Stride is a platform for all self-employed workers (SEs) to obtain health insurance.
Users with lower income qualify for subsidies, which reduces the price of their health insurance plan. For Stride, a subsidy eligible member was 4X more likely to convert.
I worked in a team of 2 PMs and 4 engineers, and involved in user research, wireframing, prototyping, visual design, QA and implementation.
With the redesign, we got more users subsidies, lowering their health insurance cost. This project had the highest bump in our E2E conversion in Stride history.
INCREASE IN USERS WHO CHECKED IF THEY WERE ELIGIBLE
INCREASE IN SUBSIDY ELIGIBLE USERS
INCREASE IN E2E CONVERSION
I started the project off with user research and identified 3 areas of improvement. I wireframed multiple prototypes to test those out. After generating confidence around our solutions, I designed those in high fidelity and worked with engineering to make sure they were implemented.
Anyone who makes less than 4X the Federal Poverty Level is eligible for subsidies and pays a lower price for their health insurance. We knew that many of our members who were under that limit didn't bother checking what their subsidy was, and therefore were paying too much for their health insurance.
We also knew that a Stride user who was eligible for a subsidy was 4X more likely to purchase through Stride.
Before jumping right into the design process, it was important to define clear success metrics. I worked with the PMs, Head of Product, and engineers to ensure we were aligned on what was success for this project.
I worked with one of our PMs, Emily in this phase. We talked to Stride users to understand their mental models on income, and why they weren't checking to see if they were eligible for subsidies.
SEs don't get a steady paycheck at the end of the month. Their income can be highly volatile, seasonal and comes from multiple sources that changes throughout the year. So it becomes difficult for them to predict how much money they will be making for the year.
Majority of users assumed that subsidies were "like Medicare", and that they were only available for lower-income individuals. Users thought they made too much money, so never bothered checking.
This findind was also backed by scientific evidence that suggested that most Americans think they're middle class, regardless of whether they made 20K or 200K.
Users didn't really understand how income affected the price of their health insurance. They had a slight idea, but didn't understand that a few thousand dollar discrepancy could make their health insurance cost a couple hundred dollar less each month.
Based on the problems identified, we generated multiple hypotheses and solution for each of the 3 problems. I then picked the top solutions and prototyped them in low fidelity, and did 3 rounds of low fidelity testing, with 15 users.
In qualitative user research, it's really hard to tease our an exact, well defined problem. Users often don't know what they want, and might not always articulate it. Low fidelity prototypes allowed me to put concepts in front of users that they could react to. This phase helped me narrow down solutions, but it also helped me tease out each problem further.
Following are the 3 problems from above and the proposed solution to each one:
It's hard for SEs to estimate their income
Guided Income: Detailed income calculator that let users input their income by job
Users assume they made too much money
Income thresholds:Giving an indication of what the exact cutoff is (based on each individual's unique situation)
Link between income and health insurance cost was not understood
Dynamic feedback loops:In-page subsidy calculator that updated as users typed in their income
Here I got into mapping out the detailed flow diagrams for each solution, and the interaction of elements and screens.
Now instead of dumping everyone straight into estimating their yearly income, the new solution offered users a choice - to estimate on their own or to use our help.
The second option allowed users to input their income by job.
In our testing, we found that users were able very quickly say if they made above or below a number we gave them. Instead of asking for permission, I designed a new onboarding flow so that the users weren't asked to check if they qualified, but rather, they were given a number to anchor on.
Anyone who answered that they were below the threshold went through the subsidy checking flow, and anyone who was above the threshold got to skip that step.
In the beginning, I tried using copy to explain the relationship between income and plan prices. But most users never read it. So I designed little feedback loops on the income entry screen that would autocalculate users' subsidy and the their plan price.
This solution aimed to visually and directly demonstrate the link between income and health insurance prices.
Here are the results from the launch of this project.
Feature 1 helped users estimate their income by breaking it down by job. This feature actually decreased onboarding conversion by 10%. When I conducted usability tests post implementation, users cited that they didn't really care about accuracy at this stage. They were often just shopping around, and while they needed help estimating, they didn't care about accuracy.
We as a business got too excited about income accuracy, but our users did not. We over designed for a feature we care about, and strayed away from user needs.
The goal of this project was also to help more users get subsidies, but we got distracted by trying to make income more accurate.
Feature 2 provided users an anchor point/threshold income before users checked for subsidy qualification. This feature was a rousing success! We went from 30% total subsidy-eligible users to total 45% subsidy eligible users.
Our onboarding conversion did not take a hit as a result of this feature.
This also had positive effects downstream. Our E2E conversion increased by 33% as subsidy eligible users were much more likely to convert than ineligible users.
Feature 3 provided instant subsidy calculation as a user was typing in their income. It reduced the looping behavior we observed in some users.
It also had downstream effects, where users were less likely to change their income later on, after they picked a plan and were filling out their plan application.
This project was one of the biggest and most successful to bring about a step change to conversion rates at Stride.
While the Guided Income feature was not as successful as we had hoped, it taught us to not get too sidetracked by what management wanted, and to keep the focus on grounded user needs.
Guided Income was overkill for a health insurance onboarding. The research done on the feature also laid out the foundation for what is our new product, Stride Finance.