Vehicle Recommendations is a system in the form of different products across CarMax.com and the CarMax App. In the span of two years I conducted discovery, uncovered human problems, and designed products using our recommendations algorithm to progress customers.

“I don’t know if this car is a good deal”

When observing customers shop online, I often noticed this scenario: The customer would find a car they were interested in but wouldn’t let themselves become too interested in the car, because there might be a “better deal.” Then they spent a lot of time going back and forth from search results to car product pages, never feeling confident that they had found “the one.”

This was inefficient for the customer and bad for the business, as most customers would eventually leave the site all together.

The Solution: Compare to Similar Cars

The final design allows users to:

• Compare the car they are looking at with cars of the same model and comparative models in a similar price range

• View comparable cars that are close to where the user lives in case the car they are looking at is further away

• Shop more efficiently and progress

Increased all email leads by over 6%.

“CarMax doesn’t have a car I want”

The Problem: Users who reached the end of search results didn’t have an avenue to continue to search, or sometimes didn’t even know what to search for next.

Solution #1: Recommendations after small set of search results

We found that when users only had 1-10 search results, they didn’t find something that they liked and often left our site.

The final design allows users to:

• View cars that fit almost all of their criteria

• Educate themselves on what make models are comparable

• See that CarMax has plenty of inventory they may be interested in

Solution #2: Zero Search Results recommendations

When customers entered in a search that yielded zero results, they were only given the option to expand their search radius. This wasn’t helpful or obvious to most users. We wanted to give them options that fit their criteria as closely as possible, based on the most important attributes: price, mileage and year.

Increased all email leads by over 5%.

“The car I wanted was sold”

With an account, customers can save cars they are interested in. But unless the customer moves fast, the cars they save are likely to be sold within the same week. Previously, we offered no help to find a different car when this scenario occurred.

The Solution: Recommendations on Saved Cars

Used the recommendation algorithm to offer similar options to customers based on their saved cars. This was especially helpful when the saved cars were sold to different customers.

This test increased all email leads by 5%.

Recommendations on CarMax.com have led to more than a 1% increase in company sales per year which equates to more than $170 million.