Autotrader is an online marketplace with over 3 million vehicle listings from 40,000 dealers and 250,000 private owners. They currently have a recommender system that is in BETA.
Target Users
The targeted users are anyone who has a current interest in the domain of vehicles. Whether it be buying or selling a vehicle, having a vehicle appraised, researching a vehicle, getting a loan or finding car insurance.
Key Goals
Their key goals are (1) to make money and (2) to be the one-stop shop for anything car relating. If anyone is thinking about doing anything with a vehicle, they want Autotrader to be the first, last, and/or only resource they go to. I would assume they make money from users listing a vehicle for sale, maybe take a cut when someone buys a vehicle, and gets ad revenue when someone is browsing the site doing research.
Ideas to help them accomplish their goals
For their business to be successful, they need to make the work as easy as possible for the customer. They have a recommender system in BETA but currently the user has to fill out as much parameters as possible to get a list of vehicles they are interested in. The user experience should be simplified where the customer has the least amount of input for the maximum amount of accurate results.
Currently, the site feels like flipping the pages of a magazine. The customer inputs the parameters like make, model, # of doors, size of engine and it takes you to the section where the vehicles match the paramters. A more intuitive system would take the type of vehicle they want, and use metadata to predict the vehicles the customer is looking for. An example would be if a customer clicked on a picture of an SUV, the system would take into account what the customer sees like color, # of doors, make, model and use those features to return a broad selector of cars resembling that one. If a customer clicks on a black Mercedes SUV, it would return dark colored luxury SUVs.
Multiple Design Scenarios
It would be wise to do a design scenario twice. The site offers many services to different people. Individuals interested in cars are one; Dealers or government agencies looking to buy/sell a large amount of vehicles may need different treatment. A scenario for different classifications or different contexts could be helpful.
Recommendation Improvements
Currently their recommendation system in BETA gives you vehicle recommendations base on the past choices. It doesn’t work very well. I tested it by picking a white truck and it recommended 3 black SUVs. Using search behavior to become more predictive could help. If a customer is looking at vehicles with 7 seats, it could predict that the customer may have a large family and display vehicles with good safety features or vomit-resistant cloth. I know I would appreciate that.