Case Study

NaviLearn Case Study

Free learning, finally within everyone's reach.

Crystal Richardson — UX Designer & Researcher

Welcome to NaviLearn screen
My Learning Path screen
Ask Navi screen

Outcomes · design scope and study, not live business results

9 → 21 screens

Every screen redrawn, in light and dark

2.88:1 → 5.12:1

Contrast on the Completed Check badge

4 → 3 doors

Bottom menu: Home, Offline, Settings

1 → 3 questions

Knowledge Check, in 3 formats

7 languages

All in one place, in Settings

4 · 15 min · 2

Participants, minutes per session, rounds

Quick Glance

Problem
Adults who are new to technology get locked out of online learning by complicated software — not by ability. My research found the real barriers were time, eyesight, fear of losing their work, and needing to learn without a connection.
My role
UX designer and researcher. I owned the research, personas, wireframes, design system, prototyping, and usability testing — and I designed the November 2026 upgrade that turns the guide into an AI coach.
The product
A free learning app for non-tech adults, built around a personal AI guide, Navi, who finds lessons, reads aloud, remembers what tripped you yesterday, and coaches — in the learner's own language.
The build
21 screens, each in light and dark, at 440×956 (iPhone 18 Pro Max), one locked design system across all of them, fully prototyped — every button, chip, toggle, and picker works, with pressed and unpressed states.
Study
Usability study in my C7 case-study deck: 4 participants, 15-minute sessions, two rounds.
Also decided
Language lives in exactly one place, in seven languages: one Settings row opens a picker with English, Spanish, French, Cantonese, Russian, Arabic, and Yoruba — an overlay over the dimmed screen, because an inline panel provably did not fit the 440×956 frame. Navi speaks the learner's language first — prompts, feedback, and errors.
The study material had to be real, in more than one shape: three worked math questions in three formats across three categories — a fractions warm-up retry, a number sentence (24 ÷ ___ = 6), and a word problem — plus two real Read With Me passages, Everyday English: At the Doctor and Fractions in the Kitchen.
Try the prototype (opens in a new tab)
Your browser does not support video playback.
The flow in order, switching between light and dark.

The person behind the decisions

Who I designed for

Carlo Rivera, 55 — a part-time retail worker and ESL student — is my anchor persona. He wants learning content in his own language, with audio. His pain points drove the design: language barriers, visual strain, information overload, and not knowing which sources to trust. Around him sit Margaret, who needs the magnifier, Priscilla, and the rest of the persona set from my research.

Every decision below was judged against one question: could Carlo do this unaided, in Spanish, on a cheap phone, possibly offline?

Frame A

Paper wireframe — Frame A: welcome screen with goal input and AI suggestions

Frame B

Paper wireframe — Frame B: curated feed with saveable lessons and tailored recommendations

Decision 01

I made the guide proactive instead of waiting to be asked

60-second reset

A tutor that waits in a chat window gets ignored by exactly the learners who need it most — people who don't yet know what to ask. My research deck's Phase 2 findings on emotion-aware AI and digital mindfulness pointed the same way.

So Navi speaks first. On My Learning Path, Navi picks the next step. When a session runs long, Navi checks in with a Brain Break offer — “It looks like things are getting a bit intense. Shall we take a 60-second reset together?” — using my research deck's own words. One tap to breathe, one tap to dismiss. The offer is a moment in the flow, not a screen buried in a menu.

The measure I hold Navi to: not whether the learner chatted, but whether they can do the next thing unaided.

Brain Break offer — Navi checks in before the learner burns out
Brain Break offer — Navi checks in before the learner burns out
Brain Break — 2, 5, or 10 minutes of breathing, with calm music
Brain Break — 2, 5, or 10 minutes of breathing, with calm music

Decision 02

I gave the learner a guide they choose, not a mascot they're given

4 guides

On Meet Your Guide, the learner picks their guide's look from four guides and names them — Navi, Sunny, Coach, or a name they type themselves. They can add their own photo. This is not decoration: a learner who chose their guide trusts the guide, and trust is the whole product for someone who has been burned by software before.

The trade-off I accepted: four guide looks and four name paths is more build than one fixed character. I paid it, because the first two minutes of NaviLearn decide whether Carlo ever comes back.

Original onboarding setup step
BeforeOriginal onboarding setup step
Meet Your Guide — pick the look, pick the name, or create your own
AfterMeet Your Guide — pick the look, pick the name, or create your own

Decision 03

I cut the bottom menu from four icons to three

4 → 3

My earlier mock had four tabs. Four competing doors made new users hesitate. I cut it to Home, Offline, Settings — and the bar appears only where it belongs: from Home onward, never on Welcome or the setup screens, where it would compete with the one thing the learner is doing. The bar is an Apple-style frosted-glass toolbar; in dark mode the tabs shift to a bluish gray from the same family.

What I chose not to do: keep Progress as a tab because “apps have one.” Progress exists as a screen. It didn't earn a permanent seat in the bar.

Original homepage — four icons in the bottom bar
BeforeOriginal homepage — four icons in the bottom bar
My Learning Path — three doors: Home, Offline, Settings
AfterMy Learning Path — three doors: Home, Offline, Settings

Decision 04

Ask Navi shows one answer, with its source — and help that refuses to just hand over the answer

1 answer

A feed of ten results feels generous and reads as homework. For a low-literacy learner, ten cards is ten chances to give up. Ask Navi shows one answer, and under it, a source line — “From your saved lesson · GED Math Basics” — because not knowing which sources to trust was a named barrier in my research.

Help is graduated: 1 · Hint → 2 · Example → 3 · Answer, in that order, and I deliberately built no one-tap escape hatch that skips straight to the answer. A guide that hands over answers stops being a guide. The guidance line says it in Navi's voice: “Try once first — I'll stay with you.”

Ask Navi — one answer, its source, and the warm-up exchange in the same conversation
Ask Navi — one answer, its source, and the warm-up exchange in the same conversation

Decision 05

Navi remembers what tripped you — and a miss is never a gate

This is the decision I'm proudest of, and it came from watching how the big tutoring apps fail: they either punish a wrong answer or forget it happened. NaviLearn does neither.

The warm-up card sits at the top of My Learning Path: “Warm-up · Fractions — Last time this was tricky. 2 quick questions, then your lesson.” If the learner misses, nothing blocks. The missed category returns the next day as a different question — never the identical one — rotating day by day until the learner gets one right. Then it clears, with a congratulatory note and a gold badge: “You got it right — this one took a few tries. Well done.”

The trade-off: a gate would be simpler to build and would inflate “completion.” I chose memory over pressure, because my learners have already had enough of systems that lock them out.

Original homepage
BeforeOriginal homepage
My Learning Path — the warm-up card leads, reassurance is upfront
AfterMy Learning Path — the warm-up card leads, reassurance is upfront
The cleared state — the badge the learner earns when the tricky one finally lands
The cleared state — the badge the learner earns when the tricky one finally lands

Decision 06

I locked one design system across all 21 screens, in both themes — continuity is a law

21 screens · 2 themes

The same three-band banner (navy, light blue, teal), flush edge to edge. The same title pill, 20px below the banner. The same typography roles at the same sizes. The same chips, the same toolbar, the same states — on every screen, so a learner never has to relearn the page. When screens drifted, I fixed the system, not the symptom: Next buttons aligned to one baseline across the setup flow, selection shown by fill and a soft ring — never by outline alone.

And a hard rule with no exceptions: light/dark parity. Whatever changes in the light set is mirrored in the dark set, and vice versa. All 21 screens exist in both, and the dark set is wired, not just drawn.

Original Welcome screen
BeforeOriginal Welcome screen
Welcome, light — the free line reads in both: “Free · No card · Works offline”
AfterWelcome, light — the free line reads in both: “Free · No card · Works offline”
Welcome, dark — the same free line, same hierarchy
AfterWelcome, dark — the same free line, same hierarchy

Decision 07

When the contrast math failed, I changed the color, not the standard

2.88:1 → 5.12:1

My competitive study surfaced a failure in my own palette: white text on my design teal (#2AA9A3) measures 2.88:1 — it fails WCAG AA, and my learners include low-vision users, so “close enough” was not available to me.

I tested three options and chose the surgical one: darken the teal to #1F7A76 (5.12:1, passes) only where white text sits on solid teal. A full-file inventory found exactly one element that qualified — the Completed Check badge on Lesson Detail — so the banner's teal band, which carries no text, kept its original color. The dark badge got the same treatment, with its check flipped to white.

The lesson I show my work with: accessibility fixes should be measured, minimal, and verified — not a palette overhaul performed for show.

Lesson Detail — before: the badge that failed at 2.88:1
BeforeLesson Detail — before: the badge that failed at 2.88:1
Lesson Detail — after: deep teal #1F7A76, 5.12:1, passes AA
AfterLesson Detail — after: deep teal #1F7A76, 5.12:1, passes AA

Decision 08

I built offline in from the start, not as a settings promise

6-day streak

My original concept promised offline learning; an early mock mentioned it in Help but gave the learner no way to download anything. I closed that gap for real: lessons carry a download action, the Offline screen holds what's saved, and screen 21 is the empty shelf — an honest “nothing here yet” state that teaches the learner how to fill it instead of scolding them for its emptiness.

Reassurance is upfront, not buried: “✓ Your progress is saved” sits at the top of My Learning Path beside the 6-day streak pill, because fear of losing work was a named research barrier. (My May mock showed 6 days on the dashboard and 12 on Progress. I ruled it: six — one number, everywhere.)

Original offline download screen
BeforeOriginal offline download screen
Offline — saved lessons, ready without a connection
AfterOffline — saved lessons, ready without a connection
Offline, empty shelf — a first-run state that teaches instead of scolding
Offline, empty shelf — a first-run state that teaches instead of scolding

Decision 09 · Language lives in exactly one place

Original Settings screen
BeforeOriginal Settings screen
Settings + Accessibility — where Language lives, the screen you see before you tap the Language row
AfterSettings + Accessibility — where Language lives, the screen you see before you tap the Language row
The Language Picker — seven languages, one tap from Settings
AfterThe Language Picker — seven languages, one tap from Settings

Decision 10 · The study material had to be real

Knowledge Check — question 2 of 3: the number sentence
Knowledge Check — question 2 of 3: the number sentence
Knowledge Check — question 3 of 3: the word problem
Knowledge Check — question 3 of 3: the word problem
Read With Me — real passages, read-aloud with speed control
Read With Me — real passages, read-aloud with speed control

The reality check

What went wrong, and what it taught me

  • My own memory braided three true things into one false one. I remembered timed 2/5/10-minute breathing choices in my May mock. They exist in none of my sources — the mock had a single 5-minute breathing line inside Simplify. The 2/5/10 choices are my design, specced in from the start of the November build — and the case study says so, instead of dressing a new decision up as recovered history.
  • A checker once reported screens “missing” that were simply unloaded. I verify a file is fully loaded before I report an absence now. Absence of evidence got confused with evidence of absence exactly once.
  • The serif on my four newest screens is Lora, not Georgia — the build environment cannot load Georgia. The switch happens on my Mac, where Georgia lives. Until then, those frames are not pixel-identical to their siblings, and I say so rather than letting a reviewer find it.

Scope is a design decision

What I chose NOT to do

  • Not gate the guide behind a paywall. NaviLearn is free. The market's biggest apps are moving their AI tutors down their price tiers — free is the direction, not a compromise. Welcome says it on its face: Free · No card · Works offline.
  • Not copy features competitors have retired. When a tutoring feature is discontinued by the company that built it, it does not go into NaviLearn. I study what the field is keeping.
  • Not use the mic-bar version of Ask Navi. I built two. I chose the typing box with a Send arrow because it matches the chat apps my learners already know. The mic version stays in my files as an iteration, clearly labeled.
  • Not restart when the plan changed. When audits found gaps between my concept, my written case study, and my build, I attached fixes in journey order instead of redesigning from screen one. Continuity is for the designer's process too.

Fully prototyped design

Where it stands

21 screens in light and dark, plus the Brain Break Offer and Language Picker overlay states — fully prototyped, with a working walkthrough. Setup collects reading level and interests, so Navi's picks are personal from the first session. A walkthrough video built from these Mocks 2 screens — the flow in order, switching between light and dark — sits with this case study for review.

Open the working walkthrough (opens in a new tab)

Like what you see? Let's talk.

Next Project — Tyo's Auction App →← Back to Portfolio
crystalrichardson690@gmail.com

Crystal Richardson · UX Design Apprentice

Amy

Crystal's portfolio assistant

Ask me anything about Crystal's design process, projects, or experience.

Hey, I'm Amy 👋 — Crystal's portfolio assistant. Ask me about her work, process, or experience.

Accessibility

100%
Edit with