Empower22 - Doug and Bharath - An EMPOWER Talk with Doug Fearing of Zelus Analytics

[MUSIC PLAYING] Everyone, thanks for joining us for our final session of day one of Empower. My name is Bharath Vasudevan. I'm leading product for the division of Quest that is hosting this event. And today, I would like to introduce and welcome Doug Fearing. He's the co-founder and CEO of Zelus Analytics. The topics we're going to cover today will revolve really around the subject of data intelligence. Before we get into a little bit more of the content, hey, Doug, can you give us a little bit more background on Zelus? Sure. So thank you for inviting me. It's nice to have this opportunity. So Zelus Analytics is a data analytics company that works with professional sports teams. We build proprietary models using teams' complex data sources to provide a competitive advantage. We work with a limited number of teams in each sport, so one per division in baseball, three per conference in basketball. We work with a few European soccer clubs and Indian Premier League Cricket Club, a couple of hockey franchises, and are working into to expand into football. And in each of those cases, we create models that help them win more games. So we're looking at player performance. We're looking at in-game strategy. We're even looking at tools for helping with player development all to help our partners compete. So most of what you described-- I think a lot of people who aren't well versed in the art remember the movie Moneyball. Is that a fair oversimplification and generalization of what you guys do? Yeah. Yeah. Sadly, I am not Brad Pitt. I'm more of the Jonah Hill character in this movie. But yeah, this is the iteration of Moneyball. So if you think-- for those who've seen Moneyball, a lot of it was trying to understand how different outcomes affect winning and look for players that excel in those areas. And how sports analytics has expanded over time is just more and more complex data sources. So that goes from spatial temporal tracking of ball and player movement to now in baseball, there are multiple cameras pointing at the batter and pitcher on the field to collect player movement data and limb tracking. So we're using those complex data sources to better predict performance and to better improve performance. Well, all right. You said multiple different leagues, and I think you guys-- if I remember right about your history-- started with Major League Baseball and then kind of expanded into other areas. So what are the similarities that you see between data or even metadata that you found across these different sports? One of the common challenges that exists across sports is working with diverse data sources. So we have-- for example, I mentioned kinematics data. That comes from multiple different providers. Ball and player tracking data, there are multiple providers for that. We have college performance data. We have league data. And each of these data sources uses their own metadata. They use their own identification schemes, and so the first step is always integration. How do we map these data sources together and make sure that we're building models on top of high quality information? I can imagine some of the challenges you're describing is very similar to what the audience that's listening today is facing. I know one of their challenges that we've heard repeatedly is dealing with sensitive information, or protected information. How does that work for you? Our entire model is based off of working with proprietary data sources from our customers, and so we have to be very careful about managing that information. We have agreements with the data vendors that we will only use their data sources in appropriate ways and to provide services for our common customers. And we need to isolate those data stacks for our clients. So being very thoughtful about how we set up our data architecture so that each client lives in its own project space and has its own clean data environment for building those models. So more like a shared nothing model. Exactly, yeah. And that's one of the challenges that we deal with is we're not sharing the data between clients, but we want to share code. We have the same model specifications. We're trying to, even where we can, save on compute costs where there are overlapping data. And so it is that balance between security and efficiency that we're often working with. One of the challenges that a lot of the companies and organizations we talk to, they have a hard time pushing projects like this through where the option for them is to effectively do nothing. I'm sure when you have similar conversations with your prospects, they have another choice of just sitting back and doing nothing. How do you address that? Because I'm sure it's a standard objection that comes up. That's a challenge for us, both in terms of trying to integrate with the internal group. We do a lot to make sure that we are seen as supporting the work that's being done within our partner teams as opposed to competing with them, so providing complementary and supplementary services. And in the data and analytics space, I think we're fortunate because pro teams, it is a zero sum game. So effectively, if we can help someone win, they want that advantage. And so it's a little bit easier to tie the work that we're doing to the outcome on the field. Now with that said, we still run into similar challenges that other industries do. There are traditional decision makers who rely more on instinct and expertise to make subjective decisions, and it's hard to sell those individuals on the value of data and analytics. And largely, we rely on our customers, our internal analytics partners to build the relationships within their organizations so that the value of our work can be distributed. You are effectively, in your own words, describing an enterprise sales cycle, right? You've got your champion within the organization. They work closely with you to get whatever you're doing better understood. Who are you typically competing with? We're competing against internal investment. Teams, because they're focused on competitive advantage, they treat the space as being very proprietary, probably overly so. And a lot of what happens in sports is that you have 30 teams using-- for example, in baseball, you have 30 teams using the same data to solve the same problems and hoping that they're doing it a little better than the team next to them without any real way to validate that. And so I think we play a pretty interesting role there because not only are we getting some cost-sharing benefits, we're able to build the expertise in-house that teams don't have. We're also providing this external reference. Teams can compare what they're doing against the work and the approaches that we use to try to improve their own techniques. So what you're effectively offering them is a build versus buy decision. But in the process of buying, it's going to help them get better themselves. Absolutely, because we're developing all this expertise building up our systems and working with multiple sports in a way that you can't get that experience within a single company, within a single team. And so we can, in addition to bringing the product to our customers, we can also share that expertise. And frankly, their job is putting the best product on the field, and that's their business. And I guess your business is just enabling them to be more successful. Yeah. And look, I think this is happening across industries. If you look at pro sports teams, they effectively have to create these small tech startups within their organization to leverage data and analytics, and that's not the core expertise. The core expertise is, as you said, the play on the field and putting the team in place for that. So there is this opportunity across sports and otherwise for vendors, for experts to come in and be able to provide those services. Drawing a parallel to what-- that we're doing, right? You're offering a one-stop shop for-- let's take this module of work that you would have to do. Let us do it for you and augment your best practices, make your business better. I think when we approach a similar problem, it's providing the right tools and the right tool-set to simplify a very challenging problem that larger organizations have in that everything is very disparate. Nothing is uniform. It was built over different time periods using different technologies and different protocols, and there needs to be some unifying presence that really cuts across that heterogeneity to offer something that an organization can use to either model out their environment or connect into some of the tools that they're already using today. So it almost seems like in the grander scheme of the people, process, and technology spectrum, we're trying to address it with smarter tools. You're trying to address it through people and a combination of processes. But ultimately, we're marching towards the same outcomes for our customers. Yeah, I think that's right. If you think about the way things have evolved in sports, but more broadly, we do have this explosion of data. And I think companies and teams have recognized that there is value in having access to all of those sources of information. They've built up the stacks to collect the information. And really, what's missing is that rich analytical layer, that rich tooling layer that sits on top of the data to make it actionable. And in many ways, that's what we're trying to solve for. We want to make the data and insights actionable for pro teams. You jogged something for me when you just talked about that. The volume of data that's really just exploded in the last few years, and there's all sorts of stats that says 90% of an organization's data is created in the last year, and that was 90% more than the year before, and it goes on and on like that. When we talk to organizations that call it a little bit less sophisticated and less mature, to them, that's a data protection problem. How do I protect all that data that I have? How do I back it up? How do I restore it? How do I forensically analyze it if something's wrong, there's a breach? For the more strategy forward organizations, they're having conversations with us around how to monetize that data, how to extract new insights out of that data. And it sounds like that's similar to what you're doing. Absolutely. Yeah. I think there's always this-- if you think about the progression I mentioned, first, it's just collect all the data. The teams recognize that there is value in having these disparate data sources. And then there's this period of hoarding the data, feeling like, OK, we have to lock it down. We don't want anyone else to have access to it. We believe it's valuable, but it's not really mined. It's sort of sitting in the raw, and that next step is where the value is really unlocked by taking those raw data sources, and being able to put them together, and create compelling models, create predictions, create interesting insights that the organization can actually use. Yeah. Making it actionable is probably the biggest benefit that you can provide them. It's not only have you done the thinking, you've done the analyzing. Now it's, OK, here's what you need to go do. Yeah, and I think there's a risk of being so focused on securitizing the data that you miss the opportunity to actually derive insights from it. So this has been tremendous. If you look at how we've taken a challenge that has nothing to do-- I would venture has very little to do with anybody who's listening's day-to-day business. We've taken it, and we've found parallels to what you're trying to do for professional sports organizations, which, again, is not the industry any of our listeners are in. But the techniques, the outcomes that you're trying to drive are exactly the same types of insights and outcomes that our listeners are here trying to do. We've already talked about different ways for them to do that. You're providing an outsourced model to your buyers. We're providing a tool and process-centric model to our buyers that help address some of the same complex data intelligence challenges that weren't present in the market, or weren't competitive differentiation between 5 and 10 years ago. Yeah. And look, I kind of hinted at this, but when you-- so prior to Zelus, I built the R&D group for the Los Angeles Dodgers. So I spent four seasons there. Through that, I was responsible for the baseball systems, so player information system, data engineering integration, responsible for quantitative analysis, which was essentially our data science group, and responsible for performance science, so sports science technologies. And I've always been excited about problems related to data and technology. Hard problems with data and technology are what has driven me throughout my career and led to founding Zelus. But if you actually are within the organization, the largest value is in building the relationships to influence the decisions. And so that's where I think it's-- we're really trying to push is, hey, we'll give you the best available information, the best available tools for-- to support that process, but we're enabling our internal teams to spend more time on building those relationships, on being involved in the decision-making process, on establishing best practices and processes that can incorporate data into-- whether it's player acquisition or in-game strategy, into decisions that lead to wins. So Doug, you just shared with us a little bit about the Dodgers. Can you describe the environment you walked into, the steps you had to take to build what you would consider a world-class organization? And did you get there? Yeah, that's a great question. So when I started with the Dodgers, all of our player performance data was essentially managed by our IT department. We had an on-prem SQL database that was updated on an ad-hoc basis within, and I think that maps to what a lot of people are doing, an ad-hoc process for collecting information. And what we had to build was a scalable, repeatable process for a larger variety of sources of data and integrating those on a daily basis. I think one of the first questions you need to address is, how quickly do you need to update? If this is a transactional database that we're only going to be doing analysis on a weekly basis, great. A batch process may be more relevant versus something where it's kind of like a streaming architecture, or regular updates. In baseball, a lot of the decision-making is daily. We want to know what happened in the last game. We want to be able to update our metrics, and so we had to build out a cloud-based data warehouse that could effectively house all of that information. So that was step one. And then you think through, OK, how are we going to integrate these data sources into our downstream processes, whether that's a website and building a caching layer on top? Or if that's models, and we need to have a computational infrastructure that sits on top of that data layer? But each of those steps has complexity associated with it, whether it's who has access to the data, how do you structure the data, and then how do you use it? I have one question for you, and I think you might have touched on this in the description. But in your environment, did you have producers of the data that were different than consumers of the data? Yeah. So to an extent, most of the data in sports is from data vendors, and that might be the league producing the data. Fundamentally, it's the players producing the data if you think about it, and so-- but the consumers of the information are the front office, the coaching staff, even some going back to the players to help them in their next game and the next series. So that was the primary path by which data was produced and consumed. We also had scouts who would put in scouting reports, and we had to collect information from them, and that would roll back into our models as well. In my mind, I'm thinking million-dollar arm where they're writing down, scrawling notes, and faxing it in. I'm guessing the ways of collecting that data have now been normalized. Yeah. I think when I started, it might have been a Lotus Notes database. But yes, that has definitely improved over time. I think teams have all built scouting applications that both allow scouts to enter the information but also access player information so that they can be really well informed when they're writing their reports. And it sounds like what you've done is you've basically taken it from working it in-house, building it in-house, understanding how to take value of it in-house, and effectively replicated that, but now delivering it more as a service. Yeah. Yeah. Fundamentally, I had the benefit of being able to see those problems within an organization. And with Zelus, we're trying to solve that at scale by working across teams and across sports. And frankly, you kind of joked trying to make this useful. You literally just walked through what every single person who's listening to this has to make from a decision-making standpoint. Do I build it or buy it? How do I do this? Is this my business or is it better served coming from somewhere else? Are there tools that can make my life easier and allow me to get to the business of what we actually do? Everything I think we've spoken about seems to be pretty relevant to modern decision-making. And I think that's really the motivation for when you and I first talked about this for what we wanted to elevate for our listeners. Whether they're existing Quest customers or they're learning about Quest for the first time, there's a lot of things that we are trying to enable from a more broader perspective, bringing in viewpoints like yours to help inform some of the decisions that the folks listening are going to have to make. Yeah, absolutely. And I do think that tension of, do you build it versus do you buy it, if you build it, it's just-- you start to focus more so on trying to create better and better tooling and internal capabilities, and it distracts from the more important mission of building the relationships, of influencing the decisions, and that separation has proven to be really valuable for us. If only there was a Zelus for every industry that we're working with, it would be a great story. Yeah, I think we're sticking with sports. That's where our expertise is. But yeah, there's a huge space here. Hey, Doug, before we wrap things up, what advice would you give to somebody who's listening to this and is looking down the barrel of having to make some of these decisions and decision processes? That's a really good question. I think if it's a professional sports team, I think the most important advice is to work with Zelus. So if there are any listeners who happen to own a team, please feel free to share my contact information. But no, I think the one thing that often ends up being a challenge is that people think this is easier than it really is. So working with data, integrating data, cleaning data, just getting to the point where you can start to do the analysis is a lot of work. And so if you can leverage experts who have done that work before, who already have the tooling, who have the capabilities there, that sort of data warehousing as a service is an incredibly valuable head start because what you really want to be doing is, as quickly as possible, pulling value from the information. You don't want to lose two years of just building your database. You want to spend as much time as you can understanding your data and using that to make good decisions. And I think that applies whether it's a professional team or any industry that has been collecting data and is now trying to figure out how to use it. So Doug, on behalf of Quest, I want to really, really thank you for coming and spending a little bit of time with us today. I know you are remarkably busy in your day job and getting the ability to squeeze us in is something I really, really want to thank you. I think the quality of the conversation that we've had, how it's going to resonate with the audience, and how they're going to be able to move it going forward is something that will continue to provide them value. And on behalf of the entire team here, I really want to thank you for spending that time with us. Yeah, thank you for having me. I really enjoyed the conversation. [MUSIC PLAYING]