Deriving Actionable Strategies Using Machine Learning and Correlation Analysis

Our goal with this project was to figure out how CarMax could best tailor it’s marketing & inventory strategies to draw in distinct segments of customers. We approached this from a three pronged perspective. First we looked broadly in terms of large demographics and found trends there. For example by looking at how customers from certain age ranges favor certain categories of cars provides a lot of valuable information that can then be applied to targeted ad campaigns but also influence the inventory choices of each individual CarMax lot. Secondly, we took it a step further. Rather than only looking at large demographics as a whole, we knew that we could use Machine Learning models such as Neural Nets and Random Forests to successfully predict the choices a customer might make. By knowing what the customer is hoping to get out of the deal, the employee can really streamline the CarMax experience as a whole by personalizing their pitch to this sole customer. The goal of these models is to provide a proof of concept and emphasize methodologies. With larger data sets, the accuracies of these models can be significantly improved. Ultimately utilizing these models led us to create our proposed Oracle User Interface Tool. By marketing CarMax products through a tool such as this, it represents the ultimate personalized experience custom fit for each unique customer and maximizes their satisfaction. Find a news article about our achievement in the CarMax analytics competition here.