Evan: Hello and welcome to the Mining Your Own Business podcast. I’m your host, Evan Wimpey, and today I’m excited to introduce Selma Dogic. Selma is the senior manager of Analytics at Carter’s. Excited to learn everything that she does with data there. We know we can trust her cuz her background is in economics and she studied at Georgia Tech just like yours truly. So, very happy to have her on the show today. Selma, welcome to the show.
Selma: Thank you. Thanks so much for having me. I’m so excited.
Evan: All right, fantastic. To get started, can you give us a little bit of your background, how you got into analytics and sort of what you do now at at Carter’s?
Selma: Yes. It’s actually a funny start, at least to me. When I started. And I’m glad that you mentioned that I have a background in econ. I decided to go to school because I was really interested in helping use data to make better or informed policies, right? So like everybody who’s young and in college, I kind of wanted to save the world in my own little way through data, right? And so I had these big thoughts about going into economic development, particularly international development. And I started working in the nonprofit consulting space. So I did some consulting for state and local government. We would do fiscal impact analyses, economic impact analyses, again, really using data to inform better policies. But what I really learned from that was that I had a passion for data and I had an even bigger passion for analytics and I wanted to really dive into that. So I went to school at Tech and got my masters in economics to do the same thing. And then, I tell people I’ve sold my soul to corporate and started working for NCR in the analytics space. I was on a team of data scientists. After NCR I went to Home Depot for a brief period of time, again, working in analytics. And then that also really, I think, did a good, it really taught me a lot about kind of customers in the retail space. And so when my former boss from NCR asked me if I would be willing to join his team at Carter’s, I was like, this is perfect, right? Because I was doing all of this kind of learning and really getting excited about consumer goods, which is a little bit different than NCR that I was like, “This is perfect.” So that’s how I got how I got here.
Evan: Very nice. Now, I’m super familiar with Carter’s. I am a father to young children. So on the shopper end, I’m aware. But I suspect – three children, I wouldn’t have been able to pick Carter’s out of a lineup. So can you give us just a little of what Carter’s is, what you guys do?
Selma: Yeah, so we are the leading brand for baby and small children apparel. And we’re doing some things when they’re a big kid too. So, you know, as your kids are kind of growing, you can remain kind of a part of our customer base. We also have brands with Osh Kosh is one of our brands, Skip Hop, So Baby Gear, really cool stuff.
Evan: Awesome. Very cool. Yeah, I think to any parents of young kids listening they’ll be very familiar.
Selma: Yeah, for sure.
Evan: Yeah, I do appreciate, I appreciate the background and I’ve got a history at Home Depot. We’ve had, I don’t know if you had any overlap, Home Depot’s a huge place, but previous guest, John Weininger and John Carroll were part of the Met Analytics team, They were great to have, some really nice insights from Home Depot. Listen to those episodes. But for today at Carter’s, can you tell us where you sit in Carter’s? There’s certainly different functions. Do you sit with a particular business unit? Are you central?
Selma: Yeah, I actually sit on the IT side, but I’m a little bit of a unicorn in that I’m not really an IT person, don’t tell them that. But I really, my role in, I think that’s why I love it so much is that I get to be the liaison between IT and the business and I would say by and large, analytics roles sit on the business, right? So you, they’re kind of embedded in different business teams. But what I’ve seen, and even at Carter’s, right, it’s no different, but the data sits somewhere else, right? So the data is all kind of managed by IT teams and they oversee the schemas and all of that on the data landscape. And then you have business teams who are trying to make sense of it, and trying to even just access it can sometimes be, you know, a big headache. And so my role is to kind of bridge the gap between the two and to accelerate and enable analytics on the business side.
Evan: Okay, fantastic. Is the business side – you’ve mentioned a few brands that live under Carter’s. You’ve probably got other retailers that you sell too. You’ve got different business functions. Is the business like a specific brand or is it a specific function like marketing or supply?
Selma: Yeah, so we kinda, our teams are structured by function. So you have like the pricing team, the marketing team, supply chain teams.
Evan: Awesome. As you, as you are liaising between sort of the data and the infrastructure owners and the business stakeholders, are you prioritizing? Presumably there’s some competing needs, everybody wants, wants, wants. Do you – is that part of what you and your team do is help to prioritize?
Selma: Yep. Yeah, so the, and I’m wondering, I think it might help too if I kind of share a little bit about the way that our broader team is structured. I sit on the global data analytics team, right? And so under that umbrella, we have integrations, data engineering, bi, and then my analytics COE. So we have the data engineering team who works with the integrations team to really kind of bring all of the data into our data landscape, right? And they do all the ETL and everything. The BI team is working with the majority of our business insights consumers, right, who are leveraging dashboards and data to drive. My team is interfacing with a smaller portion of these analytics-based business teams to enable and empower them to do to build analytics solutions. Right? And that can come in the form of they’re this. Stakeholders or the business sponsors, and we’re the builders, right? So we are hands on keyboard, developing analytics solutions, analytics models and deploying them. Or it could be through the lens of consulting and enabling them to do it themselves. So giving them the right tools, technology and training to do it themselves. And when it comes to priorities, right, because we deal with smaller groups that are more in the. So, and we can kind of scale right, by building these hybrid teams with them where we’re not the only ones who are hands on keyboard, right? Either we’re consulting them or they’re consulting us because they have the business knowledge and we have the analytics subject matter expertise, right?
So, like I said, that’s the cool part about our team is that we can really partner with the business to build these kind of hybrid teams. But no matter what, right? No matter where I’ve been there. Always these competing timelines. Use cases. Business cases, right. So the way that we really prioritize is based on feasibility, based on business prioritization. And a part of that feasibility is not only like can we do it, but when we do it what decision. Can we tie to this project, right? What decisions are we enabling? And if we can’t really tie it to a specific decision, if we can’t draw that line, then it’s really not in our interest to take it on. And so we leave it in our backlog, right? We leave, we leave the conversations to be had later when we can kind of get more buy in, more sponsorship when it’s more feasible, right?
But it’s really important to us that every project that we take on has a direct line to a strategic initiative from the top down, right? It has to be tied into the decisions that our leaders are gonna be making.
Evan: Yeah, that’s, I think, boy, you really just highlighted sort of an ideal structure, an ideal case there. It’s great. If it’s not tied to a business strategy, it’s not priority. And if you can’t action on it, you can’t. If it’s not going to drive some. And there’s no value. Yeah, it’s tough as with the data mind and the data science. Some things are fun to explore, but that’s –
Selma: And well, and we have examples of that too, right? Where it was important to us because it was, you know, we had a big leader be the sponsor, and once we brought that project to fruition and had the results and were socializing those results, you know, my number one question is, Okay, well what decision are you gonna make from this? Right? Like, from, it’s just data until you do something about it, and we couldn’t tie it to a decision. And that, for me was a kinda a failure. Right. Definitely a big lesson learned and I just don’t think that any team has the bandwidth to do things like that, to do the projects. There are projects that bring more value, that are more feasible when you tie them to a decision.
Evan: Yep. Absolutely. Doesn’t matter. Accuracy or area under the curve or whatever it is, but not improving the decision. Maybe, and you don’t have to get into specifics, but maybe you can talk through what delivery looks like in, in a good case, you work on a project tied to some strategy. It helps drive a decision. Do you just hand over a model? Do you hand over some decision criteria? Framework? How do you deliver your results to the business so that they can make a better change?
Selma: Yeah, I think actually, so for a lot of data science projects, I think the first part of it where you do a lot of the discovery and the research is often overlooked. And then the last part of it where you do a lot of socialization and sharing insights and you’re doing a lot of presentations that gets overlooked. Right? Like as a data scientist, we’re really I would say, we get really excited in that middle layer where we’re doing the development, we’re doing the data engineering, right? The KPIs that we really care about are not the KPIs that are gonna empower the business to make the decisions, but we’re thinking about MAPEs and AUCs, right? And we’re thinking about ML ops and all of that. So I could tell you, like, my answer could be, well, we do all of our model monitoring and deployment and we have these triggers set up and all of that.
But the truth is, again, at the onset, we do a lot of discovery and a lot of research as we frame out our business question and like the timeline and our milestones and everything. As we do our model development, we take steps to iterate, right, to improve. Like not only are we like doing the data engineering and feature engineering to help train better models and you know, do that, but we’re also like learning ourselves, right? Like we’re again, because we’re really having to embed ourselves in the – there’s so much research that goes into that throughout the life cycle of a project, and we’re also developing it in a way that we can enable ML ops, right? So thinking about the way that we automate flows, the way that we automate triggers at the end of a what, when we produce our forecast, when we predict, when we produce our predictions, a lot of it is how do we, in how do we allow our business users to consume that and how do we serve these on sites? So we send all of our predictions back into our data landscape through you know, the ML ops framework that we’ve built out. And then we build dashboards. That’s kind of our primary way to communicate insights with our business leaders. And that’s the primary way that they consume those insights are through dashboards. But again, that doesn’t mean that we build a dashboard when we give them the link to it and that they’re using it every day to make decisions. Once that dashboard is out there, there’s a lot of work that goes into play into having them adopt right now, this new data asset.
Evan: Yeah, I think that’s great and I would love it. However, you can expand on how they tried to adopt that. I think probably to the listener, there were plenty of cases where we built a great dashboard that does exactly what. And nobody’s logged into it in the last six months and it goes unused. So are there any ways that once you’ve delivered, you’ve been communicative and got a dashboard that’s useful. Are there any ways that you follow up? Is there some process in place that you try to ensure that folks are using them and using the way they’re intended?
Selma: Yeah. So I think even if you’re, you know, building dashboards that aren’t related to any ML models that you’ve built, the method is the same, right? The first thing, if you’re replacing something that they’re already using on their team, I think that’s really, it’s not necessarily low hanging fruit, but that’s a really big opportunity for you because they’re already dependent right on something. They’re already making decisions off of some data. And now you’re serving it in a way that’s more efficient, more effective, right? By way of like visual analytics, more efficient, but through automation, things like that, right? But again, they’re already dependent on those data and using it to make decisions. So that’s already a small win for you. The way that they’re making the decisions is big. Right. So really keying in into their process and using that to inform the way that you build your ML ops architecture, the way that you build your dashboards, that’s gonna be important because that’s gonna help drive the adoption and making it digestible for them. Again, sometimes we, as the data scientist, as the analyst, we’re always kind of like craving that next gen, right? Like the big new thing in data and analytics and that can be hard for the end user to digest and consume. So, keying into their processes, really understanding how they’re making those decisions and servicing that first, like leaning into that. And then showing them, like walking them through, Okay, this is like the data that you’re using to make your decisions. You’re having to jump all of it out and create these pivot tables and analyze, and it takes you like three or four days, right? Here’s the dashboard that does all of that for you, plus it gives you a prediction, you know? There’s no heavy lifting on your part. You just have to consume it and make that decision. Really walking them through that and making it a reality for them, I think is important.
That doesn’t mean that they’re gonna be logging into it every day. There are still gonna be dashboards, right, that are gonna take time for users to adopt. Now, that’s where you have to key into a champion, and I have seen strategies really lag when we didn’t have the right stakeholders or champions tuned in. When we weren’t talking to the right people. When you identify a champion, somebody who gets it just as much as you do, but they’re on the other side of it, that can like transcend your analytic strategy and really drive adoption, and I’ve seen that happen.
Evan: Absolutely. Yeah. Can’t get more agreement from me there getting an analytic champion. It doesn’t matter how good the results are. And sure, removing barriers to use, but if you’ve got somebody who’s with you and the champion along the way, and the process then the implementation is certainly helped out.
So it sounds like you guys have a really well organized process on, on delivery there. I want to step back. Pre-delivery, pre-model build, thinking about the data that you use that seems to sit sort of in that centralized face. Maybe you can give some color here, but when I think about Carter’s, I think apparel that’s on sale at a bunch of other retailers, at third party retailer, department store or something. But then I also think, I think you have your own stores as well. So I’m curious what the mix is like there. Is your store data a lot easier to come by? Is that more usable?
Selma: Yeah, so our brick-and-mortar data and our e-commerce data, because it’s our source system, is much easier to access and I do all the ET detailing into and making it available for our analytics use cases, our analytics users and BI users. There are some partners, like some partner data that’s much easier to access than other partner data. But I would say that our IT department, our data engineering teams, along with our, the partnership with the business, it has done a really good job of making that data available within our data system, and I think that’s another really good, great thing. So you mentioned that we’re centralized. Really our analytics model is federated in that we have a data engineering team and a BI team that does a global data and analytics team that does all the hard work in making the data available in our data landscape. And this kind of like broader environment, but we also have business teams that are super tuned in with the data. Again, and that’s what I was saying before is that if you are trying to make something more efficient or predict something that the business is already leveraging to make decisions. You’re already kind of like keyed into a big, big opportunity. And we have that at Carter’s. The majority of our teams are leveraging data to make decisions, and they’re really keyed into that data so they can tell us, Oh, the data that you’re using, Partner, Right? The selling partner doesn’t look accurate because I’m pulling it from their portal, and then they help us reconcile and build better processes. So it’s really the partnership between the business and IT that I think is transcends.
Evan: Wow yeah, that’s great. And I, the term partner escaped me. I didn’t know what to call them. There’s like some tension there. They own the data. You own the data, but I love, maybe there’s some element of that, but you even call them partners, so it’s at the very symbiotic relationship.
Selma: Yeah, and it’s definitely – the partnership is led by our business teams, right? Until we, we, we partner with our business teams in Carter’s, and they enable that partnership with our vendors and our selling partners.
Evan: Awesome. So we’ve had a couple of guests on here from other tumor goods companies, and a lot of their effort circles around customer lifetime value. How valuable do we expect this customer to be? And they’re looking, you know, they’ve got long-term products or products that you can buy throughout your lifetime. Your products restrict you a little bit. You know, you outgrow Carter’s before too long and you can have a lot of kids and I recommend folks do. But the customer lifetime value is really, is sort of restricted. Does that change any of the effort of where you point your analytics? Since you don’t have the longevity on the customer?
Selma: Yeah. But I mean, I don’t think it changes the equation too much in that most CPG companies are thinking about like cross sell, upsell, you know, what do you have in your basket? And how can we add to it, essentially, right? So they’re not just looking at, we have a customer, how do we retain that customer? But how do we cross sell to that customer? Upsell to that customer. Another part too is you don’t have to have a baby to buy Carter’s, right? So you’ll send to friends having babies. Carter makes great gifts, things like that. And then that also does come into play when we think about strategy. And again, I don’t think that it changes the equation too much relative some of that to some of the other CPG companies in that we’re thinking about how can we market to different types of customers? How can we evolve our product so that we are meeting more customers where they are? So we are aging up with our customers. We’re not staying in the, you know, baby to toddler age ranges anymore. So there are definitely things that we’re doing to kind of evolve with our customers.
And when we think about kind of analytics, what does that look like? It’s our market basket, our customer demographics. How do we market to them? How does our product evolve? Things like that. And again, I don’t think that even if we were, you know, selling to adults, which I think have a longer lifetime span, right? I don’t think that that necessarily changes, right? Like, I think that they’re doing the same.
Evan: That’s good. So awesome. You have really painted a nice picture of how your team sports different business units, how you collect data, how you prioritize analytic efforts and serve that out. Let’s scrap all that for just a second, and it’s just the Selma show. You’ve dealt with a bunch of different business units and maybe just put on your sort of selfish hat for a second and you get to point your analytic efforts and your team’s efforts at any problem that you want. Whatever you think is interesting and everybody else is aligned from the IT and the BI folks to the actual business. You just get to work on whatever you want. So in that sort of area, what’s a project that you’re –
Selma: And I think this is gonna, this is kinda a trick question because for a lot of your listeners, I think they’re gonna think, Oh my God. Like, what’s the next use case? What’s the next model right now, like for me, there are a couple of things that are really important, not just that are really interesting to me, but I think that are important given the context of where we are, right? So like supply chain issues that’s really important. How do we make sure that our products are arriving on time to our customer regardless of like where they’re buying, whether that’s online or in store, or you buy online, pickup in store, things like that.
The other thing about I think is super interesting to me is how do we, again, evolve our customer, evolve our product to meet our new customers and their shopping behavior that’s changing? So the demographics of our customer base are changing significantly. Now different generations are becoming parents and they’re buying different products and they have different motivators. And they have different criteria when they’re making their purchases, but they’re also shopping in ways that it’s not that we haven’t necessarily seen them before, but I think the mix of the way that they’re shopping, and again, their motivators are definitely different. So that’s interesting to me. Like, how do we market better to our evolving customers and how do we make sure that our products are arriving on time to those customers?
But if it was just my show and I could do whatever I wanted, I would not do those projects because I think that there’s something more interesting or something that is more valuable to an organization than the next machine learning model that you’re gonna build. And that is data literacy and data culture at a. I’m really, really interested in building out, well-structured and well-maintained data domains for the organization and having the business users. Equipped to leverage those data domains to make their business decisions. And so when we were talking about earlier, kind of like how do we, how do you ensure if you build this like really amazing dashboard that people are using it to make decisions? You can really do that by having data. And a strong data culture at an organization. And if you think about the, the companies that are doing it, right, you know, if you’re thinking about kind of like you, Northern Stars those are the companies that like, there’s analytics at their core. And I’m not gonna do any name dropping, but think about it. Like who are the data and analytics leaders? They’re always the companies that have really strong data cultures and have really strong data literacy at their organizations. And that’s what I’m interested in. Again, all of these like machine learning models and all of these use cases, they’re like super interesting to me. I think that they can be game changers at an organization, but not to the extent data literacy and data culture can.
Evan: Wow. That’s a full answer with the caveated very long. I will say Selma, I think after speaking with me for the last 25 minutes, I think you’re in a great position to help drive data culture at Carter’s. You speak very well on it. You seem very excited about it to what you do there, and that data literacy. Sure. Hopefully you’ve inspired a few other folks, but that is all the time we have that’s the last question. Selma, thank you so much for joining us today.
Selma: Thank you for having me. This is so much fun.
Evan: All right. Make sure to like and subscribe to get the next episode. Thank for joining us today on the Mining Your Own Business Podcast.