Capital One Dealer
Redesigning the tool behind every dealership partnership.
Overview
Overview · The moment
Before a dealership can offer a loan, someone has to add its agreement.
A small internal team does that work. It starts with an application that is printed, filled in by hand and scanned, then checked across about six separate tools. With everything in hand it takes 45 minutes. Missing pages and unanswered emails stretch that to weeks.
A new internal platform was being built to bring all of it into one place. My part was the first piece to move: the third-party tool where the team sets each dealership up.
The dealership
The agreements teamProcess
Our Design Thinking Process
Empathize
Watched the team add an agreement from start to finish, and mapped every milestone, tool and step
Define
Three findings, and a seven-action path for every agreement
Ideate
Two versions: one safe to build, one built around the team
Prototype
The team’s pick, plus the two things they asked for
Test
Think-aloud testing with all 3 lenders: 3 of 3 ready to switch
Empathize · The surprise
The team wasn’t missing features. They were working for the software.
I watched recordings of the team adding an agreement from start to finish, and mapped every milestone, tool and step. Most of what I found in my tool was small: related details split across pages, the same information typed twice.
The agreements step was different. Adding one agreement took seven actions, and a dealership needs several. The team had a name for it: “working with the system.”
Application in
Printed, filled in by hand, scanned
Missing pages and unanswered emailsCross-check
Across about six separate tools
Set up the dealership
In the third-party tool
1 · Split across pagesDetails entered together were shown apart2 · Typed twiceThe same information, entered again for a different purpose3 · Fields nobody usedOptions the team had no context forAdd agreements
A dealership needs several
7 actions, for every agreementReady for loans
The dealership can offer a loan
Define · The reframe
How might we let the team add every agreement for a dealership1 in one pass2, so the tool works for them3 and not the other way around?
What the tool had to do
Keep related details together
Details entered together, shown together.
Ask once
Information typed one time carries over.
Cut what no one used
Fields the team had no context for.
Handle agreements together
One pass for all of them, not seven actions each.
Ideate · The doubt
I didn’t know what I was allowed to break.
I walked engineering through the team’s day so they could see it too. But they still had groundwork to finish, so nobody could tell me yet what the system behind the screen could bear.
So I designed two versions. A careful one put every agreement on one page but kept the two-stage routine, close enough to the original to be easy to build. The other was the one I believed in: select everything, mark it all pending, review once, activate together.
Drawings of the two ideas, not the real screens
The question I took to the teamWould they want the version that was safe to build, or the one built around them?
Prototype · The team’s pick
Seven actions each, down to one pass for all of them.
In feedback, the team chose the version built around them, and asked for two things I had missed: a way to deactivate an agreement themselves, and a history for dealerships coming back. Both went into the prototype.
actions, repeated for every agreement
once, for all of them
Before · for every agreement
The prototype · once for all of them
Test · The proof
Sales, at the dealership
Fewer emails chasing missing details.
The dealership
Its agreement added sooner.
The buyer
A loan offer that is ready sooner.
They chose the bolder flow. Then they walked straight past the newest part of it.
In feedback, the team picked the second version for how many steps it removed, then asked for two things I had missed: a way to deactivate an agreement themselves, and a history for dealerships coming back.
In testing, all three thought aloud through real tasks. Each one skipped the faster component I had designed and went to the table that looked like the old one. They wanted the steps to change and the landmarks to stay.
In testing, thinking aloud through real tasks
Goals set for the full move, not measured results.
Solution · The answer
A drawing of the concept. The real screens are confidential.
One pass.
Select every agreement, mark them pending, review once, activate together.
The table stays.
The view the team already trusted, kept where they expect it.
It fills itself in.
Details entered once carry over, and notes take tags or free text.
Asked for by the team.
Deactivate an agreement themselves, and see a dealership’s history.
This work is under NDA, so the real screens and data are withheld.
Next Steps
The first piece of a bigger move.
My tool was the first piece to move into the new platform.
Reflection
What stayed with me.
The process is not a straight line
School gave each stage its own weeks. Here research, ideas and feedback overlapped, and each one was there for a reason.
No design is final
I used to think great designs last forever. People change, so the work changes with them.
Ideas come from people
Engineers, lenders, a researcher and other designers each changed what I made. The best ideas were never mine alone.
Not flashy, still core
No customer will ever see this tool. It is still part of how every dealership gets to offer a loan, and the people using it deserve the same care.
