June 1, 2020Analytics Leader Spotlight: Mark Stange-Tregear, Rakuten Rewards
Welcome back to our “Analytics Leader Spotlight” series, where we chat with individuals who are leading their team to success with the use of data analytics. In today’s spotlight, we are excited to introduce you to Aamedh Bhargava, Senior Analyst, Supply Chain at Home Depot.
Can you please introduce yourself and talk about how you got started in the data and analytics field? How has your career ultimately led you to where you are today?
A: Thank you for having me. I grew up in Mumbai, India, and pursued an undergraduate degree in Mechanical Engineering. My first job was in R&D for an automobile component manufacturing firm. This job made me realize that I had more of an interest in the numbers, rather than mechanical principles at play. That led me down the path of Industrial Engineering, and brought me to the US!
I was lucky to get into Georgia Tech, in Atlanta, and I felt at ease with the course offerings. While being math-heavy, the program was attuned to supply chain principles as well. I felt that most of those concepts were relatable to small everyday problems.
I chanced upon an opportunity to intern at The Home Depot. In a way, I never looked back. I started working with Business Intelligence in Upstream Supply Chain, where our main focus was on collecting historical weather data to enable analyses that can help make better decisions during potential hurricanes or droughts. I got my full-time role in the Analytics team, and with it, I realized my passion for Optimization. The Home Depot has a large supply chain, with tremendous resources. For me, there was an enormous opportunity to leverage these resources and bring them in sync by working in harmony. During my time, I touched parts of Import Freight Flow, Network Planning, Event Scheduling, and Cross-dock Operational Decision Making.
Online retail has been a growing space, and the pandemic has just accelerated it by leaps and bounds. There has been a lot of investment by Home Depot in improving their delivery channels, with the goal of being “The fastest, most efficient, and reliable delivery in home improvement”. That brought me to the Delivery Network Analytics team, focusing on Last-Mile Optimization.
At Home Depot, I’m sure you end up facing a lot of complex problems that you have to solve through data. What are the responsibilities of your role, and how does that fit into the position you’re in within the delivery analytics team?
A: The primary focus is on the delivery networks leading to the customer – may it be online orders delivered to home or the store for a pickup, orders made at stores for home, or even job site locations.
My involvement comes in helping strategic planning and decision making. The analyses conducted are usually forecasting and optimization problems. We work closely with operational teams to understand their challenges, present multiple scenarios, and help work through a strategy that best fits the need. While almost all the optimization models start with “minimizing cost”, there are so many intangibles to consider, such as speed to customer, ease of operation, and integration with systems; all of which makes it a tricky balancing act.
When we talk to a lot of our partners through AtScale, a big component of it is what you said around cost reduction. There’s a ton of opportunity when you get to optimizing some of this in terms of driving revenue, like making sure you’re pricing things the right way to drive incentives for people to order. It sounds like it’s more than just a cost savings problem that you’re trying to solve.
A: Absolutely. We have different teams that even work with the promise that should be shown to the customer whenever they’re placing an order and how that influences their decision. For example, we don’t want to commit to delivering something within a couple of days when it might take longer, because we would lose our customer base over time. We also don’t want to show something that might take too long, just because it’s more efficient in our system to get it there- because now it’s all about the speed in delivering the final product to the customer. So there are many different teams working in unison to try and get that solution together.
Those have to be some really fun problems to solve and then get to see it in reality, because ultimately all the work you’re doing is ending up in production for a lot of the orders that are being placed across the multiple countries that The Home Depot operates in. When you look at what you’re doing within this role and even prior roles, but maybe this role has more interesting things to talk about, what are the two to three largest challenges that you face today? How does your team overcome them?
A: As I mentioned before, balancing multiple objectives may not be at an apples-to-apples comparison. We want to make decisions that keep costs low, while improving speed or service to the consumer, and keep things relatively easy to execute. It is a tricky balancing act to get correct first, and can always be improved with Machine Learning algorithms. It would be a pity if we come up with a perfect solution that would be very hard or might take a long time to get into production and finally be implemented. So, it’s that tricky balancing act on getting it correct first, and then the imports can obviously be improved through Machine Learning algorithms, among various other sorts.
There’s also a large lift in cleaning the data and extracting the relevant fields at the right level of detail that’s required. We are reliant on receiving information from our partners, mainly from the carriers, and from our warehouse management partners. We also have Business Intelligence teams who consume, standardize, and vet all this data to make it easily consumable for us.
It is also crucial to speak the same language as stakeholders. There are so many pitfalls in not realizing the entire need before going in, building out, and trying to refine a model. While stakeholders might be focusing on a different definition of service or some kind of similar metric, they also may want multiple views; so there should be a focus on the repeatability of the solution in order to help them get different order scenarios working through.
The first component you talked about with the granularity of data and making sure that you have access to the right data. How long does it usually take for your team to be able to create the data pipeline to deliver the data for analysis from your team?
A: It’s more of an ongoing process by the Business Intelligence teams where they do realize a need to enable the analytics partners and for them to do the sorts of analysis that they want to. So, many times we might need to see invoices at a higher level, like a monthly transactional level, and there are teams that try to break it down and allot it in the right way so that we do have that historical data set to use in our analysis.
For new requests, there are challenges in trying to find the right data set and go through the vetting process in order to check if there are outliers. But it’s always an incremental process where we feel instead of having the entire process of building out a model sequentially where we wait on our partners, if we could work and create that structure that other teams can come in and plug their inputs into, we can try and make the final product as quick as possible.
It sounds like there are a bunch of people involved in that pipeline, in terms of figuring out what data you need, how to prep it, and then how to present it to the stakeholder. As the person who’s deriving the insight, who are the stakeholders that you’re dealing with, and how influential are they for you in determining what type of analysis you’re doing on a daily basis?
A: The largest stakeholders are our business partners who bring the challenges to us, and we ultimately have to go in and make those changes and execute our solutions the way they see fit. The operational teams might be those that are related to financial planning, sourcing, and those controlling those bids and pricing strategies, those that deal with the day-to-day capacity requirements. All of these are trained business operational teams, so it’s really important to understand their point of view, and the challenges they face. That helps us better serve the given solution that we can provide in order to help them.
So when you talk about business partners, those can be people internal to The Home Depot business, but they are just representatives of other groups. like financial planning, and procurement as examples. Is that correct?
A: Yes. For me, these are all my consumers of the solutions that I provide.
Let’s switch gears, I know you’ve been at Home Depot for a couple of years. Can you recall a project or initiative that you’re proud of?
A: For the most part, I’ve been with our Upstream Transportation business. We go through an annual bid cycle with carriers to set contracts on our terms of business. This is an understanding of a capacity requirement and a pricing structure that we will be held accountable to. While there have been on-going initiatives to better utilize our trucks and reduce our mileage footprint, growing demand means that we need a larger fleet. What we looked into was to right-size our fleet by making better routing decisions and smarter contract structures.
It was my first foray into a full-scale optimized solution that needed to be proven across various seasonality periods, probabilistic risk assessment, and incremental implementation plans. It not only helped reduce operational expenses, but it also gave us more visibility and control and was the launching pad for the adoption of more complex routing tools. So that’s definitely the project I am the proudest of.
That sounds like a super complex project. How long from start to finish did that take?
A: I would say we worked on that for approximately four to five months. We know the bid schedule and the sort of structure plus we had the previous years’ information in order to help us prep us in time for when we can refresh our data. It was also my first experience in building out a model that the answer, frankly, was afforded the time to learn and to try out new and different approaches in order to have a solution that scales well, gives us outputs at a relatively fast rate so we can run through the different scenarios that we wanted to see.
The fact that you can do that within a five to six month period, that could sound long to some people, but it’s also a pretty tremendous feat from the fact that you probably had so many different data sets that you have to pull together and then be able to do analysis on top of, you know, it’s underestimated how much work and effort really goes into that.
A: Sure, the projects that we work on are a mix of the short-term ad hoc questions that we need to answer while having some projects that we benchmark through the year that we have our main focus on. We also know that in this project example, would have a large impact on our business and help save a lot of transportation expenses. So we were, thankfully, afforded the time to work on it and try and come up with the best possible solution.
One thing that’s super important in being a data and analytics leader is the ability to measure success. From your standpoint, what do you do to measure the success in the work that you do?
A: It is definitely gratifying seeing your solution implemented, on a personal level. The type of work we do impacts the ease of our store associates, warehouse associates, and customers directly.
For a project, we do set measurable goals and develop reporting views to track key metrics: both financial and operational. It is always interesting to see how the forecasts align with the results. Looking holistically at the entire exercise, it is important for us to look at how repeatable the analysis is, in case we need to look at what-if scenarios or refresh it on a schedule.
In the long term, I view success as gaining the confidence of my stakeholders and business leaders. As analysts, we have to prove our solution out, first by validating it to ourselves and then gaining the trust from our leadership. The trust that we can understand what they want, promise what is possible, and deliver what they need. I try to be the “Swiss Army Knife” for them.
I think the ability for your leadership to believe in you and the work that you do, especially around predictive analytics, because there are so many variables that can go into making a prediction. So, if they have the confidence in the work that you do, there’s tons of value to that as a measurement of success.
It’s so important to try and gain confidence, and also very satisfying knowing that you’re the person they can go to when there are difficult problems that they want answers to.
As we wrap up this interview, do you have any advice to others in the industry who are building their careers in the data and analytics space?
A: Find the pleasure in what you do. Industry problems do get frustrating. You will wade through murky inputs, slow run times, and potential bottlenecks in execution.There is tremendous satisfaction when everything works out.
While we broaden our tool kit with technical expertise, we need to learn to marry science with the art. My advice would be to let go of the need to make things perfect and make assumptions that get the solution executable for the business.
Sometimes, I feel debugging is like detective work, and fixing a problem can feel as enjoyable. Find what excites you, and on its own, you will find the energy and enthusiasm to learn more and perform better.