand why it's so expensive in Los Angeles. So speaking of data leading to insights, we're here with A. Charles Thomas, Chief Data and Analytics Officer of General Motors. Charles, thank you so much for being with us today. We really appreciate it. Happy to be here. Great, Charles. You joined General Motors fairly recently when compared to the length of your career in the field of analytics. Tell us about your role at GM. What is its scope and what led you to accept the current position versus a very prominent role you previously had with Wells Fargo? Well, certainly. General Motors represented an opportunity to not only do the traditional role of a Chief Data and Analytics Officer, which is to make sure the data are a lot more important in line to help the enterprise leverage insights to drive actions that drive better results and the like. But to also do it within the sphere of the Internet of Things or rolling Internet of Things as I call it, we've had connected vehicles for about 20 years now. And we have not optimized the usage of those data for enhancing the customer experience or potentially even generating new revenue streams. So this was an opportunity to take what I'd learned over the past and really apply it into a space I love. Based on that description, is it accurate to think of the types of data that you're working with mainly in sort of the post-production environment of General Motors or does it extend to the operating environment as well? It includes operating environment. We're doing quite a number of projects in the manufacturing space to help them optimize what they're doing. There's all kinds of logistics work that can be done around inventory, around where we store parts and the like. So this isn't just a customer facing kind of position that's dealing with customer experience only. It's really, you know, spanning the gamut of how the company generates and will potentially use data in the future across all domains. Okay, excellent. What business outcomes are you targeting at General Motors? What do you expect to change as a result of your activity? Well, they're just the things that matter most to the business. We are an EBIT-driven company, so we want to find ways to drive positive EBIT and cash flow. We want to delight customers. And so everything from JD Power results and the like. We want to retain customers in a multi-generational framework so that might draw new metrics around, not just customer retention. Did you, you know, get your next GM vehicle, but do your kids buy GM vehicles and do your grandchildren buy GM vehicles? So there's quite a number of metrics that are really core to the business, but there are and there are also some transformational ones as well. So, you know, we look at things such as utilization of our dashboards because we want to transform the culture to be much more fact-based and more insight-driven. They're quite a number of metrics just across the gamut that we're using to measure our success, but it all starts with what's success to the business. Absolutely. Okay, and in that case, just for our listeners, the dashboard you're speaking of is the internal analytical dashboards that the analysts use. Yes, okay, yes. Okay, great. It's interesting when I was listening to your response, you've used a lot of phrases, especially as it leads to customer outcomes, that we hear when we work with insurance companies. And I know in your background you have Wells Fargo, which is a consumer-oriented company as well as a USAAA, where your last two major employers tell us a little bit about how what's common and what's different, maybe between general motors and your prior position in terms of how they use data and how they think about that customer. Yes, so in insurance and in banking, as you can imagine, there are going to be a little more advanced in terms of utilizing data largely because of risk. So imagine a major credit card issuer needs to really understand the risk that they're underwriting when they are issuing credit to a customer. So they've been using analytics for quite a while. Insurance is all about analytics. It's all about actuary and really assessing the risk of someone getting in an accident and the cost of that accident. So those industries are really born of needing to understand risk and understand the impact that it has on the balance sheet. I would expect that those companies would be more advanced in terms of utilizing data and analytics. However, there's a lot of data and analytics that happens within the car manufacturer. We just maybe don't think of it that way because it largely happens in engineering. And so if you think about all of the safety work, if you think about all of the work that's done to increase the performance of the vehicles and what have you, that's non-traditional what we call analytics but it's very much quantitative in nature and uses many of the same techniques and software. Where we have not historically used insights as much as we could, there's really been on the customer experience side. And so the similarities between the companies are customer gets in a vehicle and they want to do something. They don't want to just drive. They want to do some things. And the experience is wrapped around them and you want to understand what they're trying to accomplish and how you make their lives easier. The same thing as if someone goes into a bank branch or someone calls to file a claim. You want to maximize those interactions and ensure that you're relevant and timely and that you're speaking to them about the right things at the right time so that you can get them to use your products and be delighted by them. So there are things that kind of now are similar across that if you say a customer is a customer and if they like you though buy more stuff and if they don't they won't. Then if you look at it from that perspective, very similar. Interesting. So the data now has as much of a role as the product, the hardware product in terms of creating that customer affinity and long-term affection for the brand. It's not just about how good looking the cars are or how fast they are. Everybody makes good looking and fast cars. It's really about safety features. It's really about ease of engaging with the technology. It's about what you can do in the vehicle. It's about connected experience from when you're in your home to your driveway to the car to your office or wherever it might be and being able to use your mobile device with simplicity across all of those things and have an experience that's replicated regardless of where you are. A simple continuation of what you're doing. I believe in the future, I'm not saying that cars won't be important, they always will be. But I think what you're able to do in your vehicle will become increasingly important. If you have a collusion experience in the vehicle that doesn't overcome having an attractive car, having a good looking car, having a fast car, I think it will be less important if you don't have at least some baseline great experience in the vehicle. It's interesting in the enterprise environment, so much of the effort is about getting everything into a single paint of glass. When you go to the consumer side, a lot of it is about synchronizing multiple paints of glass that the consumer uses. It's an interesting. Simulously. Simulously. That's the key. They don't care that they're doing a lot of business with one line of business now, now they've helped across to another line of business. They want one experience and they want easy, they want value, they want value. want speed, they want one button if possible. And if we design, our experience is like the digital world does and like mobile does, I think will be in a great position moving forward. You've been with the concept of using analytics inside organizations as well as extending it to the consumer experience for as long as really most people have in the space. What can you tell us about what's changed in terms of how people use the data to things, really. Technology has made the ability to process massive amounts of data in a very short period of time. That has changed everything. When I started my career, you had to go pull a sample. With a sample, why do you pull a sample? You pull a sample because your machine can't process a million rows. You can only process 10,000. So you have to pull a sample and it stratifies to be write your code, you do your analysis. It takes weeks. Things now take minutes or hours. And that has compressed the insight generation component of the value lifecycle pretty significantly. What has not changed is how much time it takes to go from an insight to an action to result. And if you were to look at the things that we as practitioners have not lived up to the promise on is that we have not convinced the business to more rapidly take an insight, plug it into the operations, generate the result. The final yard is the longest yard. And so that still takes a lot of collaboration, a lot of partnership. And so I mentioned two things. The first was technology. The second is people. When I came out of school, there's no such thing as a degree in data sciences. And so I think the professionalization of the space is helping. But there's still that final yard piece where we've got to help business people get more comfortable with using this information and taking risks, calculator risks, using multivariate experiments and market tests and things like that that allow you to test ideas without betting the forum on it. We've got to get those people much more comfortable doing. So one of the things that I recommend, and if I were to coach young people in the space, is aspire to move out of the data sciences and become a marketing manager or a product manager or move into one of these decision-maker roles, where you know what you need from a data perspective, and you can move more quickly to get to the answer and drive results. So that would be kind of my aspiration for people coming behind me is to think more broadly, not just about being data analytics professionals, but being professionals who use data and analytics. And if you round out your skills, you can be very successful and drive measurable results that you can take credit for. Very interesting, thank you. And I think one thing I noticed in your bio was you actually use the term storytelling in one of the things you try to teach, I believe it's at Cal, that you have an advisory role that you try to-- do you ask them to focus on the role of storytelling? It's critical. I mean, you're dealing with very smart people, who-- you know, you're business partners, the person you're trying to influence. Very intelligent. And it got great degrees from great places. They've been doing their job for 30 years. They know their space, but they've just never used this stuff before. And so if you present it as, well, that's the fact. And why aren't you taking action? Good luck. What you really need to do is walk a mile in their shoes, help them use the language that will appeal to them, help them understand. Not that they're non-intelligent. It's just that they've never used this stuff. So the burden is on the teacher. Think back to your favorite professors in college, were they the most brilliant professors, or the ones that really helped you understand. And you walked away saying, I got it. That person spoke to me in a way that reached me and used analogies, used symbolism that now I totally get it. Or the person who went up to the board and they wrote out a bunch of equations that figured out. Those people might be brilliant, but they're not very effective. I'd rather be effective than brilliant. Understood. Maybe making that a little bit even more specific, if you care to, how do you then specifically line up the roles and responsibilities between the office of the data analytics officer and the business units and the R&D organizations, who presumably have some people who are kind of doing maybe less sophisticated versions of the same things. Yeah, so there are some recent articles by a couple of firms that really focus on, how do you bring in this coalition of peers? So you need to have the CIO at the table because you need to leverage the technology and the data that they're generating. You need to have a business person who owns the action. And then as the analytics professional, you need to come to the table with, you ask me for this very basic thing, right? And so I'm going to answer the mail on that. But then I'm going to earn your trust so that now I can counter-propose the next time you ask for me for something. Say, yeah, I'll be happy to deliver that. But let's take a step back and see what you're trying to accomplish and would it not be more effective if we try this approach that gives you what you ask for a plus? But that's really about relationships and that's about trust. But it's having the right people at the right level and how to engage with them. Now, that doesn't always mean it's the senior most person of the group. It's great when you have a mandate from the top, but very seldom are they actually in the details. So you really need to understand where the leverage points are in the organization, who owns the actions, and then how do you reach out to them and then influence them. Sometimes you're not the right person to do the influencing. Sometimes you need to get their chief lieutenant on board and then have that person influence. And so no matter how much data, how much technology, how much tools, and all this other stuff, it's all about people because people still make the decisions. And until people don't make the decisions anymore and we run everything on algorithms, analysts are going to have to learn how to influence. I'm in the influencing business. I don't own the means of production nor do I own the business outcome. In this particular case, I own the influencing of those things and holding people accountable once the influencing has happened. But that is the extent of my role. And people need to understand that there are certain skills that you must have if you're going to be effective in doing that. That's a great answer. How do you calibrate for the rate of change? There's people who want you to make that change go faster, but then there's also the reality of the organization and some of that immune system response that you might get if you try to push too hard. How do you calibrate that? Yeah, you have to think of yourself as being an internal consultancy that comes in and diagnoses not only the business challenge that your partner is facing, but also their level of adoption/maternity. And so we have certain business partners where there's really not much we do other than provide tools and maybe some training because they know what they want to do and they're ready to go fast. And so we're just want to be a key enabler. There are others who are like, I believe that there's something there, there's some there there regarding this data stuff, but I really need someone to escort me all the way through. And so our service offering and the skill set of the people will vary. So someone who's got great bedside manner would be assigned to the ladder, someone who's efficient and gets things done might be assigned to the former. And they both get what they need, but you get to mix up your service offering so that you're giving the organization where it needs. And if you peanut butter and say, well, everybody gets the hand holding a scorting, well then you've just ground the place to a halt. But if you just simply want to provide tools and training everybody, people will say, OK, what am I supposed to do with that? So you've really got to balance it and understand who you're talking to and understand their ability to digest and the speed with which they can. And then how do you operate and leverage channels that they have already established in order to make things happen? When trying to bring this sort of change to an industrial organization, frequently when I ask people in your position, they say that for a period of time, if you kill the gestation period, really the initiative for this type of thing has to come right from the top. The people at the top nurture your effort until it's had a chance for people to draw in what you are offering them as opposed to you having to explain what you have to offer them. Is that your view and your experience as well? Yes. And if the senior most person says we're going to do this, but not just that we're going to do this, but I'm going to periodically check in and make sure that we're doing this, that's where you have the greatest degree of success. Where it doesn't work is when someone says we're going to do this. And then they come back and say, OK, so where's the results? And I'm like, well, it fell apart. Three layers down because they had the 1,000 reasons why they can't do this. And you can mandate it from the top, but you're going to have to set up some sort of cadence around when you check in as a senior later and say, OK, we said we're going to do this. Did we do it? If we didn't, what's standing on the way? And then my expectations that we're going to do it next time, I'm going to see some results. And when you have that, you're much more likely to be successful. That's standing. Sort of this corollary question to what's the right way to start is how do you know you've been successful? And when I was preparing for this interview, you gave an answer to this question that I always asked in some of the public collateral that I saw that was probably the most illuminating one I've heard. And you said, I know I'm being successful when the business units start stealing my people. Essentially, I'm paraphrasing. Yeah. Well, so, you know, they're kind of operational metrics that we want to establish regarding service level agreements. And do we help the company make money or save money or whatever the objective of the activity was? So, that's kind of price of admission. That's not transformational. It's very transactional. The transformational piece is, you know, You know, do you see senior leaders interrogating data, having robust debates and having conversations about what to do next from what they see in front of them? As opposed to staged presentations where someone comes in and says, "Here's what we're going to do," and everybody nods and they move forward. Are they really questioning and seeking truth, seeking knowledge and interrogating information in order to come up with that? Another way that I'll know that I'm successful is if we can move away from a lot of these paper, binders, full of information that lots of senior people walk around with, and if we make information easy for them to engage on our iPad, and we coach them and teach them so that they can interrogate the data themselves and not have to flip through a thousand pages. The more educated the consumer, the easier it is for me to do my job and focus on the next big thing. The stealing of people is a good thing because people will see that, "Hey, the quickest way for me to be insight driven is to go take someone who knows how to do this my second nature." That I love because one, it creates great opportunities for my people and I think I should be an exporter of talent too, is that it makes my job so easy because it's just pitching catch. Because I don't have to influence, I don't have to beg, or I don't have to do anything to get people to use my stuff. They know exactly what to do with it and the quickest way for it to catch fires. They're sitting around their staff and we have one example of they're just killing it with the results and the boss is like, "Well, where's everybody else's results? Why aren't you using the stuff?" And that person, if they're wise and they're mature, it's like, "I can help you with that." I think there's a lot of opportunity for it to catch fire when you have someone who is exported from an analytics team who goes into a business function and then creates change. Outstanding. Charles, in a lot of what you've described here, there's a running theme of you getting all parts of the organization. The people you wouldn't necessarily expect to be involved with the kind of things that you're working with. You really get them involved, get them engaged. And I know if you don't mind transitioning a bit into the personal side of the history, I know your own background and your own personal history involves sort of a history with your probably starting with your father of him being drawn into and being groundbreaking in terms of what he did at his own career. And in some ways, at least in terms of a role model setting the stage for you, are you okay giving us a little bit of that background in terms of what that meant to you? Yeah, so we're from Baton Rouge, Louisiana and my parents are children of the 50s and 60s. So integration, particularly professionally, was starting to happen at that time. And my dad in 1968 was the first black professional hired by Exxon. He was refined right there, which was at the time one of the largest in North America. And he was the only one over a little while. Years before my grandfather was a janitor there. But unfortunately he passed away before he saw my dad do that. But being a first has, you take a lot of boys and darts. So I learned a lot of lessons about being a first in my own career. I'm first chief data in Amelix Officer 3, three major companies. And how to win people over, how to understand their business challenges and how to help them, and how to solve them, how to break down some of the, maybe they aren't the same racial barriers, but they're certainly barriers of woman egghead. And so how do I fit into a business environment and that sort of thing. So how do you adapt to your environment? So those are lessons that ultimately you want to be effective and you want to be successful. And so the burden is often on you to make the adjustments once you understand what you're walking into. And so that's allowed me to be able to move through various industries, various environments, and to pioneer things and not have any fear of failure because without risk there's no reward. So I learned that a long, long time ago and I've carried that into my professional career. >> Excellent. In my own case we used to live in Baltimore in the 70s. And my dad who has a little bit of a darker complexion than I do, he would be in the hospital in an elevator with a surgical unit on. He was a surgeon. And if he was the only person in the elevator people wouldn't get in with him. But I was always struck by how little that he didn't really let that bother him. He just didn't let him define it. It was just an observation that he made and then he just kind of moved on from there. And part of his reaction was to be always well-dressed, just present himself a little bit more sharply than he otherwise would need to. It's hard to do that in scrubs but still he tried. >> You were a nice shoes, I guess. But you got to put the look up with the designer glasses. You know, you've talked about in some of the material I've seen about the analytics being about really cool stuff versus ROI. How do you measure the ROI of what you do? What structures have you found out to be useful and successful? >> Well, the first person you need to get to meet in the whole process is the chief financial officer. That is the number one person because you need them to certify the benefits that you claim to the company. So that's understanding, helping walk them through methodologies that you're going to use that's going to drive incremental performance. That you can now take credit for. Getting them on your side is critical. That way you don't have to extend your own credibility and oftentimes someone of my role is relatively new. And so you don't have much credibility. And so, and success has many fathers as the saying goes and failure is a bastard. According to President Kennedy. And so as a result, lots of people will want to take credit when the results are there. So you've got to establish the methodological side that says we're going to use an in market AB test or multi-vari test. Against the challenger is going to be business as usual. Here's how we're going to construct it. In any incremental lift, we'll attribute to the change in stimulus we put into the environment. And now, can we get an agreement across all of the people who are involved before we kick off on this? Okay, yes. Great. See if you're going to certify it, right? If it works great, it doesn't find. I think it'll work, but let's try it. But having those processes that not only define the methodology, but the venues through which you will get agreement that, okay, once we find this, this is the answer. And then this person is going to certify the answer and they're in charge of that, does everybody agree? Once you have that, it would be much greater likelihood of being able to claim benefits above and beyond. Which you normally would when you don't have that agreement. Because once again, market forces or marketing brilliance or executive vision will always take credit for a great results. But when you can say, yeah, but business as usual means that you would have done it exactly the way you've always done it. And this is something new and the new trumped the old by 5%, that 5%, multiplied by $1,000 of vehicle equals blah. And CFO, do you certify? Now that gives you your greatest chance of having claimed and recognized the benefits. About three years ago in one of our conferences, the chief executive of Triple A insurance, the CIO of Dignity Health, which is a large healthcare provider, a couple other people, CIO of Chevron basically said, listen, we don't pay for discovery when it comes to data analytics. We define a problem and we try to solve it. What's interesting is, today, during the course of this year when I've talked to the same people, both privately as far as this podcast series is concerned, that's changed. Where people now will pay for discovery. In other words, there's a little bit of an R&D mentality when it comes to data analytics. Have you noticed that change? Is that something that you would say has been the case over the last few years, or do you have a different view on it? Yes, I think you've got to, and out of all of those people you mentioned, you've got to answer the mail on their immediate needs to earn the trust. If you come in with this fancy strategic visionary, it's going to take us five years thing you won't be around to see that actually executed. You've got to execute on their immediate needs, earn their trust, then they come and say, "Okay, so that's pretty good. What else can you do?" Then you say, "Well, you didn't know this, but we were working on this other thing on the side for you for just this moment." And here's what the possibilities really look like. Let's try. And you're much more likely to have a win when you answer the mail on the immediate need. That doesn't mean you don't start the transformational thing until you get agreement. You should finance that yourself. Carve off a little bit of your capacity to work on the next big thing. You might not necessarily tell them because they'll say, "I want all the resource working on my immediate problem." So you solve their immediate problem. And then if they're good business people, they'll say, "Well, what else can you do?" And then you say, "Well, I'm glad you asked that question. Here's what we've been waiting for. Here's what we can do." Excellent. Let me turn to the conversation a little bit toward the technology architecture. Some of the things that the vendor ecosystem might be interested in hearing about. What's your view on what is the right architecture for analytics within the organization? There's the typical, you talk about IoT analytics, et cetera, one of the big pushes and pulls is, edge versus core. How much of it do you bring back and consolidate and run in one fell swoop versus? kind of enabling the edge, whether it be the machinery itself or the people themselves to do a little bit of the work. Is that an interesting question for you or we can move on to something? >> Yeah, I think it is. And I think, you know, it just depends on the application, right? I mean, and I don't mean the technical application, I mean the business application. If you're wanting to solve for an immediate need that's happening in a vehicle while a customer is driving in real time, that would have a certain set of parameters around how the data need to be captured and processed and utilized. You might collect something that generates an algorithm that warns them about lane departures or warns them about speeding or if you have an insurance product that warns them about reckless driving behavior that brings them back in a line that now processes things right there on the vehicle with the algorithm that now is very informational. There are other things that require that you blend those data points with other things about the customer or about, you know, whatever it might be that you now need to bring it back to your larger data lake and to blend it together to generate an insight that now might be pushed and executed in real time once you generate the algorithm. But in order to generate the insight you might need to pull it all back into one place. So I think the key is to understand and discern which business application you are trying to accomplish and then design your architecture to meet that very specific business need. Companies will often one size fits all it approach and then they are very sadly disappointed when you're looking for something to happen in real time as close to the customer as possible but you've created an environment that requires you to pull it into process it to merge it to do some other things and then push it back out again. It can do it much faster than it could have when I started my career but it's certainly not real time. Is there something you're surprised that we didn't talk about today or you wish we had talked about today? And the second question is when people talk about machine learning AI visualization, etc. What do all those mean to you? Are these really dramatic breakthroughs or are they just increasing refinements in how data has always been used? The first one I would have said I thought we would talk more about AI and ML so now we are to me technology is always obsolete in a few months anyway. So to me it's just the better tool. It is transformational in the right hands. The company has to be mature, the people who are leveraging it need to be mature, they need to know what to do with it. But for me these are just the latest of what people think are new tools but they're really not. When I was an analyst we were building neural network models which is a form of artificial intelligence and in many cases they outperformed traditional regression based techniques and many cases they didn't. I think these new evolutions now are because they happen a lot faster they don't need an analyst to actually execute the work there in many cases self learning and they also happen in real time or in real time. That's the biggest difference but the math hasn't really changed. It has changed but I mean it's still based upon many of the same principles. My question is who's using it and if it's just the digital companies once again well then that's interesting but there's a great book out that I wrote. I was able to write a little not a forward but the jacket comments and it's really just about how do you actually put the stuff in a play and use it and that's still the great beyond. Until companies do that these are all interesting science projects and you know what happened to your fourth grade erupting volcano somewhere in the trash heap and nobody is talking about it anymore. If we don't get companies use this stuff it's all just going to be interesting and interesting is the kiss of death in this business because ultimately you want to be effective. Interesting is that's nice, ineffective but nice. Great Charles thank you so much for spending the time with us. Sure. I really appreciate it. Thanks. Thank you. [Music]