Going from Zero to Data Scientist in 8-Months
Meet the Inaugural Data Science Accelerator Program Graduates
At American Tire Distributors (ATD), whether it’s spending 3–5 minutes a day on our micro-learning platform or attending a peer-led lunch and learn, our associates are always embracing our culture of knowledge.
At the end of 2019, our analytics team Launched the ATD Data Science Accelerator Program which provided our associates an opportunity to elevate their skills in data science and machine learning. The goal of the Data Science Accelerator (DSA) program is to serve dedicated and determined associates who desire to accelerate their careers by learning the fundamentals of data science. Through an enriched learning environment and the support from our DSA mentors, we achieved this by giving graduates the ability to tackle business problems with data-driven solutions.
“This program provides our associates the opportunity to learn new skills in data science and build machine learning models,” said Tim Eisenmann, chief analytics officer and senior vice president of advanced analytics at ATD. “I’m proud to celebrate what our graduates and mentors were able to accomplish throughout the entire program and it’s one of the reasons why I’m excited to be part of ATD.”
The mentors and graduates both personally invested 10 to 15 hours per week to take part in this journey. Throughout the program, each graduate could lean on David, Salman or Sid, the DSA mentors, from our analytics team for support. Whether it was assisting with an online learning module or preparing the participants for their capstone project, each mentor was there every step of the way.
“No matter what function of the organization associates are in, we’re excited to support and help them learn the fundamentals of data science,” said David, a DSA mentor. “I’m grateful for the ability to help my colleagues learn new skills and see them add value to the business by applying this knowledge in their roles.”
The capstone project provided the graduates the opportunity to put their new skills to the test and showcase their abilities. After completing the online modules, the graduates worked with their business leaders to identify a relevant business problem. With the support of the DSA mentors, graduates worked on developing an actionable data-driven solution that could be introduced into the day-to-day workflow at ATD. Meet the inaugural DSA program graduates and learn what the outcome was of their capstone projects.
Colleen is a member of our strategy team and for her capstone project, she identified an area of opportunity with an FP&A monthly forecast model that impacts business decisions. Currently, this forecast model consists of a very manual process and only uses two years’ worth of data. With a data-minded approach, Colleen expanded her dataset with more historical data and looked at various segments to further filter the data depending on the business channel. “Working with forecast models, I knew a little SQL, but that was just to get by. Being able to learn how to code in Python and build advanced models has helped tremendously,” said Colleen.
To help her improve the model, she leveraged her new knowledge of Python and a variety of data packages (Pandas, NumPy, Scikit Learn and Matplotlib) to build a regression random forest model. This type of model handled the more massive categorical datasets better while minimizing the prediction error. Resulting in a more accurate forecast. “One of the most important things I learned is that data discovery and proper cleaning is an important step when building a model,” shared Colleen. Focusing on these important steps, Colleen identified key components within the dataset that were useful predictor features. After running her model, she achieved an R-squared value of .50. With this foundation, she identified opportunities to incorporate other relevant datasets and add additional predictor features to improve the model’s accuracy continuously.
“I’ve learned a lot from the DSA program. I’m grateful for the opportunity to add new skills and can’t thank the mentors enough for their support,” she said when reflecting on her experience.
Daniel is a member of our eCommerce team and for his capstone project, he identified an area of opportunity to enhance the customer experience across our industry-leading ATDOnline platform. He asked himself, “how can ATD provide customers with a unique shopping experience by feeding relevant data customized to their needs?” With the help of his new skills, Daniel created a model that identified the products customers needed to serve their consumers.
Through data discovery, he identified what key components could be used to achieve the unique shopping experience based on customers’ behavioral traits. “Being able to work with real-world data and examples, presents challenges you’ll never see in a classroom or tutorial-based learning,” said Daniel. Working through these challenges, he leveraged Big Query to enrich the data with more detailed specs from our database. As shopping behaviors and patterns shifted, the model continuously updated and increased its accuracy. Nearly 30 million rows (equal to one month) of data was used to create his model and deliver the unique shopping experience. Daniel was able to obtain an AUC score of 81% and is looking forward to increasing this by building out his model with more data.
“The DSA program has been an amazing experience and provided me the knowledge to make a positive impact on my team,” said Daniel.
Sridhar is a member of our IT team and for his capstone project, he wanted to create a better associate experience. To do this, he identified a way to automate how incident tickets are submitted and handled on the backend. By implementing a machine learning classification model, giving associates valuable time back in their day, the data entry can be automated. Resolution time would also improve by routing the ticket to the appropriate support team and recommending the right service.
“The DSA program was challenging, but I knew I had the support of the other participants and the mentors,” said Sridhar. Having this support gave him the ability to look holistically at our incident response process and identify how he could provide a better associate experience.
With proper data collection and preprocessing, Sridhar transformed the data using feature engineering to assist with scaling/normalizing. Using a XGBoost (eXtreme Gradient Boosting) classification model aided in automating the incident report by allowing the decision trees to work sequentially. This helped ensure the incident ticket was filled out correctly and received the appropriate priority level. Sridhar was able to achieve over a 78% accuracy rating along with a 98% precision and recall score. Overall, he would be giving back approximately 125 hours a month to all his colleagues collectively. Talk about a time saver!
“From day one, the DSA program has been one of the best learning experiences. I’m looking forward to giving back and helping the next class of associates as they go through the DSA program,” said Sridhar when reflecting on the impact of the program.
We look forward to the next class too Sridhar. When our analytics team discussed the idea of this program, no one could predict the impact it would make on the associates and the business. From the moment the program was launched to our virtual Graduation Ceremony and every step in-between, it has been a fantastic journey.
Join us in congratulating our inaugural class of graduates and thanking all of those who made the first DSA program a success. We’re just getting started and look forward to welcoming our second class of associates in the near future.