Resources
Feb 23, 2023 |
59 min
|
659 views

Data Science Hangout | Namrata Shetty-Anderson, UPS Store | Professional development for data teams

We were joined by Namrata Shetty-Anderson, Advanced Analytics Manager at The UPS Store.

Right from the beginning of the conversation with Namrata, we dove into the art of data storytelling (7:20)

...In my mind, data science is as much a science as it is an art.

To be able to tell your story, you could write hundreds of lines of code and you could build amazing models, but at the end of the day if you can't translate it to a non-data person as a data story, all that is a waste. You're not adding any value.

Businesses don't care about your code. My boss doesn't care about my hundreds of lines of code. He wants value.

So I call it the art of data stories.

I actually took a course by Kevin Hartman from Google, which talks about this. One thing I learned from that course is that you want to start with a solution for the problem in a tool agnostic environment.

Don't start with your tool when the problem is posed in front of you because what happens is the tool is going to limit your ability to tell your story. If I say I'm going to use Power BI, then I'm going to be limited to the visuals that are available in Power BI.

So what I started doing, literally, was use a blank page and a pen.

What do I want my outcome to look like?

Once I know what that story is going to look like on a blank sheet of paper, then I decide what tool I'm going to use.

Now the tool is going to depend on a few things: How quickly do I need to turn this around? How complex is this task? How good am I with this tool?

Those are the things you need to look at. People say, Python's better than R, R is better than Python. No, Power BI is the way to go. Tableau is the way to go.

No, everybody's right and everybody is wrong. Your tool of choice will depend on the problem you're trying to solve.

Don't fall in love with anything you're doing. Today you're doing it one way, you're going to find a better way to do it as you keep learning.

*The course mentioned above was Business Analytics, Communicating With Data by Kevin Hartman: https://www.coursera.org/learn/intro-business-analytics &/or https://elvtr.com/course/storytelling-with-data
_____

Namrata also shared thoughts on approaches to professional development and skill building - (22:58)

Managing a data science team is different from managing a team that's operations driven, product driven, or even a software engineering team.

It's different because problems that come to a data science team are not always well defined and the solution is not cookie cutter. There's no right answer, there's no wrong answer.

You need a research oriented approach to what we do and a drive to keep learning.

So here's what we have been doing.

1. My whole team has the option of picking their learning platform.

Some pick Coursera, some pick a virtual summit or buy a book, whatever. You tell me what your best learning approach is and you can have it.

2. The team is also encouraged to block off time on your calendar.

X hours a week or X hours a month, whatever you like to focus on learning and put it on your calendar. No one's going to bother you. Do we do it? Sometimes. They'll block it off and sometimes won't do it. It happens to the best of us.

3. We have a week long group camp

Every year we have kind of a week long group camp, as I call it, where we bring in a certified instructor (we've been focusing on R for the last two years at least) and just work on little projects with the team. For this week, the team is not bothered by anyone just to focus on learning. (more on how they block that time at 31:51)

4. We started having monthly sessions.

We started with best practice sessions where one of the team members would say, hey look, this is a project I worked on and here's how I approach the problem. This helps someone that's new on the team, someone that's learning to get some exposure to it.

You can learn everything online, you can learn from instructors but then when you have to apply it in the real world, it's not a one to one translation. Real world data is not normally distributed. Real world data has noise in it.

When we maxed out our best practice sessions, we started doing a micro assignment. My team member Liz puts together micro assignments for the team with real work data - like online reviews, customer data, sales data - and perform these operations in R.

5. We put together a central repository

Because my team does so many things, we have a repository where we have reference guides, best practices, and a lot of documentation. If someone new joins my team, I tell them you're going to spend your first week just reading everything that's on that repository.

Documentation, process documentation plays a very important role and it allows us to cover for each other. We want to rely on processes that make our people's lives easier.

____

To join future data science hangouts, add to your calendar here: pos.it/dsh (All are welcome! We'd love to see you!)

image: thumbnail.jpg

Transcript#

This transcript was generated automatically and may contain errors.

Happy Thursday, everybody. Welcome to the Data Science Hangout. I hope everybody's having a great week. If it is your first time joining us today, it's very nice to meet you. I'm Rachael. I lead our pro customer community here at Posit. This is an open space to chat about data science leadership, questions you're facing, and getting to hear about what's going on in the world of data across all different industries and companies.

And so every week, we feature a different data science leader as my co host for the day, to help lead our discussion and answer questions from you all. And together, we're all dedicated to making this a welcoming environment for everybody. So we love hearing from everyone, no matter your level of experience or area of work. It is totally okay to just listen in to Hangouts too. But we do want to hear from you as well. So there's three ways you can ask questions, or also provide your own perspective. So you can jump in by raising your hand on Zoom. You could put questions into the Zoom chat. And feel free to just put a little star next to it if you want me to read it. Or else I could call on you to introduce yourself and add some context too. We also have a Slido link where you can ask questions anonymously. And Hannah will share that here in the chat in just a second.

We do share the recordings of each session up to the Posit YouTube. So you can always go back and rewatch or share with a friend too. There's a LinkedIn group as well, which we'll share in the chat. We've been a little quiet in there lately. But if it's helpful for connecting with people after the fact, I encourage you to join that group too.

With that, thank you so much, Namrata, for joining us as the co-host today. Namrata Shetty-Anderson is Advanced Analytics Manager at the UPS Store. And I'd love to have you maybe start by introducing yourself and sharing a little bit about your work. Maybe something you like to do outside of work as well.

Sure. Hi, everyone. Really happy to be here. I was really happy when Rachel reached out. This is my first data science Hangout as a featured leader. I've attended a few as one of the attendees, and I think that's where Rachel found me. As she pointed out, I'm Namrata Shetty-Anderson. I know my name is difficult to pronounce, so you can call me Ratty. That's what everyone calls me. I'm cool with it. I'm an Advanced Analytics Manager at the UPS Store. I've been with the company for about five years now. Actually, almost six years now. And I lead a team of six analysts, and together we do everything data. We're the data nerds of the organization. And anything that has anything to do with data goes through us, whether it's bringing data on premise, whether it's building reports, dashboards, predictive analytics, machine learning, wrangling, cleaning, everything about data for this organization goes through us.

In my free time, I'm boring. So in my free time, I'm either playing with data. My husband has a PS5. He plays, you know, Call of Duty or Last of Us or what have you. And I'm playing with random data sets online, trying to learn new things. Or I'm with my dogs. I have two big dogs and a cat. And they take up most of my time. So I'm either with my dogs and my husband and my cat or I'm with data. So that's pretty much all about me.

I love it. Well, thank you so much for joining us. And it is always really fun to have someone who joined us in attendee in the past be the feature leader for the week too. So excited to have you here. I'd love to, it sounds like so many different data projects go through your team. I'd love to understand a little bit about maybe like an example project that you work on.

Sure, I'll share what I can. So like I said, my team handles everything, you know, from so we have a little bit of, you know, what we call ourselves a shadow IT or data engineering, we do a little bit of that, we have a little bit of descriptive, and, you know, just building reports, that kind of stuff. And then we have a set of us that does, you know, predictive analytics, forecasting, machine learning, that kind of stuff. So the way projects roll into my team is, say there's a company project that's ongoing, they're going to come to us for probably KPI reporting for that project. So we're going to put, you know, maybe do some wrangling in R and put together an ad hoc report. Or we also have a Power BI service, where we publish dashboards for the users. So that's where all our descriptive stuff goes.

And then we have, you know, predictive models for everything we've got, where the UPS store network. So if you know about as we have, you know, over 5100 locations across the country, so my team is in charge of forecasting how much revenue those stores are going to generate how much volume is going to go through these stores, what is the customer traffic going to look like, assisting with staffing for the stores, that kind of stuff. So we build models to assist with those things as well.

The art of data storytelling

Thank you. I know, and I was looking at the bio that you sent in, you specifically talked about, like, you get excited to tell stories with data. And I like the way that you framed that. I'm wondering, do you have any tips for us on how to tell better stories with data?

Sure. So, um, this is something that, you know, I've actually struggled with in the past, I'll say, this is something that I learned was an area of improvement for me. So over the last few years, I've really focused on this area. So definitely, I can tell you what I know. In my mind, data science is as much a science as it is an art, you know, to be able to tell your story, you could write hundreds of lines of code, and you could build amazing models. But at the end of the day, if you can't translate it to a non data person as a data story, all that is a waste, right? You're not adding any value. Businesses don't care about your code, right? My boss doesn't care about my hundreds of lines of code, he wants value.

In my mind, data science is as much a science as it is an art, you know, to be able to tell your story, you could write hundreds of lines of code, and you could build amazing models. But at the end of the day, if you can't translate it to a non data person as a data story, all that is a waste, right? You're not adding any value.

I call it the art of data stories, right? It's telling your data story, that's an art. So one thing I started doing, and I actually I took a course by Kevin Hartman from Google, which talks about, you know, how data storytelling is an art. And one thing I learned from that course is, you want to start with a solution for the problem in a tool agnostic environment, you know, don't start with your tool when the problem is posed in front of you. Because what happens is the tool is going to limit your ability to tell your story, right? If I say I'm going to use Power BI, then I'm going to be limited to the visuals that are available in Power BI, right? So what I started doing, literally, blank page, and a pen, what do I want my outcome to look like? And once I know what that story is going to look like on a blank sheet of paper, then I decide what tool I'm going to use.

Now the tool is going to depend on how quickly do I need to turn this around? How complex is this task? And how good am I with this tool? Right, so those are the things you need to look at. So everyone's, you know, oh, Python's better than R, R is better than Python. No, Power BI is the way to go. Tableau is the way to go. No, everybody's right, and everybody's wrong. Your tool of choice will depend on the problem you're trying to solve. So that is what I'm going to say.

I really love that perspective. I see a lot of comments in the chat, too. I've thought about that before, too, even when you're making presentations, like sometimes if you jump straight into the slides, you become attached to them. You don't want to give them up, even if it wasn't the right way to communicate it. Yeah, and that's the thing. Don't fall in love with anything you're doing. Don't fall in love with it, because that's what the data world is. Today you're doing it one way. You're going to find a better way to do it as you keep learning. There's always a better way.

Priyanka, I see you had asked in the chat about that course, and I was trying to find it at the same time as well. Namrata, do you know what the name of that was? Yes, just give me one second to look for it. It was Business Analytics Communicating with Data by Kevin Hartman. I did it on Coursera, but it might be offered elsewhere, too, so I wouldn't limit myself to Coursera. Okay, cool. Thank you. Marlene, I see you have your hand raised. Do you want to jump in?

Yeah, hi. Good morning or afternoon. I don't know where everybody's at, so it could be evening if you're eight hours ahead. I apologize. Hello to whatever day you're in. Namrata, thank you. I mean, I'm already digging all the things you're saying. Much love to just kind of throwing the box away and starting with a new box instead of thinking outside the box, in the box, etc.

So my question is kind of targeted, and I don't want you to feel like you're giving up trade secrets because that's not what I'm probing for. So my mom used to work for UPS in loss prevention for many moons. She was there for, I think, 28 odd years. Wow. It's really cool, though, because she was kind of a pioneer. I have to say that I think it's one of the qualities that I hope I inherited from her, where she's a bit of a sleuth and wants to provide solutions to problems.

So one thing she did, though, this was back on pen and paper, was she used to do credit card fraud tracking. So she would take known credit card fraud addresses is what she called it, excuse me, where some, she's working the downtown LA hub here in Los Angeles, and she used to say, okay, there's this one address in the middle of like a not so good neighborhood, and there are five credit cards that are all sending stuff to the same address. And she would do this by hand. And then eventually, when I was very early on in my data analytics career, I built her a little database so she could easily look stuff up and see duplicates for a single address, things like that. And it wound up saving like millions of dollars. I didn't realize until at the end of the year, one year, her boss came and said, do you realize we got a letter from Disney that said that you've saved them $5 million this year because of all like people would send, for example, tickets to Disneyland in boxes to these addresses bought on stolen credit cards or fraudulent credit cards.

And for those of you familiar with Southern California, we have Knott's Berry Farm. Around October, we have Knott's Scary Farm, which is a really big event for a theme park. Also hundreds of thousands of dollars, she would save them from fraudulent tickets being sold and purchased on credit cards. So what I'm getting to, I'm sorry, it's a very long backstory. I told the same way too. It's like an epic backstory and then a punchline.

How has maybe the analytics for that, have you started to see that maybe make their way to your team where you can maybe on a national scale or even a global scale kind of understand like the fraud part of it and or, because it's easy to do kind of just like sales profit loss that way. But I found credit card fraud really cool in that respect because it's just such a little niche pocket of saved revenue, we'll say, or loss overhead prevention, right? So anyway, that's kind of my question.

Yes, that sounds very cool. That sounds like a very cool project you worked on and your mom worked on. So the UPS store is a subsidiary of UPS. So I'm not as involved on the UPS side of things as I am on the store side of things. Although I'm learning, I'm learning, I'm starting to make that transition. But yes, fraud detection is one project that we actually worked on last year, but it was a different kind of fraud for sensitivity reasons. I can't share the details of that project, but we did have fraudulent behavior happening in the stores and we were able to use our point of sale database to actually help identify, not pinpoint the fraud activity, but identify the likelihood that, hey, this seems like something we want to go and question.

And I'm not even exaggerating. So we started last year in April is where our finance team came to my group and said, hey, look, we're seeing this behavior, what can we do? So we built a dashboard that does, it gives them all this information. You could see on an outlier graph, the fraud activity for April and March was like a huge spike. And there were spikes even like rolling 12 months before that. So it's been happening for a while. Once we addressed it with this report, two months later, up until now, it's like a straight line. So whatever action was driven by that data helped us kind of mistake proof the system such that we don't allow that kind of fraud anymore. We kind of got ahead of it. So I think that was one of the cool projects we worked on for sure.

That's so rad. Thanks for sharing and putting up with my very long backstory. Oh no, I loved it. Thank you for sharing. Yeah. I absolutely love this. Love the stories. I see Andy just said in the chat too. I love the learning, but truth be told, I'm here for the stories. So I see Caitlin had asked a question earlier and I'm going to read it for you because I know you're in a cafe. But the question was, what did your journey to leadership look like?

Oh, that's a cool one. All right. So I, as much as I may not seem like one, I am not a very people oriented person. I happen to be an introvert and the only reason I'm able to talk to you guys so well is because you're on a screen and not in the same room as me.

so it helps. So leadership was not something that was on my radar that I didn't think I was, you know, good for. But I started out as, when I joined the UPS store, I started in a, you know, low leadership level position where I had two analysts reporting to me. And I was part of the same team I am today. I think the key factor towards me moving to the leadership role was the fact that I can communicate my data story, right? In the data world, if you want to make it to a leadership role, even if you don't want to, and if you somehow make it there, the driving factor is going to be your ability to tell stories with your data. And two is to build a level of confidence in the organization, you know, because let's face it, you're going to come up with amazing data stories. And if they don't, you know, tell a story that people want to hear, it's going to be hard to convince them to, you know, still do what the data is telling them to do. If it's counterintuitive, right? Because a lot of businesses have, you know, years and years of experience, your subject matter experts are going to use gut feeling or what I like to call just, you know, their intuition, which is right sometimes, but sometimes it's not.

So I think a big factor, at least in my success in my current role was the fact that I was able to build better confidence in our data. And the way I did that was, the first thing we did was we have, we had data sitting in so many different databases, Excel files, what have you, we started working towards one source of the truth, right? We're still not there, you know, five years in, I'm telling you, we're still not there. And it's a long, long journey there. But at least we're working towards improving your data quality. Second was, I'll give you an example. So we have a, we have a model that forecasts storewide sales for the month. That helps us understand how much revenue we're going to generate, that helps us understand how much we can spend for projects, how we can manage our expenses for the month, right? It all trickles down. So we used to send this forecast to the accounting group to let them know, hey, we're going to make plan or not make plan. So you need to cut back on expenses, or hey, maybe this is a good month to have some additional expenses. Before, five years ago, the way it was done was, if we had a forecast, we would always tone it down. Because we were afraid what if the model's wrong, right? And our model wasn't great. Over the course of time, we built a model that was so strong and robust that eventually we just let the model. Today, I don't even tell people. It's just there. It's sitting there. People can go look at it and be like, oh, okay, we're going to make plan or we're not, right? So it was a journey to get to a point where people have confidence in the data.

So if you're one of those people who can drive, you know, improvements, who can build confidence and share your data stories and communicate it to a non-data environment, to someone who doesn't understand anything about data, then you're on the path for leadership is how I'll say it.

So if you're one of those people who can drive, you know, improvements, who can build confidence and share your data stories and communicate it to a non-data environment, to someone who doesn't understand anything about data, then you're on the path for leadership is how I'll say it.

Of course, the other aspect of it is managing people, which I have learned through experience as, you know, a type A who's like, you know, striving to get everything perfect. I do struggle with, you know, balancing empathy with accountability. But I think experience teaches you that as long as you're open to learning and realizing that not, you know, you have to change your ways, you can't manage everyone the same way. So the path to managing people is you just learn with experience is how I'll say it. I hope that answers the question. I know it was a very long answer.

Professional development for small data teams

That's a really great advice. Thank you. Alan, I see you had asked a question too. Do you want to jump in next? Yeah, I sure will. Hi. So much of this really resonates thinking about what overlap between what your team No Matter is doing and what my team is doing. And with a team that's as small like yours, but has really, really broad, like sets of responsibilities. I wonder what your approach is with the team thinking about professional development and skill building when it's really hard to say like, you know, your work is going to be, your work is going to look like data engineering or your work is going to look like data science, balancing everybody's expectations and career goals in a small team like that with lots of responsibilities has got to be really hard. And so I wonder what that looks like for you.

That's a great question. You know, I've been in this leader position for two years now, and I am so proud of how far this team has come. I have to say, you know, I can't take all the credit and I won't take any credit. But I am very glad that I have a very good team. I have a very reliable team. So I'm going to say that they're amazing. They're awesome. I love them. So part of it is because they're awesome, right? The second half of it is I've been trying to like, focus on so look, being in data science, managing a data science team is different than managing, you know, a team that's operations driven or a team that's product driven, or even a team that's a software engineer team, right? It's different because problems that come to our data science team are not always well defined. And the solution is not a cookie cutter. There's no right answer. There's no wrong answer, right? You have the research oriented approach to what we do. So what you need is that drive to keep learning. And not everybody has that drive. So that is something that we've been kind of working on as a team. So here's what we do. We have been doing.

We have, my whole team has, you know, the option of picking their learning platform. Some will pick Coursera, some are going to pick, you know, I want to attend a virtual summit by Power BI, or I want to buy this book, whatever. You tell me what you what your best learning approach is. And you can have it, right? We'll make sure so and, and the team is also encouraged to spend like block off time on your calendar, you know, x hours a week, or x hours a month, whatever you like, to focus on learning, and put it on your calendar, no one's gonna bother you. Do we do it? Sometimes, they'll block it off, and then they won't do it, it happens to the best of us. The other thing I started doing is every year, around the middle of the year, we have kind of a week long boot camp, as I call it, where we bring in a certified, so we've been doing R, I've been focusing, my team has been focusing on R for the last two years, at least. So what we had as a certified instructor come in, and just work on, you know, little projects with the team. So we had like a week long boot camp, or the team was not bothered by anyone, just to focus on learning.

And then the third thing, and I think this is where it kind of closes the gap, is you can learn everything online, you can learn from instructors, but then when you have to apply it in the real world, it's not, you know, a one to one translation, real world data is not normally distributed, real world data has noise in it, you know, it doesn't work. So what we started doing is we started having monthly sessions. One, we started out with best practice sessions, where one of the team members would say, hey, look, this is a project I worked on, and here's how I approach the problem. So this helps someone that's new on the team, someone that's learning that software, to kind of get some exposure to it. The other thing we started doing more recently, is we said, you know what, we're kind of maxing out on our best practice sessions, we have nothing to share at some point. So we started doing a micro assignment. So I have a, I have a very, I don't know if she's on the call today, but she was going to join. My team member Liz, she's the supervisor of the group, or part of the group, she puts together micro assignments for the team with real work data, like, hey, look at these online reviews, look at this customer data, or look at our sales data. And perform these operations in R. Because this year, our last year, our goal was R. And the team would do those assignments. A month later, when she gives them a new assignment, we meet to discuss how everyone approached their assignment. And you learn a lot. You know, you learn more when you make mistakes, than when you successfully complete an assignment, in my mind. So I think that helped the team kind of, you know, get their feet wet, if you will, with R. Two years ago, the same was Power BI, right? So two years ago, our goal was Power BI. Now, we switched to R. We'll pick something for this year, and that's how we've been doing it for the team.

I hope that answers your question, Alan. It does. It's great. Thank you. I love that idea of micro assignments, and really, really like the emphasis on learning, and kind of letting people drive a path that they're interested in. If we come back around to it, it'd be great to hear more about how you manage expectations that people have of, well, I'm learning this stuff, but I'm not getting to use it. Yeah. So maybe I'll plant that seed. I don't want to keep going. I want to hear what other folks are interested in talking about. But thank you.

I will add one other thing, Alan. Actually, two other things. The other thing that, you know, because my team does so many things, we have predictive, we have descriptive, we have data engineering, like a little bit of data engineering, not as much. We also put together like a central kind of repository, where we have reference guides, best practices, a lot of documentation. To the point where today, someone new joins my team, all I'm going to tell them is, hey, look, you're going to spend your first week just reading everything that's on that repository, and then ask questions. So I think documentation, process documentation plays a very important role, and allows us to kind of cover for each other, right? Because we want to rely on processes that make our people's lives easier, as opposed to saying, oh, hey, Matt's out of office today, so this work's not going to get done.

So that's one. And then setting expectations is, you just have to realize that you may learn something, and you may not get a chance to apply it in the real world, because organizations are not always ready for, you know, machine learning. They're not ready. They're not there yet to use these fun, cool things and data. You still have to clean the data. You have to do the groundwork. You have to do the non-fun aspects to get to the fun aspects. But don't wait for someone to come ask you to apply it. You know where the data is. You know what you've learned. If you really want to apply it, go apply it. And if you can build something that's useful and adds value, hey, I'll make sure the organization uses it. So that's how I approach that. That's great. I love it. Thank you. Yeah. Thank you.

Setting up team documentation and onboarding

Andy, I see you have your hand raised. You want to jump in next? Sure. Thanks. Yeah. Hi. My question is, so you just mentioned that you have sort of a landing page for folks that are onboarding that come in new. And our group where I work was talking about putting together a Teams channel. I'm wondering if you have some tips on, you know, just some of the must-haves for setting up a space like that for your team.

Sure. So what we use, so our organization uses Office 365 for most things. So, you know, we have, you know, our Teams group. We don't have a channel yet because we're a small team. We don't really need a channel. But we have a SharePoint. SharePoint is where all my documentation sits. But it doesn't have to be SharePoint, you know, as long as it's a collaborative platform where everyone can access it. So I don't know if you use Google Drive. Start with the simple platforms. You know, we can get to a point where we are, you know, able to use Azure DevOps or Azure Data Lake or Git or what have you. But I started with something very simple. It's Microsoft Word documents posted on SharePoint. Eventually, today that folder looks beautiful. It's got subfolders. It's got everything. But when I started, it was one folder with the Word document in it.

That's perfect. Yeah, we're also using Microsoft Teams and SharePoint. So perfect. Yeah, SharePoint is amazing. I think that's where we're starting. Thanks.

Getting leadership buy-in for team learning time

Sure. So as you were explaining, like those micro-assignments and monthly sessions, and you mentioned that week-long boot camp, I realized Catherine and I had the same question at the same time Catherine and I were slacking. But did you have to convince your leadership that the team should be given a week for learning and to not be bothered by other things?

Well, not really. Again, I consider myself very fortunate because I've had amazing leaders and an amazing team. So I've had an environment where I've been able to succeed. And I realized that not everyone has that privilege.

But there have been, so when it comes to the business has to go on, right, during that week. So the way I've been doing it is I set expectations weeks ahead. So today, the way people come to us with data requests, again, I use SharePoint for that, is I build up a form where people can go and submit their requests with all their requirements. Of course, sometimes people don't know what they want. So that's where the data person has to tell them, hey, this is what the data looks like. But I have a platform where they can go submit these requests. So maybe three weeks ahead of the training, I just set expectations with all the managers, letting them know, hey, we're going to be in training that week. So the turnaround time for these requests is not going to be as quick. So if you have anything that you can plan ahead for, plan ahead. And if an emergency situation arises that week, then I would be the one that bows out and takes care of it. Thankfully, you know, I do have the technique, not I realize not all leaders have the technical

expertise to do what their teams do. But because I've kind of come from where the team is today, um, I can manage most emergencies. So I just I was the one on call. And then I had one of my supervisors on call with me in the event, we had to step out. Two years ago, I did have to step out a lot. This year, last year, I did not have to step out at all. So it just depends on what's happening at work that week. But I made sure that the analysts on the team were not disturbed.

That's great. Thank you. Sure. Eric, I see you asked a question. You want to jump in?

Yeah, great. Great hearing your perspective on these really appreciate it. And I really agree with, you know, the emphasis you said on helping teams improve the process of not just doing the technical work, but helping others on that journey, especially with documentation, internal resources. But what I've seen is sometimes leadership up in the chain, so to speak, don't really see the value of that side of the work, as opposed to just getting that technical deadline done or getting that, you know, that piece. I don't know, I'm curious what your take is on getting leadership to recognize that that side of it of helping people, as Libby said, the glue work, I really like that is just as important as everything else. Just curious your thoughts on that.

So you're right. You're right. You're absolutely right. For the most part, people in our roles are working day to day, just putting fires out. And prioritization on the long term is not always an option, right? The way I did it is, I just did it. I didn't go asking for permission. I didn't say, Hey, look, we should do this. I just did it. And when I did it, now I can show the results. I can be like, Hey, look, when my team members are out, the business can still continue because we have tools in place to help to support the team. When you give me someone new, my team, that person can learn quickly because of this tool that we've put together. So the best way to convince leadership is just do it and then show them the results. Because to play the devil's advocate in the leadership's defense, they're not going to prioritize something if they don't see that it can add value, right? It's all about value. And I feel like in my position, I'm a mid-level manager. I have some level of freedom to make some changes, to change the rules. So I just do it. I don't go seeking permission. I do it. And then if they like it, great. If they don't, I'm fully prepared. If it gets rejected, it's going to get rejected. I can't tell you how many slides I've put together with amazing ideas. And maybe a couple have been selected, right? So you have to kind of balance that out. You just have to do it, be prepared for rejection, but do it, show the results, and that's the best way to drive change.

So the best way to convince leadership is just do it and then show them the results.

The value of subtracting

And one other thing I would also add there, Eric, I definitely recommend this book. It's called Subtract, especially if any of you are on the path to that leadership path in the data science world. I would highly recommend the book Subtract because it's not always about what you need to do. It's also about what you need to stop, right? There was a time my team was building reports week after week, sending them out. We're spending all this time, and we didn't know what those reports were being used for. And again, I'm lucky because I have good leadership. This was a recommendation. This is something I learned from our CEO. She recommended this book to me. This is what she told me. She's like, look, don't send those reports. Just stop and see how many people come knocking at your door for that report. And that's exactly what I did to the point where now I have a policy in my team where if we think something is not being used the way it should be, we put it on what we call a 90-day pause list. We don't tell anybody. We put it on a 90-day pause and see how many people come banging on our door. Sometimes they do. We put it back up. We're like, okay, sorry, we were just testing. And sometimes they don't. And after 90 days, I just sunset it. We're not doing this anymore. And I can't tell you how much time that has freed up for us to actually focus on things that matter. So subtract is just as important as add.

So that's something I would add there, Eric, for sure. Prioritization.

Absolutely. Love it. I'm on that journey as we speak, seeing some legacy processes or capabilities that, frankly, have stagnated. And why do we keep maintaining it when maybe we just focus somewhere else? That's excellent advice. I'm going to check that book out. Thank you so much. Yeah, sure. Thank you. I'm saving that to go buy this right after too.

Libby, I see you asked a question. You want to jump in?

Yeah, I asked a couple, actually. My first question was... Start with whichever one you prefer. Yeah, my first question was just how did you get into data, which might be a longer answer. But the other one, I was also really interested in how you handle failure, both in yourself and in your teammates. You talked a little bit about that, though, right? Just do it. And a small percentage of stuff is going to stick, right?

Yeah. So I'll answer the first one later. I'll answer the second one because it's a shorter answer. I read this somewhere. So this is how I look at failure. Personally, if I fail, I'm very hard on myself. And my husband will attest to that. He's had to talk me down a lot. So as put together, as I come across, I can be a total mess sometimes. I'll make one tiny mistake, and I will harp on it all day. Like, oh, my God, how could I do that?

If I do that to myself, you can imagine how I do that to my team sometimes. They might miss something, miss a deadline here and there. And I can get a little naggy about it. But because I know that I do that, I've been consciously working on, okay, mistake has happened, failure has occurred. Focus on what is the impact, right? At the end of the day, we're not saving lives, right? If I make a mistake, nobody died, right? Remember the Tesla accident, but nobody died. So it's fine. No, I'm just joking. But that's it, right? You have to remind yourself every day, at the end of the day, you're not saving lives. What is the impact of this failure? And if the impact of the failure isn't significant, then the failure is actually a good thing, because you learn from it. So I love to learn, and I have to remind myself every time I fail that this is a learning opportunity. It's okay. Over the course of time, and I hope my team will attest to this, I also start, you know, learn to do the same with them. Two years ago, when I became manager, I was picking on every little thing, right? Today, someone forgets to send something out, and my response is, no worries, it's okay. And I genuinely feel that. It took me a lot of conscious mindfulness to get to this point. So unfortunately, there's no secret to it. It's just you have to be conscious and mindful about it, and realize that the impact of it is not as significant as you make out to be in your head, and then failure just seems like a learning opportunity. I hope that answers your question.

No, it's great. Thank you so much. And then how I got into the data world.

How Namrata got into data

All right. I'm an industrial engineer. I have a bachelor's in production engineering. I had one statistics course in my master's, and one multivariate course. So no, my schooling did not lead me into the data world. I was working in logistics as a research assistant, looking for jobs when I landed this one with UPS. So it wasn't a planned job. I was an industrial engineer working in continuous improvement, doing a little bit of analytics, but not as much as I'm doing today.

I ended up in this role. I was an immigrant at the time, and I had 15 days on my visa to find a job when I landed this job. So I was in Georgia, and I drove all the way to San Diego for this job, because I had 15 days, or I had to leave the country. So I got, there was, luck played a big factor in my entrance to the data world, but once I was here, I was like, I'm not going back. I, when I joined this role, I knew how to use Power BI. I knew a little bit about Tableau. R, I had only used for school projects, and Python was, I had no idea. Okay.

I learned on the job. I focused on learning. I realized how fun the data world can be. I had, I was very fortunate to have a peer who was very, very helpful in my learning curve, being like super accelerated, and here I am today. So honestly, I wish I could tell you, you know, that I grew up dreaming of being a data scientist, but I didn't. I really did. I just ended up here, and when I was here, I was like, whoa, this is amazing. I'm going to stay.

Luck played a big factor in my entrance to the data world, but once I was here, I was like, I'm not going back.

I don't think anybody wants to be a data scientist when they're a kid. Yeah, everybody, we're taught to hate math. I mean, but I'll tell you one thing, I loved math my whole life, so that's one of the reasons engineering was the path for me, because math was my favorite subject in school. If I got a 99 out of 100 in math, I would be crying, going back to me whole, you know, being hard on myself, but I guess my love for math somehow led me here, but it was not planned at all.

Now I'm starting to think, like, what did we all want to be when we were little, and how it, like, got to what we are doing now? I was just thinking, Marlene, when you say, like, you wanted to, like, or you were talking about, like, being a sleuth and finding things in the data, I can remember, like, wanting to be a detective, and, like, it's kind of the same thing with data, too. I wanted to be a vet when I was a kid, because I love animals, but when I was, like, 10, I realized that every time I see blood, I pass out, so that was a no, but, yeah, I never thought of the data world until I really got into it.

Managing historical metadata and documentation

I see there was an anonymous question a bit earlier, and it was, how do you handle managing the hidden metadata that is based more in historical knowledge versus well-documented?

Yeah, we've struggled with that a lot, you know, because we didn't have a lot of documentation. We still don't have documentation for a lot of historical data, but what I started doing, and this has been since I joined the organization, I'm just, I just can't remember stuff. I hate having to remember stuff, so I write everything down, you know, like OneNote and Google Keep, and those are, like, my best friends, so I write everything down because I have trouble remembering stuff, and the way I look at it, if your computer is your, I mean, if your brain is your computer, why would you want to waste your hard drive with information you don't need to, right, you don't need to keep there, so what I like to do is I just put all that in my notebook so that my mind is free to actually focus on things and do some thinking.

So when I joined, this is something I learned quickly that we didn't have a lot of documentation, so I just started asking questions. Again, I have a good mix on my team. I have people on my team who've been with the organization for many, many, many years. I have people across the organization who've been with us for many years, so I just started asking. I just started asking a lot of questions, and I still do that. Every time I'm, you know, given a project I don't know much about because I don't have the history, I'm just going to ask the stupidest possible questions because no question is too stupid, so one thing you have to do is not be afraid of, you know, coming across as stupid just to ask questions. I write down those answers. My notebook became my source for everything. When I, you know, became the leader of the team, I realized, you know what, everybody on the team could benefit from this, and that's when I started posting this on our shared folder, and today we have data dictionaries. I want to say for 80% of our key data sets, we have data dictionaries to help us out with the metadata, so it was just something I started putting, and now when we have new data sets come in, we ask for the data dictionary up front. We ask that the data dictionary be created up front so that in the future no one has to do what we had to do to get here, so we just started by asking questions, putting it together, and now we have something.

Catherine, I see you have your hand raised. Yeah, just following up on that last statement, at what point do you decide as a data science team that it's not your team's responsibility to be producing those documentations and those data dictionaries, and like where is that your team grows enough that you can split off like a baby team, or is that, you know, something that you find yourself needing that you shouldn't be doing that sort of data grunt work? Exactly. That is always going to, that's an ongoing battle, right? There's always kind of an ambiguity on, hey,

does this fall on the software side, you know, the technology side of the data engineering team, or is this the PIA, which is our team, business information analysis, is this on PIA or is this on technology? That's an ongoing battle. Because we didn't have anything, you know, I just needed, we needed something, we took on the responsibility, did it ourselves, we got good, we have peers on the data engineering side and technology who are, you know, very helpful, but of course technology has other priorities, you know, they support technology in the store, you know, their priority is to make sure that systems in the store are working as opposed to, hey, does PIA have the data dictionary they need? So it's just consciously realizing that sometimes you have to say no, you know, another good book I would recommend, not my monkey. That's something that really helped me, you know, focus on, okay, this is something that we could be doing, but should we really be doing this? Just because my team has the skillset to support something, does it really fall on my team? So I just, you have to kind of politely start saying no to things and start holding people accountable for what they're supposed to do. Other teams, do you always succeed? Maybe not, but that's a start.

The other thing, the way I differentiate a project that's for our technology group versus my group is I tell the project owner, okay, are you trying to build a tool or are you, or an artifact that's going to support the business, or are you trying to identify ways to improve your business processes using data, right? Because a lot of times they'll come to us for tools, you know, hey, build us a tool where we can enter this information and it'll, you know, collect data in the database, but that's technology. So basically just defining the difference between what my team does and what the technology team does has helped us come a long way. Of course, you have to get buy-in. So my peer on the technology side, we work collaboratively. So a lot of it comes from collaboration. We put together kind of, you know, a map for the organization. Like, okay, if you want to do XYZ, the database team can help you. If you want ABC, the BIA team can help you. And it's a lot of just answering the same questions over and over again. Even though we have all this documentation, people are still going to come asking for stuff that probably BIA is not supposed to be doing, technology is supposed to do it. You just have to keep reminding people and eventually it catches on.

Learning from failure

Namrata, I like the perspective you shared on failure, and I'm curious, is there like a certain failure or something that you have in mind that you really learned from?

Failure that I learned from? Well, I learned from all my failures. I'm trying to think of which one because I have so many. Let's see.

I think for me, because, you know, the people aspect of things is always stressful for me. The data aspect, you know, it's fun. I can do it. It's, you know, when I'm talking to people, I think one thing that I learned from my failures was reading the room. A big thing I've learned is the data story that I'm telling is going to have many different versions. It's going to tell the same story, but it's going to have many different versions depending on who I'm talking to.

If I'm talking to a, you know, district manager or a senior vice president person, you know, on UPS or the UPS store side, I am not going to give them like a page-long story, right? I'm not going to give them too much detail. I'm going to cut right to the chase and give them information that matters because they don't need the details, right? If I'm talking to a peer, I may give more information, and if I'm talking to my team, I'm going to go right into the weeds, so I learned this the hard way, but it's knowing your audience, reading the room, and just cater, you know, tailor your story for your audience, and that goes a long way.

it's knowing your audience, reading the room, and just cater, you know, tailor your story for your audience, and that goes a long way.

Restarting the data science journey

That's a great tip. I wanted to make sure that as we get to the last 10 minutes here that I didn't miss anybody's questions in there, so feel free to, like, post it in the chat again if you want, but one question, and I think Deepshy sold us from you on a Slack group that we're both a part of, but the question was, if you were to restart your data science journey, would it look different, and how? I'm sorry, could you repeat that? I lost it. Sorry. If you were to restart your data science journey, would it look different, and how? Yes, it would look different for sure, because it would be more intentional than it was. I would do a lot of, you know, the research, the risk aspect of it, because I, up until, like, maybe two years ago, I never took any risks. Maybe I took a few calculated risks, but I was still on the conservative side, so if I had to start my data science journey again, I would take these risks early on and just, you know, focus on learning more than I have. I did, at some point, I felt like I stagnated, and then I had to kind of, you know, wake myself up and be like, hey, go back, learn more. This is not the end of it, right? So, if I had to start over, I would definitely take more risks early on, and maybe I'd be better off than I am today.

Open roles and one-on-one mentoring

Well, you sound great, and I see so many people are asking or saying you'd be a great manager to work for, so I did want to ask if your team has any open roles or anything that you would want to share in the chat, too. I don't have any open roles right now, but UPS is a large company, right? So, I'm sure there are openings across the country. What I can do is, I'm sure you all have my LinkedIn, but I'm going to drop it in the chat anyway. Feel free to send me an invite. Feel free to connect with me. I also have some free slots I open up on my calendar, if you, you know, go to my LinkedIn. Actually, let me just post that, too. I have free slots on my, I have paid slots, too, but I have free slots, too, if, you know, you would like to have a one-on-one chat with me about anything. If you just want to say hi, if you want to, you know, just talk about data or dogs, I'm happy to do that.

I love that you do that. When did you start doing that? What's the story behind it? Yeah, there is. So, last year, I completed my five-year, I had celebrated my fifth-year anniversary with UPS Store, and, you know, I'm a manager now, so I have to, I can't just focus on my data skills. I want to also focus on my people skills. I realized that, you know, this was a great time to kind of step outside my comfort zone and do what I am most anxious to do, which is interact with people. So, I joined the mentoring club as a mentor where I met a lot of good people, a lot of, you know, people starting out in data science, and I got to network a lot, but I just didn't like the way the mentoring club was set up in the sense that, you know, I didn't have a lot of flexibility with, you know, calendar and scheduling and things like that, and I was like, why do I need the mentoring club? I can just do this on my own, and so I just put around, like, December is when I was like, all right, I'm going to put out a link on my own, and I created some slots on my calendar, and it's been fun. I get to see, like, I get to meet a lot of interesting people. I met someone from, of course, India because I'm from India, so a lot of Indian aspiring data scientists, but I met someone from Africa. There was someone from Canada. It's fun, you know. Most of the times, you know, people stick to the free one-on-one introductory session. Some go into the one-on-one where I can, you know, help with resumes and, you know, help with specific data problems and things like that, but it's a lot of fun. It's something I do on the side, and because I enjoy data, it kind of softens the blow for me to have to interact with people, so it helps. I love that you do that. It's, like, makes me think that would be really cool to set up for myself, too, like having blocks of time. Yeah. You could just connect with people and talk one-on-one. Yeah, it's a lot of fun.

Excitement for the year ahead

Something that I usually ask in the beginning, and I hadn't asked this, but what is something you're most excited about in data science as you look to the year ahead for your team? So, this year, well, towards the end of last year, we started improving all our predictive models. We have five recurring ones that we use all the time, and one of those is what we improved to get towards the end of last year, and I got it to 98% to 99% accurate on a VTB level, which I love. I mean, 1% model error, like, how is that possible, right? I was like, am I overfitting? I was, like, all skeptical about it, but it's been doing well. We've kind of built, like, a workflow-based approach to predictive analytics, which allows us to continuously improve those models, so this year, we're going to do that for all our models, and I'm super excited because it's going to be a lot of fun. So, I think that's what I'm looking forward to the most.

Tips on mentorship

That's awesome. On the topic of mentorship, too, do you have any tips for some of us here who want to be a mentor for someone, or things you've learned in doing those, like, one-on-one sessions with people? Yeah, mentoring, look, I don't consider myself an expert in anything, right? The mindset is I'm always going to keep learning, but as a mentor, I can always help someone who's a few steps behind me on their learning journey, so that's what mentorship is to me. I don't consider myself an expert on anything. I'm just trying to help people who are probably a few steps behind me and tell them what I've learned, and hopefully, it helps them in their journey. So, if that's how you approach mentorship, you can start today. Just, you know, reach out to your network, see if there's anyone that could benefit from your learnings, and voila, you're a mentor.

I don't consider myself an expert in anything, right? The mindset is I'm always going to keep learning, but as a mentor, I can always help someone who's a few steps behind me on their learning journey, so that's what mentorship is to me.

On the other side, have you had mentors in your career? Yes, I've had. I'm very fortunate for the mentors I have. In my case, I was lucky because they were also my managers. I've had different managers in my journey at the UPS store, and I consider three of them my mentors who I've learned a lot from. One has taught me everything I know about the organization and data, and one has taught me the, you know, the work ethic aspect of it, and one taught me how to subtract. So, I've learned a lot from my mentors, and I kind of, like, you know, taken the best from everyone, and I feel like that's helped me succeed a lot.

Book recommendations

Thank you. I write random notes sometimes when I'm talking, and I remembered in, towards the beginning, you were talking about balancing empathy and accountability for leadership, and you gave us some great recommendations, like subtract. Are there any other books or resources you would recommend for leadership? All right. Books. Let's see. I'm not sure. So, because you're all data people, I'm sure most of you have read it, but I really love the Freakonomics series. I would also recommend, again, this is not specifically for leadership or data, but I love The Undoing Project by Michael Lewis. If you've read Moneyball, it's the same author. So, again, a lot of sports references. My husband had to explain a lot to me, and then Unapologetically Ambitious is another book I loved. It was more for, like, female leadership, but a lot about leadership. So, Unapologetically Ambitious, Freakonomics, Moneyball, Undoing Project, and, of course, Subtract, because I also mentioned that earlier. That's awesome. I was trying to, like, put them into the chat here right now, as well. And when I go through the recording, I will try to just make, like, a separate section that says book recommendations, too. But thank you so much, Namrata, for joining us. It's been such a pleasure chatting with you and getting to learn from you.

Is there anything that you kind of thought we were going to ask you that you want to share with us, or anything left that you want to tell us? No, I didn't really prepare for this, because I didn't know what to prepare for this. So, this is a lot of fun, really good questions. I really love the energy. I would love to connect with you guys on LinkedIn. You know, if you just want to say hi, drop me a message, I'll accept your connection request. And if you have any more questions for me, just feel free to shoot me messages on LinkedIn. I'm happy to answer them. Thank you. This was really fun for me. Thank you so much. I love the energy that you brought to today, as well. Have a great rest of the day, everybody. You, too. Thank you. Bye, everyone.