Data Science Hangout | Melissa Perry, Peloton | Design Thinking with Data
We were joined by Melissa Perry, Senior Manager, eCommerce Analytics at Peloton Interactive, Inc. Melissa is a value-driven data science leader with a passion for growing the next generation of data experts and moving beyond reporting, into optimized decision-making. ► Subscribe to Our Channel Here: https://bit.ly/2TzgcOu Follow Us Here: Website: https://www.rstudio.com LinkedIn: https://www.linkedin.com/company/rstudio Twitter: https://twitter.com/rstudio To join future data science hangouts, add to your calendar here: rstd.io/datasciencehangout (All are welcome! We'd love to see you!)
image: thumbnail.jpg
Transcript#
This transcript was generated automatically and may contain errors.
Hi friends, welcome back to the Data Science Hangout. If this is your first time joining, it's so nice to meet you. I'm Rachel. If you want to say hi in the chat, please do. The Data Science Hangout is an open space for the whole data science community to connect and chat about data science leadership, questions you're facing and what's going on in the world of data science. I'd like to remind everyone the sessions are recorded and shared to the RStudio YouTube and our Data Science Hangout site. So you can always go back and rewatch or find sessions that you missed too. We also have a LinkedIn group for the Hangout. This also helps you find people and connect with each other, continue discussions.
But together we're all dedicated to making this a welcoming space for everybody and we love when everyone can participate and we can hear from everybody. So no matter your level of experience or area of work, there's always three ways you can ask questions. You could jump in here by raising your hand on Zoom. You can put questions into the Zoom chat and feel free to just put a little star there if you want me to read it out loud instead. Or else I could just call on you to introduce yourself and add some context. We also have a Slido link where you can ask questions anonymously.
And also wanted to mention that if you are currently hiring, we love to be able to have people share those open roles in the chat as well. So feel free to use that chat to share links. We don't see that as spammy if you want to share open positions there. But with all that, I'm so excited to be joined by my co-host for today, Melissa Perry, Senior Manager E-commerce Analytics at Peloton. And Melissa, I know you've been here in the audience quite a few times. So great to have you here as our leader for today. We'd love to have you introduce yourself and maybe share a little bit about your role in the work that you do. Maybe also something you do in your free time as well.
Thanks, Rachel. Thank you for having me. I always enjoy coming into this community and hearing from the other leaders and practitioners in this space. So I'm really pleased to be here today. Well, I've been in business intelligence and data science for the majority of the last 10 years in my career. I love helping leaders make decisions. It's part of what I do, just taking a really complex environment. And I feel it's part of my purpose to help people understand the truth and most recently help people understand what's coming and what I recommend they should do because of what's coming. So that's where I really find my passion.
I've worked in many industries. I started off my career in Toyota and purchasing and being in a very quality-driven manufacturing lean environment. And then I've also worked in public education. I've worked in telecommunications. And I would say that my role in telecommunications is an analytics engineer. That's where I really kind of kicked off my obsession with moving into data science and learning how to automate decisions. There was just data everywhere in my role at U.S. Cellular Corporation. I also became an R user there, heavy R user. And with that, that exposure, I decided to level up my skills and pursue a master's in data science while I was an analytics engineer there. I went to the University of Eau Claire, Wisconsin Eau Claire, and just recently completed that. And I'm obsessed with this problem of predictive maintenance. It's one of my passion areas. How do you help an organization predict when something will fail and thus do all of the activities upstream to prepare for that event?
I just love it. But my most recent role, as you said, has been coming onto the team at Peloton. I joined at the height of the pandemic in February 2021, and I adopted a little pod of people called the e-commerce analytics pod on our data science team. And we built a dream team. Some of them are on the call today. Just great colleagues and friends and just really smart people that are passionate about helping our stakeholders to make decisions around, gosh, you name it, extended warranty. Diana's talking right now. She has a model called serial number event stream where she helps predict when our bikes will fail and helps our supply chain team prepare for that. And gosh, I just can't say enough good things. Our referral program, our member support agency, how can we help those folks to provide our members with a better post-purchase experience once they buy one of our connected fitness units?
This has been an incredible journey for me because I stepped into formal people leadership when I came onto Peloton. And not only did I take on a team and get to work with all these great people, but it's no secret that it's been a lesson in change management for me, heavy change management, every three to four months change management. So I'm passionate about sharing with you about that journey, and I'm happy to answer any questions you might have about what that's been like doing analytics in that kind of environment and building out my team.
Oh, cooking, definitely cooking. I am all about the kitchen. I like using cookbooks. I really like to be methodical and get everything ready to go so that I can enjoy myself while I cook and listen to music and things. But I work remotely, and that's something that I do in the evening, that I just love the ability to do that. I love it. I noticed that a lot of people say cooking on here. Maybe there's some relationship between data science and cooking for sure.
Growing and developing the team
So I inherited some team members who were already here. There were two when I came into the e-commerce analytics space, and then I grew people. I like a diversity in thought when I approach building an analytics team. So we took folks with direct domain experience and the member support agency who are interfacing with customers because I was trying to improve that e-commerce experience. We thought, well, let's get some folks that know it best. They've come from the front lines, but they also want to grow their career in data science. So I've grown people from that level, and then I also grow people who are kind of mid-level in their data analysis career and kind of get them to the next step.
And the most common pain point is like... And I got here in my career where it's like you spend all this time building dashboards or building a presentation, and then you're like, okay, what next? What are you going to do with that? And so I like to throw people over that loop and get them thinking about how do I deliver insights? How do I deliver recommendations? Kind of like my favorite part of it. But I start pretty nascent. At Peloton, we have in our central data science team, we're kind of like a center of excellence model. And so we sit right alongside our data engineering team, and there's quite a bit of dev work that is expected from our data analysts all the way up to our data engineers and data scientists that might not be expected on all the other teams that I'd work for. So I start real simple with folks. I pull up the terminal, and we start with Linux commands and things like that and build from there. And then we start with Git. You've got to have some development skill on our team. And so we practice what that means and kind of help them understand that I'm about to throw you for a loop here. We're about to do some things that are uncomfortable, but it's okay. You're supported, and you're going to be fine.
And then beyond that, we start exposing our team members to requests that come in that could be helpful to folks, like reporting. So there's always a need at Peloton to help people across the company understand what's going on, how many people are eligible for this promotion or this particular refund event because of where they are in their post-sale process. So there's always opportunities like that where you can really get a team member's hands dirty early on, and they can understand the data warehouse, the Mart, use BI tools. We use Looker and Tableau here at Peloton. So how can you create that report?
And then after someone demonstrates their ability there, I start to put them in front of stakeholders more so and start to build that consult up. And the consult for me is really based in ensuring they can do design thinking and build empathy with that stakeholder and start to recommend, here's what you're asking me for, which typically stops at reporting, right? But here's what I think you need to be asking me.
Design thinking with data
So design thinking is an approach to really getting to know people and their pain points as you're trying to help them. So on my team, we have like four to five strong stakeholder groups where I have a, not a dotted line, but a real strong relationship with a leader at Peloton. So let's say it's our extended warranty program. I get to know what's bothering that leader data-wise, like what is, what is, what are you on the hook for? What are you struggling to measure? How much are you doing? How well are you doing it? What is your goal? And I try to help my team to build the same level of consult and empathy with their counterparts on that leader's team so that they can really provide helpful solutions with data.
How do they design their approach? Like what is this model that we're going to build really be able to do? What are the actions that this person is struggling to do and how can I use data to help with that? So it's really kind of like a brainstorming activity, but you're putting yourself in the shoes of that stakeholder, getting to know them, figuring out how you can be most helpful.
How do they design their approach? Like what is this model that we're going to build really be able to do? What are the actions that this person is struggling to do and how can I use data to help with that? So it's really kind of like a brainstorming activity, but you're putting yourself in the shoes of that stakeholder, getting to know them, figuring out how you can be most helpful.
We do. We have a couple things that we do at Peloton that have really helped us to communicate about our successes. We have a newsletter that we work with our central project management team, and we share our wins. We also share some things that we're struggling with that we call like dirty laundry. It is just, you know, it's a startup, so nothing's perfect at Peloton. But we share how we're working through some of those issues. And then my favorite thing that we do is we have what we call a data town hall. And because we're that central analytics team, sometimes it can be hard to have that sense of community with all the other analytics teams that are at Peloton helping make decisions. So we set up this data town hall event, and we had presenters from the e-commerce analytics space that talked about some of our most exciting models. Like Diana Morrell, who's on the call, spoke about her serial number of end stream model and all the use cases that have come from that, and kind of showcase this design thinking element and how she thinks like a data scientist. Those kinds of opportunities are really powerful.
And in that same data town hall, we also talked about tech stack. We talked about programming approaches to ensure that our data models are well taken care of and documented for others, and made sure that people knew how our data engineering stack was evolving. Now, if you're asking me externally if we post, we do not at this time, but I'm always encouraging my team to blog and share their great work.
Working with non-technical stakeholders
So one of the challenges first is communication. And figuring out how to resonate with a non-technical stakeholder when your team and those that are working with you are primarily technical. You have that challenge of building trust. Trust is the first. That's the biggest thing. Right? They have to have an awareness and a belief that you're doing everything you can to provide that data product that's going to help them do better. And so I do think it does go back to that design thinking, that empathy, and having enough humility to ask people, how can I really help you be better? And not having that belief as an analyst that you know all the answers, because let me tell you, you do not. Numbers and data are only modeling what we have been modeled. They're a representation of the truth. But on the ground, operationally, there's always so much to learn. And you've got to really build those relationships with the folks that are on the ground that you're trying to help so they know you care. So they just know you care.
And I think once you get over that hurdle and they know that you're trying to help them be a better version of themselves and do better at work, then you've got that inroads to start passing them. Okay, here's this report that you were asking for. How can we make this better? Or hey, I see you're really struggling with this. Did you know that this is the journey that you're on and you've made this much improvement? Or you know what, it looks like we're getting worse in this area. I'm seeing a negative trend here. So it's really trust first, then communication and figuring out to what extent do you need to read into that person's technical capability and dial back all the things that you know. And this is the hardest thing for me to tell my people. It's like, I care about your technical know-how and your technical progression, but your stakeholder oftentimes does not, especially if they are a non-technical person. It just overwhelms them. So it's definitely a soft skill in an area that must be developed in order to truly get use cases into production.
Transitioning from individual contributor to manager
So I have been here. I think when you're ready and you want to try it, you should raise your hand and express yourself. That's the first piece of advice. You need to tell your leader. You need to tell your mentor. Hopefully you have a mentor who is, that's another piece of advice, is to get a mentor that might be doing that work that you want to do and start to talk to them, hey, how'd you get here? What did you do? So that you have that different advice coming in from different perspectives and you're getting, checking yourself because that's a big hurdle. That is such a big hurdle to clear, moving from that IC role to a manager. It's a very vulnerable state. So get someone else who's in your corner who can help you on that process.
And then the third piece of advice is to brace yourself for like a year, year and a half of constant change and wrestling with your own brain because it's really difficult to take your hands off the keyboard. And so I would say to give yourself some grace in doing that. Nobody gets it right. In the first, in that transition period, like there's always going to be stumbling points where you're going to make a mistake. You're going to step on somebody's toes. You're going to make one of your ICs feel like you're micromanaging and you're really not. You're just trying to make sure it goes well, but give yourself some grace and use that mentor to help you navigate those issues.
I don't know if anybody else has this experience, but being maybe specifically a woman, being a kid who's like meek and always told to like be quiet and not be a bother to anybody, it was a huge lesson for me as I moved through my life that I actually had to verbalize and ask for what I wanted. No one was going to just hand it to me because I deserved it. I had to say, hey, I want to do this. Can I do this? And that leaves you very vulnerable and open to somebody saying no or saying that they don't think you're ready or whatever it is, but you can sit around and spin your wheels forever if you don't at least ask or put it out into the universe that you want it in some way.
Libby, I want to respond to that and just relate back to the room because when I first raised my hand, I heard no and I had that mentor who was in my ear saying, but you're ready and Peloton said yes and so did other companies too. So just because you hear that no doesn't mean that the universe doesn't want you to do that. I love the way you said that Libby and it's so important for everyone to hear. So the person who asked that, go for it. Go for it and try it.
Libby, I want to respond to that and just relate back to the room because when I first raised my hand, I heard no and I had that mentor who was in my ear saying, but you're ready and Peloton said yes and so did other companies too. So just because you hear that no doesn't mean that the universe doesn't want you to do that.
Tech stack and team development
So the tech stack we have at Peloton here is we use SQL, use AWS Redshift for most of our data manipulation work. We use Looker as a BI front-end tool as well as Tableau for some of our reporting work that gets done. And then as we go further, then just the BI work, we use Jupyter Notebooks and Python to do our coding. And of course, like as I said earlier, all of this work, our development process requires a little bit of dev because you do need to share your work in a repository so that others can use it.
I expose them to the tooling based on where they can add value in the moment. And so if there's a tool like Tableau where I know it's so complex that you can't just sit down and start, we will pursue formal training. When it's a tool like Looker that's kind of more approachable on the front end, I might sit and help them through working in that tool. And we do some internal training as well. When it comes to like helping somebody make the shift into R and Python, I do personally recommend kind of stepping away from work and learning through a class or a boot camp where you're kind of making sure that you understand the foundations of what's going on. Or like in R, I learned through the Swirl library how to just program and get answers out of R.
We also try to provide education in fun ways at Peloton. Like we have one of my colleagues, Will, he had this idea to do a data labs concept where anybody on the team can come up with this idea. Like, oh, I wish somebody was working on this at Peloton. And if they just want to try their hand at solving a data science problem, we create a space for that where we kind of nurture the problem, help them work through it with the resources we have and coach them up so they can get their feet wet and get some exposure and get excited about data science, even if it's not their main job.
Keeping data scientists motivated
One of the very first conversations I have with my people is like, where are you going after you leave me? Because inevitably, like, we're not going to work together always, right? But my job, like, because I'm so people-oriented, is like, how do I make sure that this experience and this relationship is beneficial for the both of us? And I think in order, that's the conversation where you start to figure out what's motivating people.
Like, when I came to Peloton, I shared with my leader, like, I really want to work on figuring out how to get use cases into production. I've been struggling with that for years. Like, that was something that I said. But I've got people on my team that are like, you know what, I want to be a data scientist. So I'm like, okay, well, let me go figure out how to get out there and find them work. And so through my relationships and helping them empathize with people and figure out what their problems are, we find use cases. And I am like kind of out there talking about their work, how they've done such a great job on this dashboard thing. But, you know, really what I'm hearing is that you're really trying to make this decision.
And so, like, let's just use extended warranty as an example. My team was delivering great reporting on sales and claims for a third party warranty provider. But, you know, I had a data scientist sitting on the sideline that said, you know, what if we could predict who was best positioned to buy an extended warranty? And I said, go for it. Don't just sit there. Go for it. And so I kind of have them work on that in the background. And then I go and do the politicking to go, hey, James, I've got something for you. We're doing all this on the side that you've asked for. But you know what? I had a person that was really curious. And they were curious about delivering this kind of value. Why don't you check it out? And because we've done so much on the front end to answer that first question and deliver value of what they actually asked, they'll receive it. And then that use case continues to grow. And then it gets into prod if we're lucky. And that use case actually got turned into a feature that is used for marketing. So one way or another, that's where it comes from. And then that person grows. And then their mind grows, their eyes open. And then here they continue. Now they're a data scientist.
The center of excellence model
I mean that there's a team of folks in what we call the data science team, or the data and analytics team, and they serve the business in this, like, central function. They're providing frameworks, the tech stack, places to document things, process, structure, training, training on the tech stack, like, expertise. And then beyond that, you've got other analytics teams that are throughout the company that are also trying to make their, help their leaders make decisions, but probably more rapidly, more rapid fire. They would utilize the expertise and the tech stack from the central team to help their leaders.
And usually, their leaders want things yesterday, and, you know, that central team doesn't necessarily have the capacity to, like, intake and work all that, but they do have the capacity to keep things clean and organized. How's that working at Peloton? Well, we're continuing to evolve our story there. We're in a space now where we've got a little bit of a patchwork going on. We're, you know, growing out of a startup phase, and that's natural. And so we hone it all in and then send it all back out with our new leadership. Building those relationships, standardizing some of those relationships with the teams that are depending on that center of excellence. So it's a lot of work, but good work. Important for a company of our size to get alignment. That way your leaders don't have multiple versions of the truth.
Making the shift to people leadership
My question was, what made you want to make that initial shift from individual contributor to people leader in the first place? And did you have any concerns or reservations going into that leadership journey where you were like, okay, let's try this? I'm worried about these things. What did that look like for you?
I realized very quickly when I was a business analyst working in education that I like people. I like seeing other people do good work, improve their work using data, get more strategic, be more successful. I like using data in that way. But I recognized very quickly that as an IC, I wasn't necessarily the one that was achieving that. I saw my leaders doing that. I saw them rubbing elbows and doing the politics to get the data in use. And then I'm like, okay, I want to learn how to do that because I got to this point of frustration where it's like, how many dashboards can I build before my brain just explodes and nobody listens? And what are they even doing with this data? And why would they do that? And how does this company even make money? What's the strategy here? And I just found my backgrounds in economics, I just cared more about that. I just cared more about that social element of it and helping the work become a success, making value out of the work than I did actually touching the keyboard.
Now, don't get me wrong, every once in a while, I'll pick up a ticket and just vibe. It's a flow when you're doing something. I still feel that way sometimes when I do some work. I'm still a nerd. I still like to talk about how to model things and what data goes in. But I wanted to do it from a different angle. And as I also like to help my stakeholders do better, I really like to grow people's careers. That's like nirvana for me. Once I can see somebody, that analyst grow from a data scientist, that was like a crowning moment. I'm like, okay, yeah, this is right.
Now, like I said, I had people tell me no when I first stepped out as a leader. I don't know if you've ever heard, she's not technical enough. That was funny. Had to work through that. Still a trigger point for me because here I am. I'm social. I like to help people. And then so it called into question my technical ability. Well, that wasn't very good for the ego or the confidence. But I've worked past that. In a weak point, sometimes I'll go there and have that imposter syndrome. But it's my people and my stakeholders that build me back up.
Biggest needs at Peloton
Yeah, I think right now it's establishing one source of truth and gaining alignment. Like when we were talking earlier about the center of excellence, it's like refining and truly being a thought leader for what does that mean for how we behave, how we how we treat one another, how we communicate and talk to each other about data. What does it mean for evolving thought leadership around concepts like data governance? You know, at a fast moving company like Peloton, you say data governance, like people are, what do you mean? Like, you're about to slow me down. And I'm like, no, like, we literally we can't. Like, you have to give these folks over here a shot. You have to listen. And I'm listening and I'm going to adopt this and I want you to too.
So, like, I think it's figuring out how to show up in the center of excellence space and bring along all my colleagues. I've got like 30 stakeholder teams and like figuring out how to have them show up in the space and truly work together like a Peloton, you know. I'm sure I'm not the only person who like, you know, your company culture says one thing and then you get into the data space and it's like a completely different thing. But we can probably all laugh on the inside about that.
Thank you so much, Melissa, for joining us. It's been so fun and really love getting to hear your insights and your experience. Really appreciate it. Absolutely. Thanks for having me. This is such a joy and I can't wait to see the folks that present in the coming weeks.
