Data Science Hangout | Chase Carpenter, Chicago Cubs | Advice for Getting your First Job in Sports

Transcript#

This transcript was generated automatically and may contain errors.

Well, let's get going then and welcome everybody to the Data Science Hangout. And a special welcome to anyone who's joining for the first time today. So if this is your first time joining, just to kind of let you know how things work here. It's an all audience led informal session, where we want to focus all the data science leadership questions that are most important to you all. And so each week, I'm joined by a different featured leader, you get to pick the brain. And so there's three ways that you can ask questions, you could jump in live, you could put questions in the chat. And we also have a Slido, which I'll share the link to shortly. But just a reminder that the conversation is recorded and shared up to YouTube for anybody who misses it.

But we really do want to make this a open and welcome space where everybody can participate. And we really want to hear from everyone. And I'll say we especially want to hear from people from underrepresented groups in data science as well. And I'm so excited to be joined by my co host for today. Chase Carpenter is the Director of Strategy and Analytics at the Chicago Cubs. And Chase, I'd love to kick it off by maybe having you introduce yourself and a little bit about the work that you do.

Yeah, hey, Rachel, everybody. First off, sincere thanks for scheduling this and giving me the opportunity to chat today. So Rachel said, I'm my name is Chase Carpenter. I'm the Director of Strategy and Analytics for the Chicago Cubs. I've been with the organization for just over five years. Everything I work on is on the business side. And in a lot of ways, I support most of our major functional areas. So it's ticketing, marketing, corporate partnerships, food and beverage, retail hospitality, as well as our event operations group. We think of ourselves as a lot of ways an internal resource or sort of internal consulting group for the rest of the business.

Awesome. Thanks, Chase. And while we're waiting for questions to come in from everyone, to kick it off with asking you maybe, what's something that you're most excited about? With regards to data science?

Yeah, so I, it's a really good question. I spent some time thinking about it. You know, I so I spend a lot of time thinking about data science, a lot of my time trying to think about how to drive value out of data science, more broadly speaking, sort of data as an asset. And so I think if you even if you just look at the my title, director of strategy and analytics, I spent a lot of time trying to find areas of opportunity where we can drive business value out of the data that we have or that we can reasonably go collect. You know, I think there's a wide range of problems out there. Some problems are interesting, some problems drive business value. And so where I try and focus my time is sort of the marriage of the two.

And I think more broadly for data science as a as a profession, we've made some awesome progress, better integrating data and analytics into the business operation, certainly within sports, but you know, sort of across the sort of all industry verticals. And so that's part of that is due to tools and process and as well as at the academic level, some some significant progress in getting the right types of course material into the hands of students as they come up through both undergrad and graduate school. And it's sort of been an awesome evolution that we were sort of actually seeing this marriage of putting data into practice so that you can drive business value for sort of whatever organization you're a part of.

But it is a little more sort of manual and hands-on than at least in an academic setting, when you think about this is the optimal dynamic pricing infrastructure, where we've got some real headroom to sort of grow into that type of model.

I think that's great. Thank you. Yeah. I think keeping the human in the loop, like the algorithm loop, is really interesting. And also kind of the algorithm aversion that you might see, where you're like, oh, I don't think that's the right thing to do. I don't want the computer to make the decision for me. That's been an interesting topic. And I think that it's great that you guys use it as a reference and not as something that's automatic.

Yeah. I think part of this, I think, is the sports industry. There's two dynamics that sort of distinguish us from a lot of other dynamic pricing applications. One, we have finite inventory. We've got 41,000 seats per event. And two, that inventory expires and becomes perishable. So I can't sell a ticket to last Tuesday's game. And then the third part of that is, in addition, we have multiple sales channels. And so accurately, you can get some interesting signals in your data that may or may not actually correlate to an increase or decrease in demand. So for example, a large group ticket in order, if there is a church group or some, we may sell a chunk of 500 tickets without some proper treatment and nuanced understanding of how that data actually flows through our systems. You could make the naive interpretation of demand is through the roof, no change price. And that may or may not actually have anything to do with the underlying demand as just like when a group sale closed, for example.

No, I mean, that's an interesting what people do. We have seen a huge shift of people delaying their purchase closer and closer to events. And so that's a consumer behavior phenomena that we're seeing across all sorts of industries. I think that creates a risk for us as a business, going back to my point related to expiring inventory. So as sales get pushed closer and closer to the event date, that just creates more uncertainty for our business. And we spend a lot of time thinking about how do we sort of operate in this environment that we're watching unfold.

Right. I see an anonymous question that just came in that was, since sports games happen so frequently, how are models and analytical projects evaluated over what time period? I'd say it's highly context specific. In baseball, there's a couple lenses that most of our decisions sort of naturally fall into. One is a yearly cadence. To some extent, the baseball schedule was pretty much set in stone from April 1st to the end of September. And so a lot of the work that we do, the true and sort of final evaluations for success, certainly sort of materializes throughout the season, but the real evaluation is after the season has unfolded. More at the operational level, there is success that can be evaluated and changes actually put in practice, sometimes on a game to game basis, but more often than not, a homestand by homestand basis. And so it's roughly every other week, so we can make operational changes. The team goes away for 10 days, and then they come back. And so we can make some changes in between homestands.

Pricing algorithms and data challenges

So I was just wondering to myself, what kind of algorithm they use at the Cubs?

So I would say, so specific to dynamic pricing, we have tested a wide range of different algorithms, sort of decision frameworks, everything from your traditional classic regression-based models to non-parametric models. Jeff and I, in our prior life, worked at a consulting shop that specialized in a variety of Bayesian techniques. And so I've messed around and tried to make the Bayesian frameworks fit within the type of data that we have. I will say, having spent a fair amount of time in that area, I think the bigger challenge that we have is more at the data synthesis and aggregation layer, getting the various data sets that you would want in real time or sort of at the right granularity or frequency that we ideally want to make decisions. We run into more bottlenecks at that layer versus any individual algorithm will end up being somewhat performance, given the data constraints that we have. So I spend more of my time focused at how can I go get the incremental data at a more fine granularity so that we could sort of more meaningfully engage in more of the sort of true optimization, algorithm optimization.

Delivering results to the business

Yeah, sure. So I think you had referred to your, someone asked a question about your tech stack earlier, but I kind of wanted to understand a little bit of that last mile of how you deliver your results to the business.

Yeah, so I'd say a couple avenues. We build and maintain a small number of Shiny applications, which ends up being sort of a deliverable. A fair number of outputs will be injected back into our data warehouse and then rendered through our BI layer. But also, I'll admit, at least for our business, a lot of my counterparts across the organization still rely pretty heavily on effectively the Microsoft stacks. It's Excel, it's PowerPoint, it's Word. And it's been a big ambition of mine to get away from that mindset, but it's a real reality. And part of it is a lot of the outputs that we generate make their way into other source documents or deliverables that I don't own or sort of have much purview into. Certainly, board materials, if we're putting a board deck or a memo together, need to provide deliverables that can easily be dropped and moved in those documents.

Thanks. Can I dive a bit deeper there? Because I'm sure that a bunch of people on the call have that struggle as well. How are you going about trying to change that or working together with the team?

I'm not sure I have a great answer. I think what, internally, I am making a conscious effort to shift more and more of our deliverables into more reproducible R Markdown or just sort of fully contained scripts and processes and then sort of go kicking and screaming when we need to break that paradigm. But that's, it's a constant sort of tension or something that we're always evaluating.

So, with R Markdown, does the team also generate PowerPoint or Word from R Markdown too? I have tried to be honest, given, and this is sort of in the weeds on the specific formats and sort of the preferences that we have in those templates, I haven't been able to fully close the gap and get sort of 100% continuity there. I'm sure it could be done. It's more of a timing and resourcing thing.

Handling uncertainty and evolving roles

Sure. So, you did mention, Chase, that you deal with ambiguity and uncertainty a lot. So, I'd be curious to hear what are your most go-to techniques for approaching ambiguity and where do you basically stop and say, hey, this is the threshold where I'm about to accept it and I'm not going to dig further. And the second question was just about your journey since you started on as a one-man department. How did your role evolve and what aspect did you enjoy most from business analysis to data science?

Yeah, both good questions. I'll try and I'll take them in order. So, in thinking about uncertainty, I think there's two questions embedded in there. One is more the mechanically how do you think about and actually quantify the uncertainty from an analytical perspective. And I'd go back to a lot of the Bayesian frameworks that I was introduced to in my first job. They've just got some nice academically rigorous and just like conceptually sound manners to actually to think about confidence intervals or credible intervals. And so, I use those frameworks sort of mechanically. I think where I have spent a lot of time is getting my colleagues and counterparts in our organization comfortable with embracing that uncertainty.

A lot of the areas where I spend my time, we have to our analytical deliverables are due several weeks or months in advance of those events actually happening. And so, to sit here in the first week of November and with 100% specificity say here's my projection for the ticket sales next July, it's really hard to make an argument. But I've spent a lot of time getting our organization understand like here's the range of probable outcomes. Here are the things that would make us believe that we may over index or sort of underperform. And how much risk or uncertainty are we willing to accept and sort of what's the downside of if we misproject or basically what's the cost of being wrong. And that's taken the better part of my five years here to sort of get folks to fully embrace. COVID-19 was somewhat beneficial in that because we had so much uncertainty facing our business. It was front and center and there was no other way, at least in my opinion, to like reasonably think about what may happen in our business than through sort of a truly probabilistic manner.

And how much risk or uncertainty are we willing to accept and sort of what's the downside of if we misproject or basically what's the cost of being wrong.

And to some extent, it depends. I try and before materially engaging with colleagues, I try and get sort of push the analytical models as far as we can. So, I have an understanding on here's how much sort of statistical power, how confident we can be in these results and then also framing up, you know, we are getting into the grounds of like I don't have, we don't have specific enough data or they're just more uncertainty. You're sort of crossing a threshold of we're asking for precision that this model can't really support. So, let's take a step back and sort of understand what those limits are.

Yeah, so when I joined, there was a, the organization had to recognize the need to bring on some more resources in the business analytics area. So, there was a little bit of initial buy-in that had already sort of occurred that led to my role being created. You know, having said that, in any organization, you're ultimately building relationships with individuals. And so, I was pretty conscious in finding some quick lens and sort of, you know, early victories that I could find to gain that credibility so that we could grow into bigger, more complex and challenging problems.

When I joined, I'd say most of the problems that I initially worked on in the first 12 to 18 months were sort of pretty pure point analytics. Like, the problems were well-defined and scoped. There was a clear analytical angle that we could bring to the table, but it was, you know, pretty well boxed. As I have progressed through my five years here, more and more of the work that I spend on, where I spend my time now, is addressing bigger strategic issues for the organization and sort of, somewhat major sort of policy decisions that we are evaluating. And I certainly own and sort of bring the consumer and the data perspective to the table, but there are very real sort of non-quantitative other stakeholders and other considerations that, as an organization, we need to evaluate. And we have, you know, a very large media presence as an organization. And so, we're very conscious of the messages that we put in the market and sort of our standing just from a PR standpoint. I've not figured out how to weave that into my analytical models, but that it's a very real consideration that, you know, it's more of an open discussion and dialogue with my colleagues.

Getting your first job in sports

Thanks, Chase. I know that some of the, like, R and sports analytics meetups, a lot of people have questions around how to, like, get that first job in sports. I'm curious to hear your perspective there on maybe how you do that or how you'd advise others to go about that.

Yeah, so I think finding the first job in sports is often the most challenging one. I think once you've found a foothold, it's a pretty, particularly on the business side, it's pretty open and sort of collaborative community, and so you see a fair amount of, not job hopping, but you can sort of navigate the professional team landscape somewhat, I won't say easily, but it happens quite a bit.

In terms of putting together a strategy to land that first job in sports, I'd highlight a couple things. One, I'd go back to sort of where I started. If your preference and priority is to work in sort of the baseball or the player operations side of the house, there's a pretty clear path of developing some expertise in being incredibly smart and sophisticated and understanding player performance. On the business side, at least for me, I had really no understanding of how the business of baseball worked before I got into the business, but there's some good publicly available resources for not only baseball but all the major sports. I think Forbes does a really good job of just synthesizing, here's how the business side of these operations work, and so I'd spend some time just level setting and really understanding how businesses operate.

The second focus I have is you need to get a little scrappy and tap into your network and find friends or friends of friends, reach out on LinkedIn or if you Google Cubs front office or any team front office, you'll find sort of the full, in general, find the full roster of employees. So look at the job descriptions, look at the titles, look at the individuals, reach out, start to build some connections. The third one is be patient. In general, we are, all front offices are smaller than maybe most people would think, and so jobs don't necessarily have a predictable hiring cadence, and so be patient and do some diligence to figure out when jobs do become available.

And the last thing I'd highlight is if your heart is set on working for a front office, I'm certainly not going to tell you to not go down that path, but there are some awesome jobs available in what I'd say the front office adjacent industries. The major concessionaires all employ hundreds of analytics focused professionals. There's a great sort of cottage industry of consultants that focus on the sports space, so depending on your preferences and priorities, you can land a job in any number of sort of sport adjacent areas. So the last thing I'll say is a little bit self-serving here, but I am looking to add a couple folks to my team. We've got one job description up right now and another will be up shortly, so if anyone's interested in focusing on improving the business operations, cubs.com slash jobs, and there we've got some opportunities available on our team.

Diversity and hiring

Awesome. Something that might be helpful to dive into deeper here, and maybe with the whole group as well, is when we talk about like being able to find these opportunities and networking, I wonder how like finding a diverse pool of candidates also plays into that, and what are some of the things that your organization does?

Yeah, it's a challenge. You know, we are making a very conscious effort to try and raise the brand awareness and profile of our organization to groups who may not otherwise naturally find cubs.com slash jobs. The tactics on how you execute that are somewhat specific to the individual roles, but it's a big emphasis and priority for us as an organization that I think we had. It's just an area that we can get better and evolve, and so it's making conscious efforts to increase sort of the top of the funnel. How do we raise our brand awareness and footprint so that we can attract a more qualified and diverse group of candidates? It's something that we're spending a lot of time trying to evolve and get better at.

That's great. Thank you. I see there was another anonymous question earlier that asked about what you originally studied. Was it analytics, or how did you get into this space? So my undergrad work focused on political science and psychology, so not terribly heavy on the data analytics front. My first job, I feel like I got the equivalent of a master's in applied statistics, so a lot of on-the-job training. I supplemented that with some graduate coursework in statistics, and then I went back and did my MBA part-time with a focus on operations analytics and marketing, but I'd say most of my practical, what I use in my day-to-day really came from my first job.

Skills for open roles

And so are there certain skills that you're hiring for right now, or certain things that you're looking for for those open roles? I would say yes. So one, just so we're looking to bring on a pricing analyst, someone to help us get smarter on sort of all the things that we talked about previously with respect to pricing, and the second role is more database marketing and sales operations focused.

I think in terms of skills, I think I need someone to be successful in these roles in particular, and I think this does apply to a lot of analytics-focused roles in sports, is you need to be comfortable with wearing a couple different hats. I am not a database engineer by trade, but nonetheless, unlike working at a big tech company where you've got seemingly unlimited resources and people can get incredibly hyper-focused and specific on very interesting and niche problems, we need someone who is a little bit more of a generalist and comfortable wearing a couple different hats. And so people who have backgrounds in statistics, who have sat in the chair at another company, sport adjacent is super helpful.

I think the other thing that, interesting contrast between my first role as a consultant and what I, in my current chair is, we, I get to sort of wear the consulting hat, but we're also the operator, and so there's a much bigger emphasis and focus on practical analytics that we can actually deploy. Whereas at least in my previous role as a consultant, we got to do all this super interesting and fun analytical work, but ultimately you kind of hand that over to the client, and then you go on to the next project. We have the burden of actually putting into practice what we build, and so having, you know, more of the mindset and sort of wanting to take ownership of the full analytical process end-to-end, which is a little bit different than what, at least when I went through school, there wasn't a ton of focus on that last part of like, how do you actually move a business forward through the application of analytics?

Handling systematic data shifts

Yeah, thanks, Rachel. Chase, love, love the content so far. You know, one thing, you know, that affects a lot of us that work in financial services are systematic shifts in data, right? So think about something like COVID, right, drastically affected consumer buying behavior, and it really wasn't something we had seen historically, you know, in our data lakes and data warehouses, right? So certainly complicated my job and my team's job a lot, because a lot of our models were misfiring pretty, pretty badly, right? I've been fascinated by like ESPN's FPI, and some of the stuff they do in Georgia Tech related to college basketball, but I'm thinking about specifically with the Cubs, right, and sort of like the transitional state they're in, in the organization, specifically, you know, trading way, Schwarber and Baez. I would have to imagine that's going to make a lot of your historical data a little less useful for you and your team. Do you have any tips or strategies, or have you guys started, you know, digging into this and seeing what you can do to sort of like make your maybe previous models better fitting for the new reality you guys are going to be in?

Yeah, that's a great question. You sort of hit on the fundamental crux of what I think it means to sit on the business side of a sports organization, somewhat in an abstract sense. All teams, sports are cyclical by nature, right, and that's due to a variety of reasons, mostly a lot of it's team performance related. Our job on the business side is to smooth out those natural team performance cycles, and, you know, when team performance is really good, our job is to push the business forward, and then when team performance takes a step back, like we just saw at the most recent trade deadline, it's to sort of mitigate the loss or put a, you know, don't let the floor go as low. And so our job is to mitigate sort of that natural team performance cycle and move the business forward, somewhat independent of where we're at in any given team performance or competitive dynamic cycle.

More mechanically, to your point, we are in a drastically different demand environment and just sort of team performance cycle today than we were even four months ago. What I have done from a practical standpoint is, you know, we have pretty good historical data, and if you go back to 2010, 11, 12, 13, 14, the Cubs were not terribly competitive, and so I have, what I, what we have done is put a little bit more emphasis on more further back in time, but data and demand profiles, that is a little bit more representative of where we can reasonably expect team performance to be moving forward. So, you know, just think from like a weighting standpoint or just kind of the types of data that we emphasize in training our models. I'll say to some extent, we do try to directly incorporate some team performance dynamic elements into our models, whether that's a competitive lens or playoff probabilities. There's some somewhat easy things that we can calculate that we can sort of update over time, but it's a hard problem.

Advice for aspiring data science leaders

I'd love to hear what advice you have for maybe someone who's on the call right now that's hoping to get into a leadership position, and how they could move from being that individual contributor into leadership.

I feel like the biggest jump there, it's probably not in the hard skills category, I think, in order to move up. I sort of think of the hard skill, you know, the proficiency and excellence in what I think of like the hard data science type skills, that's somewhat of a table stakes criteria. So you need to show a sort of minimum level of competency or excellence in that category, but where I think you can start to differentiate yourself is more in the soft skills. It's the leadership, it's the negotiation, it's the team building, it's sort of all the stuff that, you know, can be described as like the soft fluffy skills.

But when you think about, I think, high performing organizations have tightly aligned strategy and data analytics functions. And so when I think about what does a aspiring leader need to do, it's starting to develop a knack for what is the strategy of the organization, and how can I influence that. Data, in a lot of ways, is a tool that can very meaningfully improve organizations. But it's sort of, I think you can really differentiate yourself on sort of outperforming your competitors, in some sense, and those soft skills.

So you need to show a sort of minimum level of competency or excellence in that category, but where I think you can start to differentiate yourself is more in the soft skills. It's the leadership, it's the negotiation, it's the team building, it's sort of all the stuff that, you know, can be described as like the soft fluffy skills.