Mining Your Own Business Podcast

Episode 7 - Building AI at Home Depot with John Carroll & Jon Weininger

Building AI at Home Depot

In this episode, we are joined by not one but two data science and analytics leaders from Home Depot: John Carroll and Jon Weininger.

You’ll hear them discuss their past and present roles at Home Depot as experts on analytics on different teams.

They offer insight into coordinating with different teams across the organization and how their respective backgrounds inform the way that they communicate with both technical and non-technical individuals about data.

You’ll also hear them talk about prioritization when it comes to project management and the ways they’ve learned from and adapted to challenges over their careers.

Finally, they share the “why” behind their work and their role in creating value and impact for the company as a whole.

In this episode you will learn:

  • How the data science and analytics teams have evolved over time, responding to company needs
  • The dynamics of the data science team and how they communicate their ideas with other data teams
  • Strategies and approaches to prioritizing projects in a complex environment
  • Common roadblocks and how to prioritize the problems that need to be solved

Learn more about why we created the Mining Your Own Business podcast.

About This Episode's Participants


John Carroll | Guest

John Carroll is the Senior Manager for MET Analytics and Technology at the Home Depot. He’s an experienced Forecasting and Analytics Manager with strong leadership and problem solving abilities.

John is a proven strategist with a Bachelor of Science of Management from The Georgia Institute of Technology.

Follow John on LinkedIn

Jon Weininger | Guest

Jonathan Weininger is the Manager of the newly formed MET Data Science team at Home Depot. He’s previously worked at the Home Depot in Operations, then in Supply Chain Analytics, and now at the Merchandising Execution Team.

Jon graduated with a Bachelors of Science in Economics from Oglethorpe University.

Follow Jon on LinkedIn


Evan Wimpey | Host

Evan Wimpey is the Director of Analytics Strategy at Elder Research where he works with organizations to transform deficient data into tangible business value that advances their mission.

He is uniquely suited for this challenge by pairing his professional experience in management and economics at high-functioning organizations like the Marine Corps and Goldman Sachs with his technical prowess in data science. His analytics skillset was strengthened while earning his MS in Analytics from the Institute for Advanced Analytics at NC State University.

Evan almost always has a smile on his face, which is at it’s widest when he is helping organizations use data in innovative ways to solve complex problems. He is also, in a strictly technical sense, a “professional” comedian.

Follow Evan on LinkedIn

 

Key Moments From This Episode

01:34 Jon Weininger and his work background
02:31 John Carroll and his work background
05:03 How does Home Depot structure their data analytic capability?
09:12 Communicating technical work to other teams
15:09 What are the roadblocks?
21:17 How to set priorities with stakeholders
25:59 Given an ideal scenario, where would you focus your analytics efforts?

Show Transcript

Evan: Hello and welcome to the Mining Your Business podcast. I’m your host, Evan Wimpey. Today I’m excited to introduce a double feature. We have two guests today, both from the Home Depot, where they work in analytics. John Weininger is a MET analytics data science manager, and John Carroll is a senior manager for MET analytics and Technology. So they’re both John’s, they’re both from the Home Depot. For some disambiguation, I’ll refer to John Weininger as J.W. throughout this show. I hate to go on a rant at the beginning of an episode already, but I think data provenance and disambiguation, sort of entity resolution, is a problem that’s easy to overlook until it becomes a problem, and then it’s a real challenge. So I don’t know if this is best practice, but at least for this show, it’ll be John and J.W. So after that opening rant, J.W., maybe you can take a minute to introduce yourself and your role there at Home Depot.

J.W. Yeah, you bet. I’m J.W. I’m a manager of the relatively newly formed MET data science team where we just celebrated our one-year birthday about two weeks ago. In total, in my evolving career at Home Depot, I’ve worked in three different groups: two years in store operations, two years in supply chain analytics, and now two and a half years in MET, first in the analytics manager role. And then as we continued some of the cool projects that the team had been working on, and some of the things that I got to be a part of and maybe helped to guide and introduce, our leadership had the creative idea of like, how can we keep this thing going? Might it make sense to try out a data science team and see if that can provide value to the business? And we’re a year and two weeks into the experiment, and have some cool stuff that we’ve completed and some cool stuff we’re working on and even cooler stuff we’re planning.

Evan: Awesome. Thanks, J.W. John?

John Carroll: Definitely agree with J.W. We’ve got a lot of cool stuff we’re working on today. My name is John Carroll. I’ve been with Home Depot for a little over ten years. I didn’t really have a huge bent toward analytics when I first started. But the business needs demanded that I get some skill sets in that area. So I did forecasting for all of the hourly payroll at Home Depot for about seven years. And then when I moved over to the MET team, which stands for Merchandising Execution Team, I got to kind of cut my teeth on what I’d say an analytics transformation. So imagine — I like to call it Ultron, but it was our SQL Server that had all these different nodes and there was just no rhyme or reason for the way it was organized. And so about 3000 different things that were just kind of a tangled mess, and that was a huge issue for the organization at the time. And, so that’s what I got to come in and cut my teeth on. And once we fixed it, it opened up all these doors for teams like J.W. and his team to come in and have some huge impacts on the organization. I am not a trained data scientist, but I certainly recognize their value. And, yeah, it’s been a great ride the last three and a half years that I’ve been in this part of the organization to see the transformation of the analytics maturity curve. If you all have discussed those before, it’s a real thing and it takes a lot of time.

Evan: Yeah, certainly. You speak of that curve. We’ve talked about that and we’ve got folks listening that are at any point on that curve, we hope. You mentioned that this team is about a year old. It’s certainly not the only analytics team at Home Depot. Can you talk about sort of how your team fits in with other analytics teams? Is there like an Enterprise Wide or does the MET sort of stay siloed within merchandising? And I’m not sure who’s best equipped to put that puzzle piece together.

John Carroll: So one piece is that we are consumers of data, right? So Home Depot does have a central team that manages the data collection and dissemination. But that’s for, those commonly used, you know, there’s commonly used data datasets, right? Then you’ve got these other teams that are creating data and insights. So we have a healthy mix between data that we’ve created and then also data that’s supported by the enterprise. So, yeah, Home Depot’s big, and we probably, I mean — from a pure analytics team standpoint, I mean, I don’t know, we probably have, at least a thousand dedicated data analyst types that are spread across the different hundreds of different parts of our org.

J.W. Evan, one small clarification is that, although the data science team is one year old and our MET analytics and insights group, which has analytics and data science as the same teams, the analytics capability has been evolving for years. Even formerly, we used to call it more of the reporting team or the BA team. I’ve worked in reporting teams in the past, and that distinction is important. So the way we simplify the complicated world around us maybe is: reporting is displaying information for everyone to keep score and show what happened. Analytics, hopefully, we’re trying to begin to look for outliers or draw attention to specific exceptions that are hopefully associated with actions and ideally followed up with some kind of ongoing measure of, “we found this wrong thing and we showed it to you. Did you do something? How do we know if it worked?” So I think kind of completing the puzzle. And then data science is going to be a little bit more involved in either predicting or looking backward in a more sophisticated way with some possible problems we’ve worked on. And then if you look backward and you can’t find what you’re looking for, maybe you have to have some input in helping to shape or design an experiment to be able to isolate some specific thing that you’re curious about. So I think that’s how our team has evolved. And then just to go on a brief sidetrack here, it’s just if you’re in the analytics group and you’re interested in dabbling in some data science-themed work but feel intimidated by it, we all feel that way. And then even some of the people I look up to admire a lot at Home Depot, one of them was Pat, he was on a podcast talking about still having a little bit of imposters syndrome, and I was like, one of the top three people I look up to the most in data science still sometimes feels that and felt that as he was growing and making stuff happen. And it’s like, that’s very encouraging. So that was kind of my attempt to connect the dots between reporting analytics and data science and then give people some peace of mind of like, “It’s okay to feel like I’m figuring it out as we go.” Everybody is, in every role and every team.

Evan: Yeah, I think that’s good. For the individual contributors with a lot of experience out there, don’t hesitate to convey that. That helps everybody when you can convey some of the things that you question. I think organizationally as well, you know, we’re trying to do these things, but we don’t know if we’re structured or organized the right way. So, it’s useful to hear you talk about the way you sort of split the team. Just, you know, you’ve got descriptive things that you do, you’ve got predictive things, and then hopefully prescriptive things that you can do to institute some change. So, John mentioned earlier that his background is not specifically in data science or analytics. J.W., managing that team, I presume there are some things that you need to convey that are very technical to John and then to other folks as well. But maybe you can just discuss the dynamic between the two of you and how you try to convey the analytics language that you guys are working in.

J.W. Yeah, I’ll attempt to take on a portion of that, and I’m curious about how John thinks about it as well. I would say that each of us has a substantial background in SQL. So even the first department we worked in where we first knew each other was eight or ten years ago, lots of in-depth, complex SQL, lots of common table expressions, lots of different sources, some of them well-documented, some of them not. I think my steps were getting really strong and SQL gave me the opportunity to also learn the different business applications. So you look at the front end and try to figure out the SQL and the back end database and try to make it match the screen that I’m looking at or maybe I search for a specific order or a specific customer or something like that. So I would say that SQL was the first stepping stone, and then someone I admire a lot, Darrell Hicks — can’t overlook the opportunity to give him a shout-out here. A DBA, very logical thinker, incredible problem solver, one of most creative, talented people I’ve ever encountered. He helped me set that foundation up to think logically through the problem and SQL can force it to do that. And it’s almost like — I’m like Python with a thousand different ways to do almost everything. SQL only has a very, very limited, or light footprint of things you can do, but you can still combine those elements to build almost unlimited things. So that was kind of my start down the analytics path. And then of course you do just the basic visualization tools. However, statistics was always one of my favorite topics and I also have an undergraduates in economics. Economic thinking is an approach to thinking, statistics is an approach to generalizing based on a limited sample. So it’s like those building blocks of being strong in math and strong in SQL and then doing forecasting type series forecasting for a while, mostly statistical like exponential smoothing, those elements all built together so that I think I can basically figure out which data science problem-solving approach applies or which view you might consider and when it doesn’t apply. And I think that I bring a lot of value to the team in finding interesting, valuable things for them to work on without blitzing them with, “I’m going to look through every inch of what you wrote in Python.” I’m not. I’m just going to ask, “What’s your baseline? How much better do you think you can do? How long do you think it can take?” And go talk to other people who have solved or look at similar problems and get creative and get some ideas.

Evan: Great. John, do you have a different perspective?

John Carroll: I would say so. So, yeah, I said I don’t have a background in data science. J.W. alluded to it. I have a lot in SQL. So when we’re talking, you know, I do understand enough to know what’s happening. And when I have to dig into the details, I can usually understand about 90% of it. But my role has, you know, as J.W. I picked up the torch this last year, my role has kind of started to get out of the weeds and get into the more strategic and really trying to support the good work that they’re doing, making sure the roadblocks are gone. And then making sure that our leadership understands the value of what our data science teams are working on. It doesn’t mean — I’m not saying that they don’t value it, but sometimes they don’t understand the time it takes to validate a finding. You know, I won’t name names, but I’ve seen, where people look at two data points and then like they assume there’s a correlation and that’s not how you do that. And so it’s like, no, that takes some time, and this is how we’re doing it. And you’re still going to get the same answer. But understand, when you get it from us, we have the statistical rigor behind those answers that give us confidence in the decision you’re about to make. So that’s kind of where I sit, I will take — and J.W. is a great communicator, but we’ll say we have slightly different audiences based on our roles. I have to communicate a little bit differently than he does and especially up the chain when we have people making some fairly sizable and sometimes paradigm shifting decisions.

Evan: Yeah. That’s certainly a very fair point. And I think [that example is] a great way to try to convey things: “two points — I can draw a line and I can extrapolate however far I want from that.” It’s a dangerous thing, but it’s all too easy to do. And then J.W. mentioned starting from a baseline, like, what is the baseline for which we try to improve? I think that’s so often an overlooked question that we just think we can throw a model at this and build something good. What do we expect to improve on? The baseline is great. So, John, you mentioned something in that answer: “try to remove the roadblocks for J.W.” And so, you know, open-ended, I’m happy if you want to talk about any of what those roadblocks may be. But I’m going to do some forecasting of my own. Typically, data scientists like to complain about not being able to get good, clean data quickly and easily. So I don’t know if that is a common roadblock or if there’s anything else that you’d like to touch on there.

John Carroll: Yeah. That’s one of them. Let’s talk about data engineers. If I had to say, when J.W. first started like a year and two weeks ago as a data science manager and started hiring people up, you know, I don’t think that we had in mind that we wanted a data engineer specifically. We knew it was a function that was going to need to be, you know, considered. But we quickly found that one of the impediments to getting a model out there was data engineering. And even some of our former team members spent a lot of time on data engineering and got — I am mostly data engineering when it comes to the mindset, but I look forward to that end result more than I look forward to diving in and trying to do all the things, all the SQL scripting and whatnot. So it becomes a burden and after a while, it can affect satisfaction, you know unless you find an individual who, that’s what they want to do. And so, of course, when we started hiring, we weren’t telling people, you know, data engineering is going to be 50% of your job. Right? Like you have to spend two months fixing this data before you can perfect your Python model. So we did identify that and we had some fun conversations around, you know, how are we going to make that happen? In the meantime, it was like, I’m sorry data scientists, this is going to be something that you have to learn and get good at. Now, I will say, I think everyone whether you’re doing reporting or building those insights tools or you’re data science is supporting a very large volume of stakeholders, I think you have to have some data engineering — don’t call it expertise, but experience, you know, you’ve got to have a mindset that can scale whatever it is, whatever it is you’re doing. But to have a dedicated individual who can float around to the different projects and optimize this or set this up so that when the data scientist moves on to the next project, there’s that dataset ready and waiting. That’s definitely a thing that we desire, and where I’d say we’re starting to make inroads, too. We’ve got a contractor to kind of help with that. I know I went far down deep in that hole, but I know that’s one of the questions I think you were kind of pushing us towards that. So here’s my answer to the data engineer question.

But the other stuff: we have a team structure where a lot of the initiatives, I would say 90% of the initiatives of our organization, the merchandising execution team organization, that 90% of the initiatives that they’re pursuing at some point flow through my team. And so, we don’t have a lot of spare time. But we also have all these other stakeholders with these great ideas. And it’s just going to be so easy and wonderful if we can give you a week or two of time to go and address this. So one of my things is to make sure that our stakeholders are focused on the goal, and that if they want some of our time, they’ve got to help us consider what gets deprioritized. It’s not a, “do all the stuff that we were planning to do, and then this other thing.” We’re full of work, we’re probably at about 110% capacity right now, just being candid, and we’re managing that pretty well. But man, it’s like we get a lot of great, great ideas from our stakeholders and they’re usually right about the opportunities there. It’s just how we assess that value. And that’s again, kind of where I’ll come in to do a lot of blocking and tackling. I know J.W. probably isolates me from a little of that as well. There’s probably a pet project or two that comes up every now and then that they’ve identified, then they’ll bring it to me and it has this huge value and we do deprioritize something. But a lot of that happens where I’m just talking with our stakeholders and saying, “Hey, this team is for those strategic long-term initiatives, and I know you want us to pull some data for you, but that’s not what they’re here to do.” I had to have a lot of those conversations last year when J.W. switched roles with the results they’re getting, they’re seeing the value and kind of aligning their requests to what they’re working on, which makes it a lot easier to say “yes” to some of those things.

Evan: Yeah, it’s a double-edged sword. When you’re successful and you deliver a lot of value, people want more value. They have more requests. And nothing draws a crowd like a crowd. People are asking for analytic support, more people want analytic support. J.W., did you want to comment or do you want to say from your perspective, what it’s like to try to prioritize the things that come into the team?

J.W. Yeah, I can comment on prioritization. You have the traditional “effort” and “impact”, right? You might make your four boxes, it’s a pretty standard concept in business settings, probably with good reason. I’m sure other people are more sophisticated in their approaches. Something else I like to think about is, I started calling it sort of “market timing”. It’s like organizational interest and kind of the likelihood of getting a strong adoption from somebody who can help to focus attention and resources. I won’t get into the super specifics of it, but something could be moderate effort and high impact, like the organizational timing, maybe the interest isn’t there, or maybe they tried something similar or some X number of quarters ago where there’s, you know, maybe they have something really good and we think we have an incredible version of it. Well, let’s let the really good version ride and continue to make an impact and then, “oh look, you exceeded your goals by this amount. We think we can help you exceed them even by a little bit more.”

So then the other aspect, of something I would say very much I’m on the learning curve around, is that when you have more interesting ideas than you have team capacity, you can either get good at rapidly triaging. “Triage” is our sort of phrase. It’s for initially assessing the size of the prize, or doing some napkin math, or cutting some reasonable corners and understanding the assumptions and potential downsides of doing that. Right. So if I can get faster at approximating the size of the prize, it makes it possible for me to consider more things before I pick those two or three things to be really good at. Currently, as someone who likes to go wide and in-depth where I can, I think I’ll just be satisfied with my curiosity going wide. But that helps me to narrow the focus which makes it more enjoyable to be on a team. Okay. Our director, my focus, challenges me to focus like, “let’s aim at one or two or three, one or two things at a time, maybe, you know, per quarter or half.” And those refocusing conversations, some of them have only been 3 to 5 minutes, have dramatically impacted the trajectory of our work. I think they’ve helped us to build very powerful Legos, this one on top of each other. Or, maybe I tried to put four or five once, two by two. So I appreciate that. I think that the senior leaders have gotten good at that and have really empowered their teams and give us that peace of mind and confidence to go after the right things. The other thing is having the humility to know that sometimes we don’t. And then it’s like, “what are our feedback loops so that we can pivot?”

Evan: Yeah, I think that will resonate with a lot of folks who are trying to juggle priorities and who maybe don’t have the leadership that says, “yes, reprioritize with this new initiative, but just add it to the backlog and do everything.” So I think those are really useful thoughts.

I would ask one more question, and I’ll ask the same question to each of you. You talked about some of the MET’s initiatives that come in and how you prioritize them. Let’s just scrap all that. It’s John or J.W. Show and you get to set your initiatives and everybody has full buy-in. And yes, the interesting problem that you want to tackle is the problem that everybody wants to be tackled, and you can put all your resources towards it. So what is that? What is that problem for you? Where do you want to point out the analytic efforts?

John Carroll: So, when we hear about problems and I think I mentioned this before, we have a lot of people who work in the stores and a lot of our most successful projects have been born from problems that our field brings to our attention. There’s a lot of smart people out there who haven’t been as blessed as me to be in this role or as blessed as J.W. to be in his role, but they have a lot of great ideas. So, we have no shortage of ideas from those individuals. And then it’s, how can we bounce off of other stakeholders and try to arrive at the couple that is going to drive the most value? One of the questions we get a lot from our field is, what’s the value of what I’m doing? That’s something we’ve been perpetually interested in doing. I would say probably a little more than dabbled in it and coming up with some great answers. But it’s not enough just to say what you’ve done. I think there’s this notion of like, how can we expound on that? You know, once you get that kind of description of the impact, how then can you start predicting and prescribing, right? How can we get more of that impact or how can we maximize that impact? I think, with information, I find that the devil’s in the details. So anything we can do that will get that impact to a more granular level, and not just it’s the impact from a sales perspective, sometimes it’s from an experience perspective, which is a little bit harder to measure. And then obviously the savings. So it’s very simple business concepts at a high level, like basically taking that balance sheet that we report out on every quarter and just drilling it down to as far down as we can go using that concept, I think, has a lot of opportunities.

Evan: Yeah. Very exciting. Thank you, John. And J.W., I pose the same question to you.

J.W. I’m going to potentially give what will seem like a cliché answer, but it is not. It’s heartfelt. I’ve thought about this a little bit. The vision and strategy stuff is fascinating to me and what Simon Sinek spoke of “Start With Why,” and also “The Infinite Game” are extremely informative. So everyone is listening to hear “why” and make sure I read about what it just causes. Even if you only watch the 3 to 5-minute YouTube video, I consider that as how you think about how you approach your work and everything else and how it all fits together. So the cliché part is, it’s going to seem a little snarky, but I’m very bought into the vision and mission of what MET is doing. You’re asking me if I could set the priority, and I would probably just have the priority that our leadership has set for our organization. That’s convenient for me and it’s motivating, even when things go well and when things don’t. I hate to give such a cliché answer, but if I could set the priority, I would continue to do what we’re doing because it makes a lot of sense and it seems to be working. I would also just want to emphasize that some of the people who started Home Depot say a phrase that I think about often, but don’t say as often as I would like, which is: “When we take care of our associates, that’ll take care of the customers, and everything else takes care of itself.” I think I would challenge John Carroll and myself and our team and everyone in our immediate organization and everyone in this building: for the projects that you’re working on, are you looking at it through that lens multiple times throughout the project? You know, if you’re building an app or you’re looking at it throughout the whole lifecycle of an app, if you’re making that answer, if you’re changing the process, is that always top of mind and are you working backward from that result? And then just to emphasize the commercial: Do you have measurements in place to help you approximate how much you are taking care of the associates? Do you have feedback loops? Would they agree? If one of them just jumped on, if we went to a random store, asking a random person like; “Here’s what we’re working on. Do you see how this connects to making your work more revealing?” Do they see that connection?

Evan: Wow. Yeah, that’s great. It doesn’t seem too cliché at all. It seems very sincere and it seems well aligned with, John, the answer that you mentioned. Now you’ve got the associates and the things that are going to empower them. I think that’s great. I think those are two encouraging answers. And — are you guys actively hiring right now? I don’t want to encourage people to apply if you’re not actively hiring.

J.W. Our specific teams are not actively hiring. However, the broader data science community within Home Depot has a few roles open, and if you go on Google and just type in… I think John will know. I’m embarrassed to slightly go blank on this. I think it’s called “Career Depot” or just typing “Home Depot jobs” or typing in “data science”. You will find it. Sometimes there’s a critique that the descriptions are generic. That might occasionally be true, but that is usually not true. The descriptions are relatively specific to the roles. I know we get to give extensive input when we post our new jobs, so there are likely roles we’ll bring in at Home Depot. If any of this resonates, if you want to explore further like the people who don’t know it yet. You can be bold and ask. The worst thing that you’ll hear is no or no response. But you don’t know until you ask.

Evan: Awesome. Well, thank you so much, gentlemen, for coming to the show. If you are in the field of data science and you are encouraged by these answers and want to be a part of what Home Depot is doing, then certainly use your Google and find some job that fits you. If you’re out there building a data science team or capabilities, take some of these answers and help them shape the way that you build your team. And hopefully, you can be as empowering, too, to the employees that work there and to the vision that your leaders have. John and J.W., thanks so much. It’s been a pleasure to have you all on the show today.

J.W. Yeah. Thanks.

John Carroll: Thanks for having us.

Evan: All right. Make sure to like and subscribe and tune in next time on the Mining Your Own Business podcast.