Decreasing abandonment on vehicle financing application

The problem

DriveTime's two-page financing form had a 56% abandonment rate with 48% of users exiting on the first page alone. We also couldn't pinpoint which questions drove the most drop off because of the way the form was originally built.

DriveTime's previous financing application at the time of the experiment.

Objectives: Our primary business goal was to increase top-of-funnel conversion by reducing friction on the financing form. Our secondary objective was to identify which questions were driving the most exits.

Known obstacles: The original form was able to collect leads from either page, but the redesign would require full completion. This meant we needed an increase greater than 3.3% to offset the partial leads we would be losing.

Process

1. User research
In conversation with our core customers we learned that they had often been rejected for financing elsewhere so our approval promise drove confidence. They were also concerned with receiving too many calls and emails.

2. Competitive analysis
Exploring other financing websites showed that most were focused on conversational language and providing context to help inspire confidence with users as they moved through the form. LendingTree in particular inspired me to look into the step-by-step form.

Most financing sites had several elements in common.

3. Iterative experimentation
We ran preliminary A/B tests rearranging the form fields across the first and second page and discovered that moving the phone number field to the second page increased the total completion rate by 1.5% over the control experience.

4. Ideation and solutions
Based on our research, we focused on four key aspects in the new form design.

• Reorder the form fields: Prioritize low commitment questions first to build momentum and move phone number toward the end of the form.
• New form structure: Take a one field per page approach - allowing us to add more content, build momentum, and identify friction points.
• Provide context: Tell the user why we were asking for specific information and hopefully ease any concerns.
• Multivariate experiment: Test all possible combinations of UI elements and verbiage in order to find the best approach.

We tested a combination of progress bars, verbiage, and buttons.

Some concepts that were left on the cutting room floor.

Final prototypes

Results and next steps

The top performing variation of the new form increased completions by just over 6%, which cleared the 3.3% threshold we had set. However, when we worked with our analytics teams to dig into the data further we saw that the increase was largely due to lower quality leads. Based on this, we made the decision to move forward with the original form.

The highest performing variations.

We also took away key learnings that informed our future designs

• Progress indicators: Providing a specific number of steps was more helpful to the user than a general progress indicator.
• Communication: The content that provided the most amount of information performed the best, reinforcing our prior research.
• Drop-off points: The highest exit rates were the form entry page, second page (name), and Social Security Number. This supported our theory that momentum would help drive the user through the form, and that sensitive information like SSN drove more friction.