Evan Wimpey: Hello and welcome to the Mining Your Own Business podcast. I’m your host, Evan Wimpey. And today I’m super excited to introduce our guest, Xhoana Laska. Xhoana is a data scientist at National Coatings and Supplies. So we’re super happy to have her on the show. Listen to her talk about her role and what she’s doing there at NCS.
So Xhoana, thanks so much for coming on the show.
Xhoana Laska: Hey Evan. Yeah, thank you. Thank you so much for having me. I’m excited to be here.
Evan Wimpey: All right. Fantastic. Well, to get started, can you give us just a little bit of your background and then maybe a little bit about NCS for folks who aren’t familiar?
Xhoana Laska: Yeah, sure.
So yeah, my name is Xhoana Laska. And the first two letters of my name sound like J in English. Just a little fun fact from my Albanian roots. So yeah, I’m originally from Albania in Southeastern Europe and my background is actually in pure mathematics. I hold a bachelor and a master’s degree in mathematics.
And after relocating to the US, I furthered my studies at NC State University, where I got my master’s degree in statistics. And yeah, so currently I work as a data scientist at National Coatings and Supplies. Otherwise known as NCS, which is a nationwide distributor specializing in automotive paint.
I am part of the supply chain team, and my work focuses mainly on inventory management and forecasting the demand, replenishment, and shipment sales.
Evan Wimpey: Alright, fantastic. And Xhoana, if I’m not mistaken, you’re, maybe you’re not the only data scientist, but you’re one of the first data scientists at NCS.
Xhoana Laska: Yes. Correct.
Evan Wimpey: Yeah. Maybe you could walk us through sort of what it was like getting started there. Maybe if there’s a project—one of the first projects that you can take us through how it got started, how it got off the ground.
Xhoana Laska: Yeah, so I started my journey at NCS, working as a pricing analyst, not a data scientist, and I’m referring to customer pricing, not the vendor pricing.
Yeah, at that time, my first analytical project was to implement a new price segmentation on 20,000 customers on thousands of products. And basically that implementation would involve not changing the customer pricing itself, but just the pricing structure, and that should be done while making sure that no data redundancy occurred.
So yeah, just to give a better idea of the complexity level of this project. So basically this 20,000 customers were assigned to 500 different price profiles, and they had to be converted to just nine price profiles. And so, yeah, there was a lot of data to be handled. And the existing process at that time was pulling all the data from our ERP system into Excel sheets. Manipulate the data, do a bunch of calculations there and then come up with the new price profile for every customer. Now, the issue there was that I had to pull the data by groups of customers because if I was pulling the data for all the customers at once, then the data size would exceed the Excel limit of one million records.
So that was not feasible at all. And that’s when I started thinking, how can I leverage Python for this project? So yeah, after I got a good understanding of the old pricing structure and the new one and the conversion logic from the old to the new pricing structure, then I first evaluated like, okay, each step of the process, what is the time that each step of the process was going to take?
I did an evaluation and it turned out that it’ll take me a whole year to do all these calculations in Excel and then implement the change changes into the system. So also, I evaluated how much time it would take me to build the code in Python, test everything, and then implement the new prices. And it turned out to be that that would take me approximately two months.
So we are talking about like 10 months of labor hours saved and around, I guess, 1,700 labor hours saved, which is a lot. And yeah, so first I had to get people to see and to understand the benefits of automating this process and doing data analysis on a program other than Excel, which they had been using forever. So yeah, I talked to my managers and I proposed this new solution.
I shared with them all the calculations on how I came up with these time estimations, and also shared with them, how like what the impact would be from a seemingly small human error on the customer pricing. And we aim to be as accurate as possible with customer pricing, because, you know, we are customers ourselves in our everyday life.
We are very sensitive when it comes to pricing. So we want to make sure the data is accurate—the pricing data is accurate. So yeah, after I shared all this benefits and advantages, my manager approved this proposal. I coded the program, implemented it and yeah, it took me around two months. We received no customer complaints about the pricing.
The data was accurate. No redundancy. So yeah, that was a win for me. My first analytical project went great. And I like was able to get people to buy in this new approach.
Evan Wimpey: Yeah, I love it. It sounds like a huge win. I think even for, for organizations that have a well-established data science team, it’s hard to get stakeholder buy in. It’s hard to get a project off the ground. So I really love the estimated in labor hours saved. And you de-risk it too, by, you know, if there are mistakes, like here’s how robust this solution will be. So we’re, we’re not worried about mistakes to help, to help de-risk it. So at that, at that point, I would think it’s kind of a no brainer for a team, a management team. Now, is there anybody internally? Is there anybody on the other side that has to absorb the work that you’ve done or who has to consume your, your Python output or run the Python scripts? Is there, is there any change that you’ve imparted or everything is you’ve already built it and it just goes straight to customer pricing implementation.
Xhoana Laska: Yeah, that’s a good question. So yeah, we have our IT team that supports all the analytical projects that we work on. So we are one of the largest distributors in the PBE industry, the paint, body and equipment industry in the nation. And we have our ERP system where all the day-to-day business activities are managed.
So basically we have a system with where all the data is centralized. And here I’m referring to transaction data planning, financial, accounting data, and so on. And so not, I don’t want to go too much into the technicalities of the data architecture, but basically we get this data from the ERP system, and those are fed to our data warehousing platform.
We use Snowflake for that and that’s where I pull the data from and build the machine learning models and applications, mostly in Python. Then it’s important to have like a testing environment where we, where we test this new processes, this new time-consuming processes and make sure that everything is good before we go live into the production environment. So yeah, we have this systems in place now, and of course, for a data scientist, it’s important to be able to tell a story through the data and do that in a digestible ways. And we use Power BI to create is dashboards and reports to create or to tell the story through visualization.
So yeah, it’s important to have a data pipeline which sources the data, transforms it into structures that are needed and make the data scientists work easier. So, yeah, that’s what we currently have in place and has helped me a lot.
Evan Wimpey: Do you, as a former data scientist myself, and as an admittedly lazy data scientist, I leaned on a data engineer or someone whose sole focus was on sort of that data accessibility, data quality.
Can you serve the data to the right place? Do you have data engineering resources or is this sort of a full stack data science role where you’re having to source and check the data yourself.
Xhoana Laska: Yeah, so as I said, it’s the IT team that makes sure that data quality is You know, like the data is accurate and is clean and it’s organized into structures that I can work with.
So the data is organized into tables in Snowflake where I can use SQL to retrieve the data and then run my analysis in Python. Yeah, so we also have like an integration platform that, like orchestrates the data and make sure that all the systems can communicate and run all the processes needed.
But, like, from my perspective, like I mostly use SQL, Snowflake, Python and Power BI for my reports.
Evan Wimpey: Are these things, like, were you already building with, with Snowflake before you started doing data science work? So the IT had already teamed, the IT team had already built out this capability and so data was sort of available for you.
Xhoana Laska: Yes, yes, the Snowflake was already there which was great, you know, but still the deep analysis part was, was still done in Excel sheets. So there was no, you know, development environment where they can run heavy analysis. Yeah, I proposed utilizing Python for calculations that need so much time.
And of course there are, I cannot implement my machine learning models in Excel. So, you know, it was crucial to, to have a scripting language or, or a development environment to do all that.
Evan Wimpey: Gotcha. Okay. Yeah, I think for some of our audience, like thinking about where to make hires to try to improve capabilities.
And it sounds like bringing in data science, a data scientist, given or allowing a pricing analyst to do some data science work was helpful because you already sort of had some level of IT and infrastructure support. So you didn’t have to go look for data and check data like you had a snowflake that you could just query data from.
Xhoana Laska: Correct.
Evan Wimpey: Were you new to Python and was NCS new to Python?
Xhoana Laska: So I was not new to Python. I had some like prior knowledge, and I learned Python using online resources and where I got like my fundamentals in Python and data science combined. But yeah, NCS was not familiar with Python when I first started there. Yeah.
Evan Wimpey: Awesome. So you’ve had, it seems like you had a very successful first effort, then maybe you got to take a 10 month vacation because you saved, saved 10 months of work.
Xhoana Laska: It didn’t work like that. I wish.
Evan Wimpey: That’s too bad. Well, so then presumably there’s some excitement around data science capabilities now.
So what do you do after that? You’ve got a good project. You’ve got, you know, stakeholders are. Wow. This is impressive. What’s next? How do you decide what to work on next? Is this something where you’re looking for capabilities or looking for opportunities or where folks are coming to you to say, can you do this?
Can you do this? Can you do this?
Xhoana Laska: Yeah. So like, as we mentioned, there is not data science history at NCS. I can, I translate that as having tons of opportunities for innovation. So there are tons of opportunities to automate business processes or to build and enhance forecasting models and so on, but I want to say like, I’ve been fortunate enough to be part of a team with open minded people and having those transparent meetings on what the.
The business most critical needs are. I would say like when it comes to deciding on what to work on, it has been both a combination of taking initiatives and assessing what the priorities are. So for example, with the first analytical project that I just mentioned earlier, it was my initiative to follow that innovative approach to implement the price, the new price segmentation.
And yeah, so afterwards, considering the pressing need that the company had to efficiently reduce overstock, to prior to two projects to priority at that time. And the first one was, I had to build an inventory optimization process that would maintain inventory level across 200 plus stores and three warehouses that we have.
And the second project was to to improve the replenishment forecasting model. So, so basically we were, we were fighting that overstock problem from two directions. Yeah, I took on these two projects. And yeah, that’s how it went.
Evan Wimpey: Has NCS already invested in other data science resources?
Are they looking to grow? You know, you’ve identified some capabilities and once folks sort of see success, you know, the inventory work goes well. The replenishment work goes well. Is this, is this an opportunity you think to grow within or for NCS to grow data science capability?
Xhoana Laska: Yeah, I’d say we are going to that direction.
We have already developed an analytical team that works on these projects, you know, by assessing the priorities again. But yeah, we would like to expand on that in that regard and yeah, see what we can do with all that massive data that we have and gain some more insights and some more actions to, you know, boost revenue and optimize and do all these cool things that data scientists do.
Evan Wimpey: Fantastic. Yeah. Very cool. You know, you can, you’ve got math stats background. You came in not as a data scientist, but found your own data science work to do and delivered a very successful project. Can you talk maybe specific to you, but maybe more general, if somebody is starting a role in data science at a company that sort of has no pre-existing data science capabilities, what do you think are important things for, maybe important things for them to focus on, or what are important skill sets to, to come in with?
Xhoana Laska: Yeah, first of all, I’d say it’s instrumental to first gain deep insight into how the business works. And you need people that do a good job to fill you in.
Of course, you yourself, you need to pay close attention to the details. For example, when I started, I had to first understand the distribution network. So we are a distributor company, I had to understand how the products flows from the suppliers to our warehouses to our stores and to the end customers.
And then I was able to develop that inventory optimization logic. Yeah, and then I’d say, based on my personal experience. As you mentioned, having that solid mathematics foundation that has helped me to being able to break down a problem and figuring out paths towards the solution. And here I’m referring to any kind of problem, either if it’s a business problem or a coding problem, you know, so on.
What else? I would say probably. You know, knowing some scripting languages, it’s either if it’s Python or something else or something else that it’s being used in our field is a big deal because you have to have an environment where to run all the, all the analysis. One other thing that I think it’s crucial is to be resourceful. So there are times where when not all the answers lie within your organization and you might, you might not be familiar with something, but you don’t have to stay stagnant. Not do anything about it. So it’s up to you to research those answers and come up with your ideas Either if they, you know come up to be like not the best ideas, but it’s that, you know, you need to do the research and be proactive for sure I’d say the last but not the least taking initiatives for sure, like that worked for me.
And I guess if you believe that there are better ways to get something done, then take that initiative.
Evan Wimpey: I am very impressed with, with your initial analytics effort as a pricing analyst, almost sort of relatively new to the, to the company sort of make these pricing changes. It’s going to take a long time.
It’s very easy to just. Okay. Well, that’s as directed. That’s what I’ll do. But being able to take initiative and to take initiative, not just to say this might be a better way, but to quantify here’s how much better it’s going to be. And here’s your ways to de risk it. I, you know, I think it’s great. I think oftentimes a data science capability starts from the top down and it’s hard to get traction, but starting from the bottom up with a real problem that you’re already trying to work on and building this analytic capabilities, it’s, it makes it a lot easier for folks to say, Ooh, that’s nice.
We should apply that in other ways.
Xhoana Laska: It’s like going again to the, going back to the pricing project. So, especially being aware of the impact, a small error in the pricing could have, like we would, we would receive like tons of emails or calls from, you know, people in the field and like letting us know, you know, this group of customers is having this problem, they are getting price at a higher price that they should.
So what’s going on? Like not being aware that those things could happen, like I had to be extremely cautious and careful with the data and making and test everything first, and then, you know, go from there.
Evan Wimpey: I think that’s great. And as a consumer from now on, when I’m in a grocery store or fast food anywhere, when I complain about the prices, I’m going to start complaining about, so it’s a bug in a, some data scientist model of the change.
That’s why I pay too much.
Xhoana Laska: Yeah, we are humans too, you know?
Evan Wimpey: Perfect. Xhoana, I want to ask you one last question. You walked us through how you. Took initiative. And then also how you’re scoping out and prioritizing new projects. Let’s just put all that to the side. Let’s just say it’s the Xhoana show.
You’ve been there for some time. You get a chance to work on any type of project that you want. And. Everybody at NCS, all of the stakeholders are aligned to your vision. They say, just tell us what it needs to be. What kind of data science projects should we work on? What is it that you want to do?
Xhoana Laska: Oh, that’s an interesting question.
Well one thing that stood out to me when I first moved to the States was this idea of. returning the products after you purchase them. And that doesn’t happen everywhere. So that was surprising to me. And I definitely liked the idea, you know, you buy something, you can return it if you don’t like, or, you know, an outfit that doesn’t fit you and all that.
So from a customer side, that’s very exciting, you know. But on the business side, there are some implications that go with it. You know, there are, from a distributor side, like the way that I look at it now is, when customers return products, there are costs associated to it, like costs associated to having that product shipped back to us or restocking it costs to resell it.
So many factors that, you know, are part of that equation. And well, I would like to build a web-based interactive application that provides a full picture of the product returns. Like, for example, think of it as having the product return rate there. And what are the products that customers most frequently return?
How often does that occur? And stuff like that. And also, it would be interesting to have the capability within that tool that forecasts the returns. I’m all about forecasting. Sure, yeah. And I guess, yeah, that would be very important to when it comes to those agreements that we sign with the suppliers, for example, if we know beforehand what’s usually the return rate with a supplier and we have an agreement.
Okay. So the allowance of the return rate is 5 percent. If we see that actually the return rate is 10 percent, maybe we need to renegotiate that agreement, you know? So I guess that would be helpful. And in general, you know, I enjoy working on problems that involve bringing solutions that can potentially save money or optimize and be efficient at the end of the day.
Evan Wimpey: Awesome. Very cool. Xhoana, thank you so much for joining us today. It was great to learn about you and your role there. Some very exciting work. Xhoana is on LinkedIn. Is there anywhere else? Do you have—do you write blog anywhere? Is there a place we should point folks to?
Xhoana Laska: Yeah. I’d say yeah, my LinkedIn, yeah, it’s X-H as I mentioned earlier. So Xhoana, X–H-O-A-N-A, Laska. You can DM me there and find me there. But yeah, thank you so much, evan. It was so much fun and thank you for having me again.
Evan Wimpey: Thanks for coming on the show and thanks for listening, everyone.
In this episode of Mining Your Own Business, Data Scientist Xhoana Laska shares how she and her team are bringing data innovation to the field of industrial paint and coatings, an industry that became prominent in the late 1800s.
From process optimization to demand forecasting, Xhoana Laska enjoys creating data-driven solutions for complex business problems. And that’s at the heart of her work as a data scientist for National Coatings & Supplies | Single Source, Inc. (NCS SSI).
Evan Wimpey | Host