Transcript#

This transcript was generated automatically and may contain errors.

Hey there, welcome to the Paws at Data Science Hangout. I'm Libby Herron, and this is a recording of our weekly community call that happens every Thursday at 12 p.m. U.S. Eastern Time. If you are not joining us live, you miss out on the amazing chat that's going on. So find the link in the description where you can add our call to your calendar and come hang out with the most supportive, friendly, and funny data community you'll ever experience. I would love to introduce our featured leader today. We're talking to Jamie Shive, Director of Data Science at the National Hockey League or the NHL.

Sure. Hi, y'all. So if I move very slowly, my camera doesn't lag with my voice. This was an issue when we were trying this out. So I'm going to move as little as possible so that my voice syncs with my mouth. So if I look insane, that's why. So hi, I am Jamie Shive. I'm the Director of Data Science at the NHL. I have worked at the league for a year and I believe around four months. So for the hockey fans in the room, I started at the league in the middle of February. I actually started in the middle of the Four Nations Face-Off tournament, which was a very, yes, I love seeing a lot of nods.

I am a hockey fan. I'm arguably much more of a hockey fan now that I work here. A little bit about the work that I do. So Director of Data Science, I oversee our fan data. So we're doing predictive and descriptive models and analytics on our fans to help improve the fan experience. I lead, I'm a part of a team of three. So I've got two wonderful, brilliant, incredibly kind and amazing data scientists that I oversee. The whole point of my job is to help meet the fans where they're at, which is nice because selfishly, if I do my job really well, I get to have a better experience because of what I do.

Jamie's background

I'm a mathematician. That's what I was trained in. I have studied at Virginia Commonwealth University. I got a bachelor's in Applied Mathematics, went on to do my PhD in Systems Modeling and Analysis. Actually, my research focus was in discrete mathematics, graph theory for any mathematicians in the room. So that is what I did my dissertation in. I had an extraordinarily nontraditional path to find my way into sports. This is not the plan. This is not what I thought I would be doing when I was growing up. It is where I landed. I could not be happier to be here. This is the right place for me to be. My fun fact, one interesting thing about myself is that I do recreational ballet now as an adult, but for the about 15 years when I was growing up, I did ballet very seriously and I wanted to be a ballerina.

Oh, mine was gonna be so boring. It's watching hockey. That's what I do in my free time. I am, I call myself club agnostic or club secular depending on how you want to think about it. I don't have a team necessarily. And it's probably for the best interest of the league if I just remain impartial here. But I do, well, I will put on any hockey game that is on. In order, though, I like to watch any sport. So it'll be hockey first. Baseball if it's on, basketball if it's on, football if it's on, but it's usually hockey first. Ballet is what I like to do for fun. Now in my free time. I take a class once or twice a week, just recreationally as an adult.

Fan analytics at the NHL

Yeah, absolutely. I did just see pop up in the chat. Somebody asked why and why we don't call it fan analytics. We do. Oh, we do. Trust me. It's not popular. I have been told that it is corny and cheesy to call it fan analytics and I do not care. I'm corny and cheesy and I'm going to call it fan analytics anyway.

So working with fan data, so folks will come and interact with the league in some way for the most part. When you're a hockey fan or someone who's gone to a hockey game or something like that, you're a fan of a club, you're a fan of a team. And those teams will collect data about when you go to games and what you do. So sometimes that data will get rerouted through the league if you come and interact with the league in some way. So sign up maybe for an NHL.com account, maybe play bracket challenge or something like that. And then we will take a look at sort of the behaviors, the ways in which you engage, try to find trends. And through those trends, we start to take some action items off of those. So what is driving you to go to a game? What is driving you to make a purchase? When you go to the game, what do you do in there? Do you buy food? Do you buy merch? Do you get to the game 20 minutes late because there's traffic? And how can we help make sure that you have the smoothest experience possible?

Maybe a project that I've overseen, for example, we did a fan sort of personas clustering exercise. That was something that I oversaw. So that 30 or so million fans in our warehouse right now, all with unique personalities and unique ways of engaging with the league. But in some senses, there are ways that we can group you all together to make it a little easier to digest and figure out how to action off of you to meet you where you're at. So we took the 30 million fans and squish you all down into about 14 unique personas of fans based on their behaviors, digital engagement, so ways in which you've come and interacted with the league specifically, just so that we can figure out how to better reach you. It's a lot easier, instead of trying to find 30 million different ways to reach a fan, if we maybe just tried 14.

Yeah, so Python and R over here, it's pick your favorite. So both of my data scientists and then a variety of other analytics folks under the team, they get to pick whichever they want, whichever makes the most sense. And through doing that, there's a little bit of learning that everyone can do. So everyone can kind of do a little bit of both. I like to think that everyone here can read both languages, which is a good start. So Python, R, we've got Tableau as our BI tool. We've our Snowflake Warehouse. And so SQL all day. I think that's our standard tooling here.

Night and weekend work in sports

Boy, do I get out of here at five. The nice thing about, I would say, so fan and like fan work specifically. So not working with, I'm not working with hockey stats data, I'm not working with things that are happening in the game. Fortunately, for me, I'm not the one who's putting the like very fun and niche stats up on the screen during the hockey game that says like, this is the second time someone has gotten four goals on their birthday. So I don't have to be a part of that. Though, that would be lovely. But working with fan data, there's a little bit of a slower pace. I don't work nights and weekends unless I'm feeling particularly passionate to do so. But that's a me thing and should definitely never be anything that anyone else takes from me. And the nature of fan data is that I like to be strategic and thoughtful. And so it's a little bit slower moving. And that means that I build in a lot of extra time, under promising over delivering on a variety of projects to make sure that myself and my team we don't, we can leave at five on Monday through Friday.

League vs. club analytics

So the way that the data exists at the league versus at the clubs is quite different. So we've got every club has its own data team, just like mine, and own data warehouse, own set of tools, own philosophies around what it means to be data driven, and how that shows up and who they support and what verticals they support. So they are doing their own data and analytics largely around getting season ticket members or like how to fill the arena with people, what are the right theme nights to host? How do they get people to buy more hot dogs or more popcorn, or put the beer in like a really giant saber or like an ice skate boot? And then how much is that going to sell? That's the sort of stuff that the clubs do. And then we at the league over here, figure out how to engage the fans every other way, hitting them with emails during the offseason, trying to get them to come and play games in our game zone. And using analytics to drive those. So finding the fans who like to come and play and then telling them to maybe make a bracket during bracket challenge. And then we will take the results of all of the analyses that we do and share them back with the clubs so that they can help that can help them make more informed decisions about the fans that they're working with.

I also in my effort, when I started here to learn as much as I possibly could about the history of the league and the clubs and how we could best support them, I set up time to talk to sort of one data slash analytics counterpart at each club. So 32 different folks, one from each club, ballpark in the data science and analytics realm. We now talk to each of those people and some others, maybe one to two representatives from each club quarterly, where we talk about the projects that they are working on. If they're considering a switch in vendor, moving tools, maybe moving their warehouse, migrating that something like that. If they are getting pushback from their C suite about trying to implement AI in some way, we talk about all of those things. It's very casual.

I have worked really hard to maintain those relationships, because I think it's really important to know what the clubs are working on so that I can figure out how we can best support them. Most of the time, my support is commiseration. But sometimes it does turn into actual projects where, say, my data scientists are working on the personas model that they were just working on. What if we took that and filtered that down on a club by club basis, so every club could see their club sort of against league average and with a personas model like that, that tells the clubs what makes their fans unique to them against sort of the overarching hockey fan. That's interesting information. And how many of them are in each persona? And what can they do off of that, then we can have a conversation about that.

I also host a business strategy and analytics workshop that happens in March, where I get all of my, all of those people, all of my friends, we are all in one room where they talk to each other about what they work on the problems they're having. It's basically just a rinse and repeat of the conversation except we've got like a giant community. That's my favorite part of my job.

Fan persona segmentation

So I do. One thing I will say is when it came to creating these personas, one thing that I did sort of benefit from as a fan was being a fan. So when we did the personas exercise, and came out with sort of these 14 really crispy, distinct personas, I could look at them and say, yeah, I've seen that fan before. Like I know that one. Had I not had sort of that level of knowledge, I think I would have tapped in probably one key sort of stakeholder from each function of the organization to have them take a look at what we've produced and see if they have an understanding of who these fans are. If the answer is yes, great. And if not, it doesn't necessarily mean that we go back to the drawing board. But it does mean that we maybe try an additional iteration to see if what comes out makes a little bit more sense.

One way that we did approach this project that I don't think is new or novel, but did make a lot of sense for the business specifically here was to create personas that were just around digital behavior engagements. So it wasn't personas that included demographics, it wasn't demographics and psychographics and fan behavior, it was just the fan behavior. Because layering in the demographic data and the psychographic data produced giant clumps of nothing that told us anything about the people that were actually showing up in the building. We wanted them to be more distinct. I'm not going to get personas that represent all 30 million fans, but I can at least get them to ballpark represent the behaviors and then adding in the demographics and psychographics after to see where those people fell in those digital behavioral engagement buckets meant that we could take these personas and share them with every different vertical of the organization. And if they wanted demographic information, then we could add that in. If they wanted psychographic information, we could layer that on top of these personas, rather than having those be the defining features.

If they wanted demographic information, then we could add that in. If they wanted psychographic information, we could layer that on top of these personas, rather than having those be the defining features.

This is the reason that I cared so much to do this project is, okay, so we know who's in the building, we know how they're behaving. What do we want them to do? Like, what's the goal? What would be the goal of knowing who they are, if it's not to take an action item off of it? So my overarching goal has always been, find people in a cluster, find people who are this persona, and see if you can move them. And what would it mean to move them? And can you? Is really the more important, I think, the more interesting question.

One of the behaviors specifically is like email engagement, who's opening, who's opening emails, who's clicking on things. And say, we've got a cluster of folks who always engage over email and love to play in sort of our game zones, playing bracket challenge or lines, and we have a whole game zone, by the way, this is not an ad, this is just, it's fun to play in there. But, they don't buy merch. They just, they're kind of just like our chronically online folk. They don't really buy merch. But, they don't buy merch. So, do we want them to? Are they going to? There's a couple of folks in there who are right on the cusp of maybe do buy merch. Why don't you hit them with an email? You know they open emails. Like, go and give that a shot. If you're going to try to get them to do a behavior, tap into the ones that they're already doing. Maybe make an advertisement or something on GameZone or add in some incentives or something like that to have them say buy merch, if that's the goal here.

The evolution of fandom

So this kind of started, I'm sure it predated me, but I've always found this interesting and would like to continue the work to see if this sort of trend continues. This is largely coming from sort of our social space, social analytics and digital analytics is the evolution of fandom, meaning how do people enter in and become fans and what are they, who are they a fan of and how do they sort of engage as fans?

So to give an example and sort of where I'm going with this, when I think about fandom and how I grew up becoming a fan, it was my family as a fan of a team. I'm immersed in it. I was born into it. I didn't really have an option to be anything but the fan, the team that my parents and siblings and friends were fans of. And that's how I became a fan. That's the entry point was through my friends and family. And I'm a fan of that team. Die hard. You cannot convince me otherwise. Doesn't matter who's on it. Doesn't matter who's leading it. That is sort of the traditional fan model. That's how it's been in hockey for a long time. We're seeing the early signs of evolution from fan of team to fan of players.

Players move teams. So the change then is we're approaching fan analytics sort of from the like on a club by club basis. This is what your fans are doing. But I'm thinking, I mean, we're years away from this. What it would be like to change the perspective of like fans of this player behave in this way. The reason that this is happening, the why of this and sort of we've seen this a little bit more in the data, the way that we've seen this show up in data is people changing their favorite team on the website. Like what?

The why behind this change is theoretical, still early and anecdotal, is the community now is sort of expanded to your social media community, your friends, your acquaintances, people you follow, people you interact with just through DM and not really much more. They're the ones who are putting the players into your universe. So it's less of a team focus and now more of a player focus. And so the avidity is for the player and players can move teams all the time.

The why behind this change is theoretical, still early and anecdotal, is the community now is sort of expanded to your social media community, your friends, your acquaintances, people you follow, people you interact with just through DM and not really much more. They're the ones who are putting the players into your universe.

And so the use of social media by and large for our younger generations. And so when we sort of have been doing these analyses on like, okay, who likes teams, who likes players, and there's a variety of different approaches to that. But a lot of it is when we're getting the signals that either someone is engaging through social media in a variety of different ways with a variety of different teams, or they're actually literally telling us that they have a new team. And then we can go and look and see like, ah, they changed this and that team made a transaction with this team. And now that player went there and we've got it. It's pretty clear. That's why that's happening. And so of course, that is where we are seeing it when we overlay demographics, it did tend to be the younger generations.

This is important to know, because those are the ones who are going to be filling the building and are filling the building soon. And as a gate driven league, meaning we make most of our money, basically all of our money by people going to games, it does matter knowing that the people that are in your building are there with team spirit, but really with player spirit. And they've got their own sets of behaviors and things that motivate them. They have their own economic and societal factors that we need to consider in order to meet them where they're at.

Building a diverse data science team

So, yes, yes, we need more. We need more diversity and background. Truthfully, to do a job like this. If you're on this call and you work in Python, you work in R and you do analytics like that, that's just the bare minimum. That's not something that most people can do it. It's the psychology element of it and sort of understanding behavior is something that I think is really important, because that is how you see a trend and can put yourself inside of it and think to yourself whether this is something that makes sense or not. And if it does make sense, you can speak to it. And if it doesn't make sense, then you can go and figure out why.

So, of course, my background, pure mathematics, someone on my team also has background in mathematics and then another sort of in sort of like sports science and sort of analytics based on. So that has also been really helpful because she comes at every project with this like and then this is what happened on the ice and this is what the players were doing and this is how the fans responded to that, which is like that I can't, that's quite quite outside of me. The more diversity that we have on the team, the better.

We're talking about fans here. Fans are diverse. And I want the representation of the fan base to be represented on my team as well so that we are reaching every fan where they're at, not just the ones that we hear the loudest.

Breaking into sports analytics

Building a portfolio, if someone wants to work in sports, pick a sport for starters, doesn't have to be hockey. If I were, say, someone they included a portfolio in an application, I would click it whether it was hockey or not. I would want to see a variety of projects that touch a lot of different aspects, areas of the business. It could be on ice or it could be fan. What I'd really want to see is, you've done an analysis. Tell me why it mattered. Tell me what you can do with it. So the action off of that. So you're seeing a trend of something. Who would you tell about it? Why would you tell them what matters from it?

My the most important job, most important part of my job and the job of my team is communicating what we see. It doesn't really matter so much if we just put numbers in front of people. The numbers are scary and hard and they don't want to see them. They want to know what to do with them. So being able to show that aspect of it somewhere within your portfolio, like here's what I see, here's what you should do about it. That's the part that I would care much more about. That's the part that I would take and say this person clearly has the technical background, but they have the soft skill, the one about communication that I care the most about.

The Heated Rivalry effect

Yes, very much yes, in that all of a sudden, they were there. All of a sudden, we noticed that the numbers were really different on demographic slices, who was coming to the games, who was interacting in a variety of different ways, who was new to the warehouse that we'd never seen before, and when did they enter, and was it after Episode 4 of Heated Rivalry or Episode 6 or 7? It's all of them. We absolutely saw it. A lot of this analysis lives with one of our social analytics, our manager of social analytics, Hope, who is wonderful. She is the one who is sort of in charge of the, well, the follower accounts are off the charts. The video views are off the charts. What are the demographics of those fans? It's they're new. We haven't seen them before.

It was coincidental but beautiful that it worked out this way, that one of our sort of league initiatives for this season, which is it's always there, but we were working really diligently towards this North Star of growing the female fan base, which is a great idea. We had sort of set up a variety of baselines and sort of had these ways to track and trend and do analysis and figure out where we were missing them, where they were coming in from, and how we can continue to nurture their fandom, and what do they need from us to keep going. It turns out they just need more of them, and that's fair.

So when Heated Rivalry and others came into existence, all of a sudden, everything was literally off the charts. Everything was off the charts, and we have a variety of wonderful employee resource groups here. One of them is for sort of the pride folk. The Heated Rivalry was posted in that Slack channel just as like a hey for anyone who's interested, and that was how I heard about it. Then the first episode happened, and then the second episode happened, and then we knew that something very, very big and different was happening for us at the league because we could see. Truly, we had nothing to do with Heated Rivalry, and that's okay. It's really lovely that that helped to show folks that hockey exists. Hockey's for everyone. Please come. Please enjoy it. You're welcome here, and that we could use that to sort of just nurture and grow.

Career advice

Absolutely. I do have a piece of advice. My piece of advice is when it comes to choosing where you work, so you know you want to do data science, you know you want to be in the data science space, choosing where you work is largely down to choosing work that aligns with your values. Doing data science somewhere, make sure that the work that you are doing aligns with your values. That is the only way to ensure that what you are doing nourishes you and makes you feel good at the end of the day because you're going to be doing it for eight hours a day.

The way that I do that here and the way that I ended up at the League here, my general core values are authenticity, transparent and honest communication, accessibility, equity, equality, and so everything that I do in the fan space is making the game as accessible as possible to all of the fans who are here and making sure that they can be themselves so that they can enjoy the game as much as possible. Every piece of work that I do touches those and so my day is filled with me touching my values. So my piece of advice is to figure out what your values are, find a job that does those.

That is the only way to ensure that what you are doing nourishes you and makes you feel good at the end of the day because you're going to be doing it for eight hours a day.