Carvana

Improving operational efficiency by automating outdated workflows.

Project Overview

Challenge

One of Carvana’s main business functions is buying used cars from customers across the US. However, their process for doing so was outdated - involving manual data entry into Google Sheets, Salesforce, and other tedious tasks. Our challenge was to build a modern car title processing tool that aligns with Carvana’s claim to be the “new” way to buy and sell cars.

Solution

Designed an in-house web application for to streamline Carvana employees’ workflows and automate the vehicle titling process.
After the initial launch of this tool, we added work prioritization features so Carvana specialists can focus on the most important business tasks.
Automare manual employee workflows with AI.

What I’m Proud Of

I led design with the support of a junior designer and partnered with cross-functional stakeholders. This included 1:1 pairing sessions with developers to work through technical constraints in real time and helping the team develop a comprehensive QA process. I was also able to influence my team by proposing product improvements informed by user data.

Team

27 cross-functional partners including devs, analysts, product managers, and operations leaders. I was Lead Product Designer with one supporting designer.

Timeline

6 Months (3 months to MVP and 3 months of post-MVP iterations)

Impact

After launching the first two iterations of the title application processing tool, we saw our teams get more done in less time. Here are the numbers:

Employee training time reduced by 3 days
Reduced rejections of title applications by 25%
Increased title application processing volume from 100 titles/day for the entire department to 95 titles/day per employee

Awards

This project, affectionately called “Get Titles Faster, OK?!” or GTFO, was recognized at Carvana’s quarterly company-wide HALO Awards ceremony. The HALO Spotlight Award is focused on celebrating a team, project, or product that drove Carvana’s business forward in a huge way. GTFO was recognized for driving customers happy through a more efficient titling process! It was an honor to lead designs on this effort.

Background

Too Much Paper

The GTFO project was launched because Carvana’s title application process was slow and inefficient. The operations team handled around 28,000 sheets of paper each day, which totals about 10.2 million sheets of paper every year. This paper came from DMVs, banks, and, most importantly, customers selling their cars.

Without good technology, it was hard to locate paperwork, prioritize work, and measure how long it took to complete tasks.

BUSINESS PROBLEM

Inefficient Technology

To get a vehicle title that says Carvana owns the car they bought from a customer, employees must take all the paperwork customers send in and review it to complete a title application. Our operations team had MacGyvered a customer relationship management (CRM) tool, Salesforce, to keep track of the process.

However, Salesforce was very expensive and the business needed a solution that would save money and increase processing speed.

User Research

Shadowing to Understand User Workflows

I flew down to Carvana’s headquartered in Arizona and shadowed five users to get a sense of their operational process and pain points, From there, I was able to map out the employee workflow to be in the in-house application I was designing to replaces Salesforce.

Iteration 1

Minimum Viable Product

After mapping out our operation team’s workflow, my task was to create a Minimum Viable Product (MVP) of an in-house application for ~100 Carvana employees. The MVP’s goal was to make the process more efficient, reduce errors, and save money.

Employee Work Log

I designed a log for each employee of all the paper title applications they completed per day. It also displayed how long it took to complete work, an important employee success metric.

Title Application Guide

To assist employees in filling out title applications, I designed a questionnaire tool that guided them through each step of the process and ensured they had all the proper information.

Usability Testing

High-fidelity prototype for usability testing

Prior to launching the MVP, I conducted usability testing with 5 employees to validate the high fidelity prototype.

Highlights:

4 out of 5 specialists noted that the MVP reduced their pain point of having to switch between screens!

Key Findings:

End of day process should be updated to allow employees to select which portion of their work record they’d like to print to send to the title vendors
Title applications that cannot be processed should be labeled more clearly.
Employees should be notified if they or someone else has already worked on a title application.

Iteration 2

Adding a Work Queue

Problem: After launching the MVP of the title processing tool, we started to see some interesting trends in user behavior. Since employees were able to choose titles to complete at random, they were cherry picking the easier work and avoiding tasks that were more involved or required more work to process. 

Goals: For this iteration my goal was to (1) reduce the number of title applications that were not completed due to complexity, (2) increase visibility of how long an employee spends on an application, and (3) improve employee performance management.

💡 Solution

The MVP only provided visibility into the title application an employee was currently completing. I designed a packet queue so users could check in up to 25 application packets at a time.

This discouraged users from cherry picking because there was a record of the work they intended to do and their leadership could see if they abandoned a task.

User Feedback

Although one of the main goals of this iteration was to help managers track employee performance, I interviewed 12 employees on packet queue design to ensure the feature was user-centered. Here are some benefits of the redesign they highlighted:

Saves them from the repetitive, clunky process of going to another system to update their packet location.
Prevents them from being stuck with the hardest packets b/c other processors took all the easy one.
Processing packets in bulk saves them time

Iteration 3

Leveling Up with AI & Automation

As we were working on optimizing the vehicle title processing tool, the state of Montana told us they were ready to test e-title flipping! E-titling involves transmitting vehicle titles electronically, making it unnecessary for Carvana employees to deal with paper documents or enter information into our in-house application manually. This was a huge time-saving, cost-saving, and paper-saving development for us.

While our engineers worked towards a fully automated e-titling solution with Montana, I designed a portal for employees to manually kick off the process.

Iteration 1

Initially, I designed a very scrappy solution for employees to use to kick off the automated e-titling process under the assumption that this page would be useless in 1-2 months as our engineers worked to enable us to submit all title information via API.

Iteration 2

However, automating Carvana’s vehicle title application process proved more complex than anticipated. Six months later, users were still using the “scrappy” portal.

So, we launched a second iteration of the portal to reduce errors and improve usability. After reviewing usage data and error rates as well as synthesizing feedback gathered during weekly employee meetings, I worked with my Products Managers to prioritize the following revisions to the automation portal.

Priority Revisions

Add ability to merge and delete documents so are only using the necessary files
Organize page in order of tasks users need to complete so it’s clear what to focus on
Reduce time users spend downloading and viewing documents in other tabs

VIEW PROTOTYPE

Reflection

Balancing employee needs with the time and cost saving business benefits of automating workflows has been an interesting challenge. A key part of my approach as a Product Designer has been involving users in the process through interviews and testing session and uncovering ways their expertise in the car buying space can be better applied.

Automation gets a bad rep from putting people out of work to robots run amok. At the same time, time and cost savings are often the driving force behind automation. My strategy has been to leverage automation to serve a higher purpose. to eliminate monotonous tasks so employees can use their creative problem-solving abilities to the fullest.

Next Steps

While our engineers are working on automating the once tedious and manual title application processing workflow, I am designing a tool that allows our employees to apply their brainpower to handling rejections. This is likely to launch in 2024. Stay tuned!