WithClutch

WithClutch is a startup in California on a mission to transform Credit Unions into true fintechs. For this, it uses a series of services and APIs.

Given the track record of Clutch's founders, the company just started with the mission of being a vehicle refinancing system in the USA.

There, many people take out financing at the point of sale and end up paying much more than they should for financing. Impulse purchase.

And there, there are two types of financial institutions: Banks and Credit Unions.

Banks dominate the market. But they usually have higher interest rates.

CUs still use Jurassic systems for customer acquisition and retention and internet banking. But they have much lower rates, because they have some tax exemptions and because they are not for profit.

Example of Credit Unions’ popular LOS (Loan Origination System)

The idea was to use the vehicle refinancing system with the CUs and help them become true fintechs.

And because each CU has very different ways to operate, Clutch has made the choice of being a white label system.

So when I joined, Clutch was just that, an auto refinance fintech.

This was Clutch's interface when I first joined. It was improved from the vendor's previously shown but it still had a lot of issues.

Explaining the flow

1. The first artboard is a fictional CU and then we have the flow where:
2. We ask for their phone number
3. We send them a verification code
4. We ask them for the last 4 digits of their Social Security Number
5. We ask them to confirm their information
6. And we show them offers of refinance

Services like Cognito, Experian (and other credit score checking agencies) are able to assemble a comprehensive map of the user's financial situation. Something Credit Unions never had.

But as you can see there are still a lot of things that need improvement here. The bounce rate was huge at above 80%, and bear in mind that these are users with high intent. I think many of you, by looking at this screen can guess why the bounce rate was that big…

A UX audit here easily identified:

But this first page was the most important to us. We wanted to reduce the bounce rate.

So, with this audit in place we:

1. Improved the UX copy
2. We inserted the branded header
3. We divided the form into information that we already knew and information that we didn't know.

Click to see new prototype

The bounce rate simply reverted. We were using ecommerce bounce rates as a reference. Our goal was a 30% bounce rate, but these simple UI changes turned them to 22% percent which was pretty good. Ok, but that was the easy part.

Let's skip into the future

So we were already getting data from users that could get us to build plenty of other products. CUs loved what we gave them. Because, not only our UX was better, but we also provided them with important documents that they used to generate manually and now we just handed that over to them.

We could build more products and the idea is that we had more mice in the maze so we could have more accuracy in our experiments.

So, together with the multidisciplinary team we did a prioritization exercise. Often this was done in an excel file too. And this is not the real matrix we used.

Products that CUs usually offered.

Products prioritized were:

• New auto loan
• New personal loans
• Personal loans refinance

So we already knew what we wanted to build. We already had more engineers. While engineering was preparing the infrastructure, database, APIs, etc, we were always ahead:

• Performing usability tests using user brain
• We were using Twillio to find users to talk to and talking to them
• We were prioritizing features within these products.

So this is what we had after a couple of weeks:

Let's fast forward into the future one more time.

Well, we already had more mice in the maze. Our UX was already pretty mature. We had a design system in place. Now there were a couple of metrics that we were not happy with and that we knew we could help CUs with:

• Share of wallet
• Dropoff rates

Share of wallet

At the time we worked with over 17+ Credit Unions with a combined 2.5 million Membership base. So we did a soft pull from over 9,000 applicants to see the share of wallet among various common products.

When looking at debt categories, we get an idea of what percentage of Members have a particular type of debt. Within these credit files we see nearly all Members have a credit card or auto loan.

The average Credit Union Member has $139k in debt and this is the share of wallet of the CUs with their members.

So that's one business metric we wanted to improve to each loan category.

Drop-off rates

In order to do this, we also needed to understand where the dropoffs occurred the most and it was pretty obvious. As usual, it was at the top of the funnel.

It's important to point out a particular characteristic of loan applications. There are desired dropoffs. These are candidates who would never have been approved. If they don’t complete the application, this is net beneficial for the Credit Union since many LOS systems charge by application, and since an application that won’t fund is bad use of labor.

The Loan Portal

So the solutions for these are many. Together with Product Management we decided that what  we needed to measure was the recapture rate. Meaning, the percentage of people that comes in for one loan product and ends up with another.

As we saw in the beginning most loan applications end up being unnecessarily complicated and lengthy without giving the user any reward or indication of approval. Rather than requiring multiple forms the idea was:

• Reduce the burden and eliminate fields that don’t influence decision making
• Take the user to the “what could my loan look like” part sooner to prevent abandons
• Collect as much data as possible via APIs and external services.

Enters the Loan Portal. Please note that the following prototype was built for a usability test with specific tasks and therefore only works within the tasks’ happy path. Click the image to see prototype.

Ps: First screen is not part of the product.

Figma file

Results

The final Loan Portal was not achieved in one iteration. It had several states of feasible blocks until it got where it got. But the results were amazing.

We measured the share of wallet and recapture rate by re-running a script through the soft pull after one month. We also used CU's own figures statement to get to this result.