One of the greatest challenges faced by Carmax and its customers is the transfer costs associated with transferring a desired vehicle from one Carmax location to another. In order to mitigate this issue we propose the use of an app that utilizes machine learning not only to calculate transfer costs, but also suggest vehicles according to customers’ tastes and preferences. In our proposed app, all else equal, vehicles with lower transfer costs will be displayed more prominently in the matches than those with higher costs. We believe this would result in a net effect of customers paying less transfer fees while maintaining or even raising their satisfaction in purchase experience. This is because we believe that when people purchase used vehicles, they are not necessarily looking to purchase specific models, but rather vehicles that fit their budgets and preferences. By addressing the challenge of decreasing transportation costs though this approach, there is a greater likelihood that a candidate vehicle will both suit the customer’s tastes while also keeping transfer costs to a minimum. We came to this hypothesis after navigating through Carmax’s website, where we noticed that there are many filters but none pertaining to factors that may influence transfer costs. Another disadvantage of the website is that when there are too many or too little filters applied, there’s the possibility of search results being too narrow or broad, respectively. We value the freedom of customers to continue having access to Carmax’s nationwide inventory, and hence do not wish for them to feel like their choices are being limited by the desire to save money on transfer. At the same time, we also believe that because there is an element of “guessing and checking” with manually setting filters, customers might potentially narrow their search to vehicles with expensive transfer costs when other vehicles that would equally satisfy their preferences and needs go unnoticed. Our proposed app does not, however, preclude customers from searching for vehicles via existing means, as it does not seek to replace the website or the existing CarMax app, but is rather a tool that facilitates the search process while also providing customers with purchase options that could potentially reduce transfer cost.
The beauty of unsupervised machine learning is that it recognizes patterns that humans might not readily notice, meaning that even subconscious preferences could potentially be recognized. The app collects further information regarding users’ tastes through an interface that operates similarly to Tinder, and the user has the option to enter broad preferences such as vehicle age, type, price, and mileage, to narrow down candidate vehicles that appear on the swipe interface. Images and basic information of vehicles in Carmax’ nationwide inventory that fall within the user-defined parameters, if any, appear on the screen, and the user can either like or dislike the vehicle in addition to clicking on the image for a more complete summary of the vehicle. With every additional swipe, the app’s predictions will become more accurate, and with sufficient swipes, users can access their matches that list vehicles that fit their many preferences, with priority given to vehicles with lower transfer fees. Because transfer fees are not constant, however, and can change at any time due to factors such as inclement weather, machine learning will also be incorporated to predict transfer fees based on factors such as distance, season, desired pick-up date. We assume the following factors to impact transfer costs:
Location - distance is not the only factor that determines transportation costs, but the location of the origin and destination relative to major transportation lines. All else equal, transporting a vehicle from a big city to another big city would be less costly than transporting a vehicle between two small towns
Season - shipping tends to be more expensive during winters since many roads and highways are more dangerous.
Natural disasters - if an area is impacted by a flood, hurricane, tornado, etc., then shipping costs can temporarily spike.
Vehicle size - all else equal, larger and heavier vehicles are presumably more expensive to ship than smaller ones. We imagine a Ford Super Duty to cost more to ship than a Honda Civic.
Delivery Flexibility - some deliveries may be less expensive if the buyer is willing to wait - multiple orders that have overlaps along their respective shipping routes can be consolidated for less expensive shipping. If a user were adamant about a particular vehicle, future delivery dates can be accurately predicted with past data as well as real-time information regarding things such as weather and patterns.
Log in or sign up for Devpost to join the conversation.