Overview Problem Methods Findings Redesign Results Reflection

Case Study · 06 of 06

WhatsApp, enhanced.

Small interventions, outsized friction. A 12-week self-directed usability study that diagnosed low predictability as the through-line — and used Figma redesigns to prove the fix.

Type

Self-directed

Methods

Performance + think-aloud

Validation

Paired t-test

Duration

12 weeks

Outcome

4 of 5 tasks improved (sig.)

Heuristic evaluationPerformance testingThink-aloud protocolTask analysisFigma prototypingPaired t-testNielsen heuristics

My Role

Sole researcher and designer — study design, moderation, statistical analysis, and Figma prototyping. End-to-end ownership of a 12-week project.

Approach

Mixed-methods: heuristic evaluation surfaced candidate issues; performance testing quantified them; think-aloud protocol explained why they happened; redesigns fixed them; t-tests confirmed the improvement.

Platform

WhatsApp iOS. Six tasks covering core messaging interactions: group creation, message reactions, media sharing, status updates, voice messages, and contact search.

Problem statement and objective for the WhatsApp study, alongside a quote from an international student in Korea who struggles to find hidden features in the app.

Results

Statistically significant gains.

4/5

Tasks improved
(p < 0.05)

↓31%

Average task
completion time

↑22%

Subjective satisfaction
(1–10 scale)

The Problem

WhatsApp is widely used. That doesn't mean it's easy to use.

With over 2 billion users, WhatsApp is one of the most-used apps in the world. But ubiquity doesn't equal usability. Users adapt to friction — they work around it, forget it, or assume it's normal.

The research question: are there specific interactions where WhatsApp consistently imposes unnecessary cognitive load or time cost on users — and if so, can targeted redesigns measurably fix them?

I chose this as a self-directed study because it offered a contained, testable product — ideal for demonstrating end-to-end mixed-methods capability including quantitative validation.

WhatsApp overview: founded in 2009, acquired by Facebook in 2014, with over 2 billion users — yet held back by poor ratings tied to user experience issues.

Methods

Three-stage mixed-methods design.

Qualitative discovery → quantitative measurement → Figma redesign → statistical validation. Each stage informed the next.

Three research questions guiding the study: how WhatsApp can improve feature discoverability for international users, which hidden features users most need surfaced, and how the group chat admission process affects engagement.

Heuristic evaluation

I evaluated WhatsApp iOS against Nielsen's 10 heuristics to identify candidate usability issues. This gave the study a structured starting point and surfaced six tasks worth testing.

Performance testing · Baseline

Six participants completed six tasks on the original WhatsApp. I recorded task completion time (seconds), error rate (number of incorrect actions), and subjective satisfaction (1–10 scale). Task order was randomized to control for learning effects.

Think-aloud protocol

Concurrent think-aloud during performance testing. Participants narrated their decision-making in real time. This surfaced the "why" behind the performance data — revealing that users were slowed not by technical failure but by unpredictable outcomes from familiar gestures.

Thematic synthesis

"Low predictability" emerged as the through-line. Across all six tasks, users hesitated when they couldn't anticipate what would happen next — whether from inconsistent gesture affordances, buried navigation, or feedback that didn't match their mental model.

Figma redesigns

Five targeted redesigns addressing the highest-severity issues: improved gesture affordance signifiers, surfaced group creation flow, enhanced message reaction discoverability, clearer media attachment indicators, and visible status update entry points.

Post-redesign performance testing + paired t-test

The same six participants completed the same six tasks using prototype flows on the Figma redesigns. Paired t-tests compared baseline vs. redesign performance. Four of five tested tasks showed statistically significant improvement (p < 0.05).

Key Findings

"Low predictability" as the through-line.

Across all six tasks, the performance data and think-aloud transcripts told the same story: users weren't confused by complex features. They were slowed by small, recurring moments of unpredictability — interactions that should have been obvious but weren't.

The pattern mapped cleanly to two of Nielsen's heuristics: consistency and standards and visibility of system status. WhatsApp's gesture vocabulary wasn't broken — it was inconsistently applied.

Task-level findings

Group creation — Users looked for "New Group" in the main chat view rather than under "New Chat." Buried entry point cost ~14 seconds on average.
Message reactions — Long-press revealed reactions, but many users short-pressed first (expecting a context menu), got no feedback, and assumed reactions weren't available.
Media sharing — The attachment icon (+) didn't signal media vs. document distinction. Users opened and closed the wrong sub-menu in 60% of trials.
Status updates — Status entry was found only via the Status tab, not contextually from a contact's profile. Users expected profile-proximate access.
Voice messages — The lock-to-record gesture was unknown to 4 of 6 participants. Requires prior discovery or instruction — not intuitive.

Redesign

Five targeted interventions.

Each redesign addressed a specific predictability failure — making affordances explicit, surfacing buried entry points, and aligning system feedback with user mental models.

Group creation — surfaced entry point

Added "New Group" as a direct action in the main chat list header (alongside the search icon), eliminating the need to navigate through New Chat → New Group.

Message reactions — visible affordance

Added a subtle emoji icon on hover/long-press preview that communicates the gesture affordance before the user commits — reducing failed short-presses.

Media attachment — labeled sub-menu

Replaced the generic + icon with labeled icons (Photo, Document, Contact) visible in the attachment tray on first tap — no sub-menu navigation required.

Status update — profile-proximate entry

Added a status ring affordance to a contact's profile photo (consistent with Instagram/Snapchat mental model), with a tap-to-add-status CTA in your own profile.

Voice messages — onboarding tooltip

Added a one-time contextual tooltip explaining the lock-to-record gesture on first voice message attempt — dismissed after acknowledgment, visible to new users.

Statistical Results

The numbers confirmed the diagnosis.

Paired t-test results

Compared pre-redesign vs. post-redesign performance across all participants for each task. Four tasks reached significance at p < 0.05. The fifth (voice messages) showed improvement but did not reach significance — likely due to the small sample (n=6) and the confound of prior familiarity.

What statistical significance means here

With only 6 participants, the effect sizes needed to be large to clear the threshold — and they were. That's a meaningful signal: the usability failures were severe enough, and the redesigns targeted enough, that even a small sample returned confident results.

Impact summary slide: the finalized prototype enhanced user satisfaction by 21%, reduced task completion time and errors by 47%, and improved usability, visibility, and consistency.

A "significant improvement" without a p-value is just a feeling.

Reflection

What self-directed research teaches.

This project was a deliberate demonstration of full-cycle mixed-methods capability — from qualitative discovery to statistical validation.

The most important thing I learned wasn't about WhatsApp. It was about how the methods compound: the heuristic evaluation told me where to look; the performance testing told me how bad the problem was; the think-aloud told me why it happened; and the t-test told me whether the fix was real.

That sequence — discovery → measurement → diagnosis → intervention → validation — is the same one I bring to every client engagement. The platform changes. The discipline doesn't.

WhatsApp, enhanced.

My Role

Approach

Platform

Statistically significant gains.

WhatsApp is widely used. That doesn't mean it's easy to use.

Three-stage mixed-methods design.

Heuristic evaluation

Performance testing · Baseline

Think-aloud protocol

Thematic synthesis

Figma redesigns

Post-redesign performance testing + paired t-test

"Low predictability" as the through-line.

Task-level findings

Five targeted interventions.

Group creation — surfaced entry point

Message reactions — visible affordance

Media attachment — labeled sub-menu

Status update — profile-proximate entry

Voice messages — onboarding tooltip

The numbers confirmed the diagnosis.

Paired t-test results

What statistical significance means here

What self-directed research teaches.

Continue exploring