Health Insurance · Self-Service Platform

Everyone thought the website was broken.
The real problem was trust.

Fewer than one in four people who started buying a policy online actually finished. The easy answer was “the site is clunky.” It wasn’t. I found what was really happening, redesigned around it, and the beta cut drop-off roughly in half.

See the results View the work
~50%
Drop-off reduced
70%+
Self-service adoption
$1.8–2.5M
Projected annual ROI
12 wks
Discovery → beta
Scroll
my-coverage.app/dashboard Overview My policies Family members Certificates Orders & history Good morning, Priya Everything’s active. 4 people covered. Monthly premium ₹4,820 Coverage status Active Health certificate Ready to download Claims & coverage timeline Covered members Priya · self Arjun · spouse Aanya · daughter Dev · son + Add a family member

The unified self-service dashboard — one place to see coverage, family, and certificates.

01 — The Trap

The agent had never just been processing forms.

They were the reassurance. That was the product, as much as the policy.

When someone buys health insurance, they’re scared of getting it wrong. Pick the wrong plan and your family isn’t covered when it matters. Miss a field and your premium is off. The human agent absorbed all of that fear — reassuring people, catching mistakes before they happened, quietly signalling “you’re doing this right.”

When the company moved the journey online, it moved the forms but not the reassurance. So people did exactly what you’d expect: they started, got anxious, and called support for the hand-holding they used to get in person. The website wasn’t failing because it was ugly. It was failing because it left people alone with a high-stakes decision.

The business read this as a UX cost. It was actually a trust cost. Every abandoned session became a support call, a data error, or a lost customer — an estimated $800K–$1.2M a year leaking out the bottom.
02 — The Insight

The brief wasn’t “make it prettier.” It was “rebuild the reassurance.”

Research made the pattern impossible to miss. Stakeholder interviews across product, ops, legal, sales and the call center. Live purchase and support calls. Usability tests on the legacy flow. A deep dig through drop-off analytics. Three findings did the heavy lifting.

Two moments caused most of the damage

Over 60% of drop-offs happened at exactly two steps: adding a family member, and uploading documents — the precise points where stakes felt highest and guidance was thinnest.

People wanted “am I safe?”, not more options

They didn’t need fourteen plan variants. They needed to know the choice in front of them actually covered their family. Progress and safety cues beat feature breadth.

The fear of certificates was about the unknown

Customers dreaded requesting a certificate because they had no idea when it would arrive. Instant, visible issuance turned out to be the single biggest unlock.

Where people abandon the purchase % of users still active at each step · legacy flow 100% Start 87% Choose plan 41% Add family 23% Upload docs 19% Review 17% Payment <25% Complete 60%+ of all drop-off
FIG 01 Drop-off by step. Two adjacent steps — add family, upload docs — account for 60%+ of all abandonment.

Who we designed for

The Family Decision-Maker (35–50, mid-income) wants reliable coverage for their dependents, fast, and is terrified of missing something. The Policy Administrator is an existing member who just wants to manage changes without phoning support.

Thinks: “Will this actually cover my family if something happens?”  Feels: anxious, rushed, skeptical of online-only.  Does: starts more than once, calls for reassurance, abandons if it gets complicated.

Design implication: transparency and safety nets aren’t nice-to-haves. They are the feature.

Two personas, one fear to design around FD Family Decision-Maker PRIMARY · 35–50, mid-income "Will this actually cover my family if something happens?" GOALReliable coverage, fast FEELSAnxious, rushed, skeptical of online-only DOESStarts twice, calls for reassurance, abandons → Drove: reassurance cues & the guided wizard PA Policy Administrator SECONDARY · existing member "I just want to manage changes without phoning support." GOALSelf-serve policy management PAINScattered visibility, slow updates NEEDOne dashboard, real-time status → Drove: the unified order dashboard
FIG 02 Personas as decision drivers — the decision-maker's fear shaped the wizard; the administrator's frustration shaped the dashboard.
STAKEHOLDER INTERVIEWS·CONTEXTUAL INQUIRY·USABILITY TESTING·FUNNEL ANALYTICS·COMPETITOR BENCHMARKS· STAKEHOLDER INTERVIEWS·CONTEXTUAL INQUIRY·USABILITY TESTING·FUNNEL ANALYTICS·COMPETITOR BENCHMARKS·
03 — The Build

Replace the agent’s reassurance with the interface — right where people used to panic.

1

Guided plan selection

Personalized recommendations and side-by-side comparison, with progressive disclosure so people aren’t drowning in options. The rest stays out of the way until needed.

2

Frictionless family addition

The worst drop-off point got the most care. A step-by-step wizard with auto-validation on age and relationship, plus a live premium preview as each person is added.

3

Instant health certificates

An online form generates a downloadable PDF on the spot, with status tracking in the dashboard. The 3–5 day wait and the branch visit both disappeared.

4

Unified order dashboard

One place to track applications, view policies and manage changes — with notifications and quick actions for renew, update and cancel.

Card sort → an IA that feels safe and reversible Objective: minimize cognitive load while making every action feel safe to take. TASKS SORTED BY USERS compare plans buy a policy add family member get certificate track order renew / cancel Users grouped by intent, not by internal product structure. RESULTING IA · FOUR CLEAR ZONES Discovery & Comparison Purchase & Enrollment Management & Certificates Orders & History Separating "buy" from "manage" let each flow stay simple — and let us add safety cues without cluttering the purchase path.
FIG 03 The card sort that shaped the IA. Four intent-based zones kept each flow simple enough to feel safe — the structural basis for the whole redesign.
BEFORE AFTER Add dependent Full legal name DOB (DD/MM/YYYY) Relation code National ID number Upload proof of relationship (PDF) ⚠ Submission failed — relation code invalid Premium impact: — (shown after submit) Document review: 3–5 business days Submit & wait No feedback · jargon · offline wait → 59% quit here STEP 2 OF 3 · ADD FAMILY Who are we covering? Priya Sharma Daughter Age 9 ✓ eligible New monthly premium ₹4,820 +₹610 / updates live ✓ You’re covered the moment payment clears Add & continue Live premium · inline checks · instant cover → +38% complete
FIG 04 The family-addition wizard, before / after — the exact step where 60% used to quit.
Rebuilding the worst step: three fidelities Each round chipped away at the anxiety that drove the 60% drop-off. LOW-FI · fewer fields Cut jargon & field count first Still no feedback during entry. MID-FI · inline validation Added live checks (age, relation) Missing: cost & coverage clarity. HIGH-FI · premium + reassurance New premium₹4,820 live ✓ Covered when payment clears Result: +38% completion at this step
FIG 05 The fidelity progression. Low-fi stripped jargon, mid-fi added validation, high-fi added the live premium and reassurance that lifted completion.
Every component answers a fear PROGRESS 2 3 Progress indicator “How much is left?” REASSURANCE You’re covered if payment clears today Confidence badge “Am I actually safe?” CONTROL Cancel this policy? Keep Confirm Action confirmation “Can I undo a mistake?” VISIBILITY Certificate ready issued instantly · download Status card “What’s happening now?”
FIG 06 The system did one job: every component answers a specific fear.

The system did one job

Every component existed to rebuild reassurance. Progress indicators so people always know where they are. Confidence badges and “you’re covered if…” language at decision points. Action confirmations so nothing feels irreversible. Status cards so waiting never feels like a black box.

BeforeAfter
Multi-session slogSingle guided session
Offline uploads and delaysReal-time validation and previews
No unified viewOne order dashboard
Constant support dependencySelf-service certificate download
04 — The Proof

The honest version of the numbers.

Straight talk: these are projections from the beta, not a year of validated results. I’m flagging which ones I actually trust, and why — because a portfolio that over-claims is worth less than one you can believe.

Numbers I stand behind

~50%
Drop-off cut in key journeys (45–55%). The number that matters most — tied to the two worst abandonment points.
+38%
Family-addition completion (35–40%). We rebuilt the worst step and watched completion climb.
70%+
Self-service adoption for certificates, removing ~80% of branch visits for that task.

Real, but softer — depends on sustained behavior

+120–150%
Online conversion lift
50%
Faster order-management time
72–78
SUS score, climbing toward “good”
If you take one number, take the drop-off cut. It’s the most defensible, and it proves the central bet: rebuild the reassurance, and people stop quitting.
05 — The Call I Had to Make

The hardest moment wasn’t a screen. It was a conflict.

My job wasn’t to choose between user trust and legal risk. It was to find the design that served both.

Legal and compliance wanted every disclaimer surfaced up front, before the customer committed to anything. Reasonable risk management, from their seat. But research was unambiguous: front-loading dense legal language was exactly what spooked anxious customers and sent them to the exit. The thing meant to protect the company was helping kill the conversion.

So I made a call. We kept every required disclosure, but moved it to the moment of relevance using progressive disclosure — plain-language summaries first, full legal text one tap away. I took that to compliance with the drop-off data and the user-fear evidence side by side, framed as risk reduction for both sides: fewer abandoned sessions means fewer error-ridden workarounds, which is its own compliance exposure. They signed off.

What I’d do differently

Two things. I’d test personalization harder — there was more conversion and SUS to win, and we left some on the table. And I’d bring legal into co-design earlier, in the room while we sketched. The compromise that took weeks could have been a thirty-minute whiteboard session in week two.

The Team

Role — UX Leader: strategy, research, interaction design, governance alignment
Team — Me + 3 senior product designers, 2 UX researchers, and partners across engineering, product, legal & operations
What I owned — Research strategy for sensitive health data, the end-to-end journey, the trust & adoption metrics, translating compliance constraints into usable requirements, and the executive case tying user trust to revenue.