01 · The bet
A loan marketplace that only works for prime borrowers is not a loan marketplace
Credit Karma saw secured loans as an underserved market. There was no good place for people to connect with these products, and that gap was exactly the point: the company wanted to be the loan marketplace for everyone, not just the prime segment that every other marketplace already served well.
An underserved product is usually underserved for a reason, though. In this case a large part of the reason was that almost nobody knew what a secured loan was.
02 · Problem one, comprehension
Users did not register that we were showing them secured loans, or what one was
The marketplace presented offers as cards, and secured loan offers sat in that same stream styled like everything else. That produced two failures at once. Users did not notice a different kind of product was on offer, and if they did notice, they had no idea what made it different or why it might work for them when an unsecured loan would not.
This was a comprehension problem, not a persuasion problem. The two need opposite treatments.
Persuasion problems get sharper copy and stronger calls to action. Comprehension problems get structure: name the thing, show what it requires, and make the tradeoff visible before someone commits to it. So I designed several treatments of the secured loan card that varied how explicitly the product was named, how the collateral requirement was surfaced, and how far the card visually separated itself from a standard offer.
Those went into unmoderated usability tests on UserTesting.com, deliberately recruited across people who knew Credit Karma and people who did not. The mix was the point. Familiarity hides comprehension failures, and the users most likely to be confused by a secured loan are the least likely to be power users.
The two strongest treatments then went into an in-production A/B test against the standard card design.
37% more users continued through to the partner's lead generation from the secured loan card than from the standard card. It is worth being precise about what that number is, because it is the question an interviewer should ask. It is not offer impressions and it is not funded loans. It is the step where a user commits to going to a lender, measured against a live control running in the same product at the same time. That is the exact step the card was redesigned to affect.
03 · Problem two, friction
People were dropping out when we asked for their car
Pre-qualifying for a secured loan requires the collateral, which in practice meant asking users to enter their vehicle information partway through the application. That is where the flow bled users, for two reasons that need different answers: some people did not have the details on hand, and some felt it was too much to ask at that point in the process.
The second group is a design problem you can work with sequencing and framing. The first is not. If the information is not physically in front of someone, no amount of clarity gets it out of them.
So the answer was not in the loans vertical at all. Credit Karma already held vehicle information for any user who had ever used the platform to compare auto insurance rates. It was sitting in another vertical, in another team's system, unused by us.
We connected the two so the application could pre-fill a user's auto information from their insurance history. For those users, the hardest step in the flow stopped being a data entry task and became a confirmation. Pre-qualification completion rose 15%.
The design was the smaller half of this job. Most of it was internal: building relationships across verticals, making the case for why the loans funnel deserved insurance data, and getting the work onto another team's backend roadmap. None of that shows up in a Figma file, and it is the reason the number exists.
04 · Constraints
Every offer had a contract behind it
This was a marketplace, so the design was never accountable only to the user. Compliance, lender contracts and eligibility rules all had a vote.
Disclosure placement is the clearest example. Some lenders required specific disclosures at the top of an offer. For others, the same information could sit further down. The requirements varied lender by lender and they were contractual, not stylistic.
That rules out a single fixed layout. Designing to the strictest partner would have made every other offer worse. Designing to the average would have broken the strict ones. The work was building an offer presentation that could flex to each partner's requirements and still read as one coherent product, while Credit Karma stayed committed to surfacing the best available offer for the user within its lender partnerships.
05 · The front door
A landing experience where no single vertical was the main character
Alongside the secured loans work I contributed to a redesign of the Credit Karma app's landing experience. The old front door put one vertical in the foreground. The new one showed users the best available options across all of them: mortgages, credit cards, insurance and personal loans.
This was a large group effort with every vertical contributing to a shared vision, and I want to be accurate about my part in it. I was one of many contributors, not the owner. What I took from it was a harder version of the marketplace problem I was already working on: when every vertical is somebody's priority, relevance has to do the work that hierarchy used to do.
06 · How the work got validated
Research was not optional, and I did not sign off on my own work
Every design at Credit Karma had to be vetted through research before it moved forward. That is a good culture to learn in, because it keeps separating the designs you like from the designs that work.
My preferred method was unmoderated testing on platforms like UserTesting.com. Moderated sessions go deeper, but unmoderated let me reach a much wider range of people quickly, including plenty who had never touched Credit Karma. For a usability question, and especially for a comprehension question like the secured loan card, that breadth mattered more than depth.
On ownership: I worked under a Staff Designer who guided my design decisions, helped me build relationships with PMs, lender partners and engineers, validated my work, and gave final sign-off before anything went for development approval. The secured loan work was mine to do and theirs to approve. Both of those are true, and the second is a large part of why the first one landed.