Personal Loans Repeat Member Experience
Team:
Product Designer
Product Manager
4 Software Engineers
2 QA Engineers
Product Analytics
My Role: UX, UI, visual, product design & user research
The Problem: Back in 2018, the funnel application experience for repeat users was the exact same as first time new users but with pre-filled data that wasn’t fully up to date. Only 70% of members who click on the MC banner card get past PI1 and only 67% of those who login from repeat email get past PI1.
The Objective: The goal of this project was to optimize the funnel experience for repeat members to make it faster and more convenient for these members to get a 2nd or 3rd loan.
Project Context: We launched about 8 experiments in total for both individual application experience and joint application experience and worked on this project in Q4 of 2018 and Q1/ Q2 of 2019 with multiple iterations. For this case study, we are focusing on the individual application.
Metrics:
Provide a personalized experience for repeat members
Task Success - Help them get through the application with minimal errors
Improve App_Create by 5%
Improve App_Success by 2%
Current Experience
User Problems:
"There’s no real benefit to being a 'member of the Club' because I am treated exactly like a first-time applicant."
"LendingClub doesn’t know me that well even though I have been doing business with them"
"Why am I going through the same application again?"
User Goals:
"I want a better deal with LendingClub since I’m already a member"
"I want to be recognized for being part of the club"
"I want to easily go through the application because I’ve done this before."
UX Metrics: Task Success:
How many users complete PI1and PI2?
How many users edit their personal information?
Data Insights:
Based on historical data since 2017, for repeat customers whose 1st loan is joint, majority (71%) of them has 2nd loan joint too.
When previous joint customers choose to apply joint loan again, 99% of them choose the same co-borrower.
Repeat customer’s income is not the same >85% of the time
48% of repeat apps have the same loan purpose as first loan
78% of repeat apps have the same address as the first loan app
84% of repeat apps have same number of people applying as first loan app
V.1 & V1.5 A/B Test
After creating three concepts for both individual and joint application I consolidated all the working pieces from the three concepts and created this final variant design to test. The thought process was to incorporate all the data insights and goals into the designs. We wanted to make this application a lot easier and simpler for the user to get through. For the PI2 flow, I combined living situation and phone number together and all the work information together to group them into themes. If they changed their work status then they would have to update their whole work form, if nothing changed then they just click through. The prototype shows the edit interactions.
We launched a v.1 and a v.1.5 experiment splitting up testing PI1 and PI2 separately to isolate the impact. The v.1 which was focusing on PI1 underperformed. We learned that App_Create was up slightly but not significantly mainly because the joint application rates were down since we pre-selected the loan type. The test has not reached stat sig but we took our early learnings to do a v.2. The v1.5 experiment performed well which was the updated PI2 summary screens. The Overall issuance went up 1.1% stat sig, more users converted at a higher rate when we simplified PI2 and minimized the number of steps involved in PI2.
V.2 A/B Test
We took the learnings from the v.1 and used the PI1 control flow with the new pre-fill behavior. We did not update the PI2 behavior so both variants had the same PI2 behavior so we could focus on the new pre-fill behavior in PI1. We did not pre-select the loan type since it resulted negatively in our v.1 and we updated some of their PI1 information to be more accurate.
There was a significant lift in variant A at issuance level. The pre-filling method for A is more accurate. Less users in variant A changed their address; about 16% users in Control have changed on address page but only 12% users changed on the address page. Making these updates with the PI2 summary screens resulted in a ~4.5% issuance conversion lift.
V.3 A/B Test
We mastered the PI2 flow by grouping the information together and by pre-filling their information with the most accurate data but we still felt like we could optimize PI1 even more by combining the pre-filled address and the terms of agreements.
The combination of the Address + Agreements page for variant A performed better but overall the variant A funnel performed slightly worse by .76%. Variant A underperformed because the Loan Amount + Loan Purpose screen performed worse. The experiment was turned off and the results were inconclusive. We couldn’t continue the project due to other business priorities however if we were to revist it again we would need to figure out why that screen had a high drop-off.
Final Live Repeat Experience (V.2 Design)
This new experience led to a 4.5% conversion increase and $194 million in origination and ~$11.6 million in revenue