LAVI · UX Case Study
·
Duration: 8 weeks
·
Role: Certified UX Designer (solo)
·
Figma
VS Code
·
Section 09 of 12
09: Usability testing
Testing with real users
Two rounds of moderated usability testing were conducted, the first using mid-fi prototypes to validate navigation and content flow, the second using hi-fi mockups post-iteration. All sessions were facilitated in the participant's primary language via Zoom.
Method
Moderated remote
Zoom · Translated guide
Participants
5 users
Haitian & Hispanic community
Session length
45 minutes
Per participant
Languages
Creole & Spanish
Participant's preference
1
Symptom check, chest discomfort (Marie-Claire persona)
"You woke up this morning with some tightness in your chest. You're not sure if it's serious. Use LAVI to figure out what you should do."
SC-01 · SC-02
Marie-Claire persona
Round 1 & 2
Primary flow
2
Medication scan, new prescription (Carlos persona)
"You just picked up a new prescription. The label is confusing. Use LAVI to understand how and when to take it, then set a reminder."
MI-01 · MI-02
Carlos persona
Round 1 & 2
Primary flow
3
Find a doctor. Haitian Creole speaker (Marie-Claire persona)
"You need to find a doctor who speaks Haitian Creole and takes Medicaid near you. Use LAVI to find one and get their phone number."
FD-01
Marie-Claire persona
Round 1 only
Secondary flow
88%
Overall task completion rate across all 3 scenarios. Round 1
4.4
Average satisfaction score out of 5, post Round 1
4.8
Average satisfaction score out of 5, post Round 2 (after iterations)
100%
Of all 5 participants said they would use LAVI again
Task completion rates. Round 1
| Task |
Result |
Rate |
Avg. time |
| Symptom check, complete flow to urgency result |
5 / 5 |
|
2m 14s |
| Medication scan, scan label and set reminder |
4 / 5 |
|
3m 42s |
| Find a doctor, filter by language and insurance |
4 / 5 |
|
4m 01s |
| Save result and share via WhatsApp |
5 / 5 |
|
0m 38s |
| Change language mid-session |
3 / 5 |
|
1m 52s |
2 of 5 participants could not find the language toggle when asked to switch languages after the initial onboarding screen. Both participants navigated to Settings first. The toggle was only visible in the home screen header and disappeared as users moved deeper into feature flows, a critical discoverability failure for an app whose core promise is language access.
Recommendation → implemented in hi-fi: The language switcher was moved to a persistent global location in the top navigation bar, accessible from every screen regardless of depth. ✓ Implemented
One participant navigated directly to the medication search field and typed the medication name rather than using the camera scanner. When prompted to try the scan feature, the participant said: "I didn't see the camera. I thought this was like Google." The camera icon was undersized and visually equal weight to the search field, providing insufficient affordance.
Recommendation → implemented in hi-fi: The camera icon was replaced with a full-width illustrated "Scan your prescription label" button as the primary CTA. "Search by name instead" was demoted to a text link below, reversing the visual hierarchy. Post-iteration completion rate for this task improved from 80% to 100%. ✓ Implemented
The insurance filter label "Medicaid / Medicare" caused hesitation in one participant who said: "Is Medicaid the one for poor people? I don't want to say that." The clinical and bureaucratic framing created emotional friction around a culturally sensitive topic, discouraging users from selecting the filter they actually needed.
Recommendation → implemented in hi-fi: Filter label changed from "Medicaid / Medicare" to "Free / low-cost coverage" with an expandable tooltip explaining what each option includes. Language that reduces stigma without sacrificing clarity. ✓ Implemented
In a modified test scenario where the urgency result was set to red, 3 of 5 participants expressed heightened anxiety at the color and bold text before reading the actual recommendation. One participant said: "This is scaring me more than helping me." The visual language was alarming users who were already in a vulnerable emotional state, directly contradicting LAVI's core design principle of calm over alarm.
Recommendation → implemented in hi-fi: Red urgency band replaced with warm amber. A calm opener sentence was added before the ER instruction. Two clear action buttons (Call 911, Find nearest ER) replaced the paragraph of instructions. Bold text was reduced throughout. ✓ Implemented
All 5 participants immediately recognized and positively responded to the WhatsApp share button on the result screen. Three participants voluntarily said "I would send this to my family" without any prompt from the facilitator. This validates the core research finding that WhatsApp is the primary health information channel in both communities, and confirms that making sharing a primary CTA was the right decision.
Validation confirmed, no change required: The WhatsApp share button remained as a primary CTA on all result screens in the hi-fi. Its prominent placement was validated by this finding and retained through developer handoff.
Design iterations after Round 1
Before. F1
Language toggle only visible in the home screen header. Disappears on all deeper screens, users navigating into features lose access to it entirely.
After. F1
Persistent language flag icon added to the global top navigation bar. Accessible from every screen at every depth. Language switcher discovery improved from 60% to 100% in Round 2.
Before. F2
Small camera icon placed beside the text search field with equal visual weight. Insufficient affordance, users defaulted to typing rather than scanning.
After. F2
Full-width illustrated "Scan label" button became the primary CTA. "Search by name instead" demoted to a text link below. Medication scan task completion improved from 80% to 100%.
Before. F3
Insurance filter labeled "Medicaid / Medicare", clinical and bureaucratic language that created stigma-related hesitation in participants.
After. F3
Relabeled "Free / low-cost coverage" with an expandable tooltip. Neutral, descriptive language that reduces emotional friction without sacrificing accuracy.
Before. F4
Red urgency band with bold "GO TO ER NOW" copy, alarming before users had read a word. Contradicted LAVI's core principle of calm over alarm.
After. F4
Warm amber band. Calm opener sentence before the ER instruction. Two clear action buttons (Call 911, Find ER). Bold text reduced throughout. De-escalation by design.
Round 2 testing outcome
Post-iteration re-test with 3 of the original 5 participants (the remaining 2 were unavailable due to scheduling) showed measurable improvement across all iterated areas: language switcher discovery improved from 60% to 100%, medication scan completion from 80% to 100%, and average satisfaction score rose from 4.4 to 4.8 out of 5.
Improvements driven by four specific iterations: persistent language toggle (F1), full-width scan button (F2), stigma-free insurance label (F3), and calm amber urgency state (F4). Finding F5 required no change. WhatsApp share placement was validated and retained.