Case study - Member experience

Empowering people to make smarter health insurance decisions

A self-guided Open Enrollment experience for web and app, built by turning advocate-only content into something members could use alone. I led it end to end - a 114-person survey, a 12-participant usability study, banner and content-state logic across 100+ employer customers, and nine pieces of embedded education. V1 launched in October 2024; 77% of members rated the information clear, and the inline definitions were the single most-praised feature.

My role
Lead product designer - discovery, IA, prototyping, usability testing, UI
Team
Product manager, content team, engineering, advocacy operations SMEs
Timeline
July – November 2024, research to launch; follow-on redesign scoped April 2025
Tools
Figma, Confluence, UserTesting
Open Enrollment home - what members must do, and by when

Overview

Context · Problem · Constraints

01

Context

An advocacy model that buckles once a year, on a predictable schedule.

The company's core offering is health benefits advocacy: a member calls or messages, and a care advocate helps them navigate their benefits. During Open Enrollment - the annual window when employees pick a health plan for the coming year - that model buckles.

The volume is structural, not incidental. We served well over a hundred employer customers, each with its own OE window staggered from mid-September through December, each with its own plan lineup, enrollment rules and support arrangement. One customer's window ran nine weeks; another's ran three. Every one of those windows generated a spike of calls and messages asking the same handful of things.

We already had the content. It lived in the internal advocate tool - structured data from our content platform, rendered for advocates: level of support, key dates, what's changing, ID cards, how to enroll, plus a plan comparison tool and coaching material.

The content was written for the person answering the question, not the person asking it.

It contains instructions like "do not share details with the member until the external date" and assumes the reader already knows what an out-of-pocket maximum is. Handing it directly to members was not an option.

The source experience - advocate-facing Annual Enrollment page

02

The problem

Members had no way to answer their own Open Enrollment questions, and the only existing content was written for the person answering them.

01Members don't understand the system they're asked to shop in. Not the plans - the machinery. How a deductible relates to an out-of-pocket maximum, what happens at an out-of-network provider, why a plan type changes which doctors you can see.
02The content model was built for a human intermediary. Sections arrive curated, or flagged Under Investigation, Not Applicable, No Change. An advocate reads a status flag and translates. A member reading "Under Investigation" gets nothing.
03We could surface the information but not the action. Members enroll through their employer's own system. Designing a research tool that deliberately stops short of the task the user came to do turned out to be the defining interaction problem.

03

Constraints

Fixed before design started, and they shaped the outcome more than any aesthetic decision.

No premiumsPremium data wasn't ingested on the platform. Research would go on to show premium cost was the single most important piece of information to members - and we could not show it.
No enrollmentEnrollment had to redirect to the customer's benefits system. No delegate or dependent view; new hires and qualifying life events out of scope.
Partial dataNo pharmacy-level coverage comparison. Content completeness varies per customer, per section, and changes during the OE season as customers finalise details.
Conditional by customerFull, partial and unsupported customers each needed different content - with no OE banner at all for unsupported ones. Some customers have multiple populations with different plan sets, so members must identify their own group before anything can render.
TimelineDevelopment had to start immediately to hit the autumn OE season, which constrained research to one fast usability round.

Discovery

Research · Synthesis

04

Research

Prior workshop material, benefits-administrator interviews, market review, and a 114-person member survey.

What benefits administrators told us
~75%enroll in the high-deductible plan, largely because the financial-account benefits are attached to it
~16%take no action at all during OE; most people re-pick the same plan unless something in their life changed
Logisticsthe most common questions weren't about plan selection at all - where is my medical card, how do I contact my vendor

Employees don't engage with the resources already available internally, and ben-admin teams deliberately don't tell anyone which plan to pick. Published benefits research pointed the same way: roughly half of employees don't fully understand elements of their package, and among those who regretted their choices, 60% blamed a lack of understanding or information.

Member survey - n = 114, representative of our member population

The goal was to separate two things that get conflated: what members consider important, and what they find hard.

Most importantPremium cost - ranked first by 58 of 114 participants. Then deductible and out-of-pocket maximum, then coverage for regular provider visits and medications. Least important: anticipated life events, financial accounts, plan type.
Most challengingWhat costs are paid out of pocket, what services and procedures are covered, prescription coverage and terminology. Ranked least challenging: the difference between plan types like PPO and HMO.

89% were somewhat or extremely satisfied with their previous OE - but of the 56 "somewhat satisfied", 14 described a lack of clarity between offerings; of the 46 "extremely satisfied", 11 gave that rating because of education their employer provided: webinars, calculators, guided selection tools. 72% re-enrolled in the same plan as the prior year.

Satisfaction wasn't tracking plan quality. It was tracking whether someone had explained things.

05

Synthesis

Two findings in direct tension. Resolving them became the brief.

The most important information was missing, and its substitute was the most confusing

Premium was the #1 ranked need and unavailable. The nearest things we could show - deductible and out-of-pocket maximum - were also the things members told us were hardest to understand.

Members rank plan type as least challenging - and they're wrong

Self-reported confidence was high. Observed comprehension, once we got to testing, was not: several participants read "plan type" as medical vs. dental, not HMO vs. PPO. A member who thinks they understand plan types won't go looking for an explanation - which ruled out a guidance section members would have to choose to enter.

Design response: education embedded at the point of confusion. Every term in a comparison table becomes its own entry point. The member doesn't decide to learn; they tap a word they don't recognise while doing something else.

V1 goals
01Say clearly when OE is happening and what members must do
02Let members see and compare every plan available to them
03Answer service-level questions - how would knee surgery be covered?
04Teach the underlying concepts inline, wherever a member hits one
05Be explicit that enrollment happens elsewhere

The educational set was scoped to nine pieces: health insurance & OE, coverage, cost sharing, networks, plan types, financial accounts, who is covered, places of service, and FAQs on adding or removing people.

Design

Exploration · Iteration · Testing

06

Exploration & process

Most of the design work was negotiating with the content model.

Three overlapping views of the same plan data

Available elections (what exists, including how plans bundle - electing a medical plan usually auto-enrolls behavioral and pharmacy), plan comparison (side by side on deductible, OOPM, networks, financial accounts), and service coverage comparison (how one service is covered across all plans, by place of service and network). I explored these as tabs in a single Elections section, as separate destinations, and as one long scrolling page. Tabs won for the prototype, with plan-bundle logic explained in a dedicated guidance sheet rather than added table complexity.

The banner system

The banner is the only entry point most members would ever see, so it needed its own exploration. It had to express pre-OE, active and post-OE; single versus multiple populations; and full, partial or no support - with no banner at all for unsupported customers.

Banner states - starting soon, in progress, ended

Designing for content that isn't there yet

Every section needed to degrade gracefully, because completeness varied by customer and shifted through the season. I designed explicit states for missing plan information and unavailable section content rather than hiding sections silently - a hidden section leaves a member wondering whether they missed something.

The population picker was a similar negotiation. When a customer has multiple enrollment groups nothing can render until the member identifies theirs - but members often don't know which group they're in. The picker got a "Which group should I pick?" affordance, populated from the population descriptions in the content platform and shown only when at least one description had been filled in.

Population selection - picker and the "which group?" sheet

Collaboration

The nine educational pages were drafted against member language, then reviewed against clinical and compliance requirements, with a standing disclaimer clarifying that examples are illustrative. With engineering, most of my specification effort went into conditional logic rather than visual detail: which banner appears for which combination of period and support level, how plan bundles sort, when the population picker appears, and what each section shows for every content status.

07

Usability testing

Unmoderated, August 2024 - 12 participants, six web and six app, all of whom had enrolled through an employer's OE the prior year.

Tasks were built around a scenario with real stakes: you're planning knee surgery next year and need to work out which plan covers it best - including if your surgeon is out of network.

What worked immediately

Every web participant found and used the primary CTA into plan comparison. Plan comparison itself was the most valuable screen on both platforms - one participant described the linked table as far better than mentally integrating several static tables from prior years. Every app participant located and understood how to enroll.

What broke
The out-of-network filterClose to undiscoverable on mobile. The first two app participants couldn’t complete the task, so I changed the prototype mid-study to pin the filter and its feedback to the top. It helped - but only two of six app participants ever filtered by out-of-network. Several tapped the plan-type dropdown expecting network options to live there.
Coverage comparisonMultiple web participants answered coverage questions from the top summary table, comparing deductibles, without noticing the service-level table below it. Some found it several minutes in.
Highlighting backfiredI had highlighted the member’s current plan in green to orient them. Participants read green as recommended - the best option.
Tabs went unnoticedSeveral web participants missed the tab navigation inside Elections entirely; one clicked around for a long stretch, then answered every question correctly once found.
A filter read as an actionAsked where they’d add a spouse, two app participants went to the coverage-tier filter and tried to change it - treating a display control as the mechanism for changing their actual coverage.
Members expected to enrollMost app participants assumed they could select and enroll in a plan in the app. One wanted the "online" badge on the How to Enroll screen to be a working button into the enrollment portal.
One useful non-finding

Participants reached the same guidance through completely different routes - tapping a term, using the "Understanding…" button, or navigating to the guidance section. Same for navigation: some returned to the OE landing page, some used the main nav, some the sidebar crosslinks. Rather than consolidating on one path, I kept all of them. Redundancy was doing real work for different navigational habits.

Plan comparison - tested version, then two iterations

Beyond the August round, I built measurement into the release itself: an OE feedback form embedded in the experience - reason for visit, five-point scales on clarity, completeness and accuracy, and one open field - so we could validate against real members mid-season rather than waiting for a post-mortem.

Solution

Three pillars

08

Entry and orientation

Banners on the benefits home surface the period state - starting soon, in progress, ended - with days remaining and dates, tailored by support level and suppressed entirely for unsupported customers. The OE home page carries key dates, enrollment group, what the service can and cannot help with, available elections, and FAQ entry points.

09

Plan shopping

Three views of the same plan set - plan comparison as the default, service coverage comparison, and individual plan detail - filtered by zip code, coverage tier and network. Plan detail includes a current-year versus next-year snapshot.

Plan and elections information

10

Embedded education

Every term, plan type and place of service opens its own definition sheet in context - PPO, HMO, HDHP, EPO, POS; inpatient, outpatient, office, freestanding facility, telehealth audio and video. Behind those sheets sits a full guidance section: health insurance and OE basics, coverage, cost sharing, networks, plan types, financial accounts, who is covered, and FAQs on ID cards, premiums, and adding or removing people.

FAQ and educational guidance

Impact

Outcomes · Reflection · Next

Outcomes & impact

V1 launched in October 2024. The embedded feedback form collected 190 responses across three customers in the first weeks.

77% found the information clearVery or somewhat clear, half of them very clear - averaging 4.11 on a five-point scale. Completeness: 78%, averaging 3.89.
Education was the top reason for visiting57% came to educate themselves about OE and insurance - ahead of enrolling (44%), premiums (30%), plan options (28%) and basic logistics (25%). That validated the central bet: members weren't only looking for data, they were looking to be taught.
Inline definitions were the most-praised featureUnprompted, in open responses - exactly the mechanism designed to reach members who wouldn't seek out a glossary.
The comparison table reduced real effortDescribed as easy to digest against reconciling static documents, with several members noting the ordering from most to least important information.

11

What didn't land

44% of members arrived intending to enroll, and couldn'tThe largest single source of frustration. Members didn't read the boundary as a boundary - they read it as breakage, several assuming something was broken with the app or the content.

"It just kept going in a circle. I never found a way to confirm my choices for coverage."

30% came looking for premiums we didn't havePredicted by the survey, unavoidable given the data constraint, and confirmed at scale. Content and technical issues surfaced too - slow initial loads, nav items appearing only after refresh, and some inaccurate dates from a pilot population.

12

Reflection

Self-reported difficulty is not difficultyMembers ranked plan types least challenging; in testing several couldn't distinguish a plan type from a benefit category. The survey told me what members would seek out; testing told me what they'd stumble on. Designing only for the first would have missed the second entirely.
A boundary announced at the end reads as a failureI treated "you can't enroll here" as an information problem, and testing confirmed everyone could find the instructions. But good instructions at the end of a research journey didn't satisfy the 44% who came to enroll - it frustrated them. The constraint needed framing at the entry point, not resolving at the exit. That's the correction I'd make first.
Redundant navigation was right, and I nearly optimised it awayWatching participants reach identical content through three unrelated routes looks like unclear IA. It wasn't. Consolidating on the cleanest path would have stranded people whose instincts differ.
I fixed a presentation symptom, not the causeMaking the filter sticky mid-study was the right call on the timeline, but it muddied the evidence and the filter still failed for four of six app participants. Stickiness addressed visibility, not findability - members were looking in the plan-type dropdown entirely. That's an IA problem. Relatedly: compression on mobile is not simplification. The right question isn't how much scrolling is acceptable, it's what is the minimum a member needs on a phone.

13

What's next

PremiumsA standalone initiative, since the data isn’t ingested. Two independent signals point at it: #1 ranked need pre-launch, and 30% of members came looking for it post-launch.
Reframing the boundarySetting the "you enroll elsewhere" expectation at entry rather than exit - the highest-value fix available from the launch data.
Simplifying mobileDetermining the genuine minimum viable set for a phone rather than reflowing the desktop IA.
Pharmacy comparisonStill blocked on data suitability, and the largest known gap in service-level comparison.