UX Research · Product Design · Design System · Redesign Concept
Uniform Teeth — when the app is your orthodontist.
Every screen now answers "what now?" and shows who's watching. A past project, revisited: Uniform Teeth hired me for the app UI on their existing prototype — UI only, no research or audit in scope. Years later I revisited it independently and redesigned it end to end.
Background
The service is loved. The app isn't.
Uniform Teeth is orthodontist-led clear aligners, remote-first: 3–5 clinic visits, everything else in the app. It sits between in-office Invisalign and mail-order kits with no doctor. Clinics rate 4.5–4.6, the app 3.3–3.8. Between visits, the app is the orthodontist.
Challenge
Three problems. One nervous patient.
- Navigation hides "what now". Three tab-bar versions, an unlabelled button, no stable Home.
- Trust handled as fine print. Consent under a modal, a countdown on payment, card data in plain text.
- Progress you can't see. No "aligner X of Y"; the before/after photos hidden in a sub-tab.
The delivered UI followed the client's prototype; this looks at the app as it shipped and aged.
Discovery
197 screens. Five competitors. One gap.
I audited all 197 screens of the live app, read the reviews on App Store, Play, Trustpilot and Yelp, and went through five competitor apps — Invisalign, SmileDirectClub, Byte, Candid, Impress. None of them shows the patient that a doctor is actually watching. That became the core of the redesign.
| Who | Profile | Needs |
|---|---|---|
| Busy professional · primary | 28–40, chose Uniform to skip monthly visits | Check in in seconds, fast replies |
| Second-timer | 30–45, teeth shifted after teen braces | Proof it will hold this time |
| Complex case | 25–45, priced out of Invisalign | Clear costs and refund terms |
| Parent of a teen · secondary | Pays and supervises | Proof a doctor oversees it — out of scope |
Thinking
Make the supervision visible.
Four principles, each answering a fear from the reviews. Success criteria were set before design: user confidence, on-time photos, 30-day retention. Then ~130 wireframes, clickable before any visual design.
| Principle | In the app | Fear it answers |
|---|---|---|
| Doctor presence is a feature | Name, face, "Reviewed by Dr. X", messages the doctor wrote | "No one is really watching" |
| One "what now" per screen | One labelled tab bar, one primary action | "The app will fail me" |
| Progress you can believe | Aligner X of Y, before/after, review history — never 0% | "It won't work" |
| Money without pressure | No timers; totals and refund terms next to Pay | "I can't get a refund" |
Outcome
One core loop, rebuilt end to end.
The weekly photo check-in is the core of the product. It went from 7–8 taps to 4, and every step now says what happens next.
| Step | Purpose |
|---|---|
| 1 · Reminder | Push: "Time to switch to aligner 2. Take your progress photo first." |
| 2 · Home | "Day 7 of 7 — change today". One action: take a progress photo. |
| 3 · Photo check | Automatic quality check: "Looks good — clear and well lit". |
| 4 · Under review | "Keep this aligner on until Dr. Lachica confirms." |
| 5 · Reviewed | Push from the doctor: "You're on track. Keep going!" |
| 6 · Phase complete | "Nice work, Sarah. On to Align." The next set is on its way. |
Payment
Before paying, the patient sees what's included, the schedule and the refund terms next to the Pay button. The card is masked, and the confirmation repeats the numbers.
Progress
Progress opens on the before/after with "Reviewed by Dr. Lachica · On track". Sharing goes through a privacy step, and the first option is the care team, not social media.
Consent
Four consents are separate rows, each with one plain-English line. "Agree & book" unlocks only when all four are checked.
The first weeks
Six moments from the first days: the box arrives, the patient checks it, puts the aligners in, treatment starts, the first photo, and Home comes alive.
Design system
Real photography, one accent colour, locked colour roles: orange is the action, green is "on track". Built on a Figma design system with light and dark tokens.
What shipped
Results
Checked before any user study. Forecast before any launch.
No live users yet — these are my own expert reviews of the hi-fi design, plus a test script for the moderated study. The walkthrough found 12 issues — all fixed.
Test script
Five real-life scenarios, written for the moderated study. Not run with users yet.
What was found→What it coversFive scenarios for mid-treatment users — written, not yet run- A week has passed — switch aligners tonight with peace of mind
- "Is it even working?" — find proof of progress
- A meeting collides with the clinic visit — reschedule
- Day three, it presses harder — is this normal?
- New card — make sure the next payment goes through
Cognitive walkthrough
I walked five personas through the actual screens. 12 findings, 6 high — all fixed.
What was found→What was found12 issues, 6 high-severity — all fixed- The ring said "4 days left" while Home said "check-in due"
- After sending a photo — no instruction whether to switch or wait
- Progress showed no sign that a doctor had reviewed anything
- No way back to an existing appointment
- No way to report "something feels off" on your own
- Empty chat didn't say when the care team replies
Heuristic evaluation
Nielsen's ten. 28 findings: 3 high, 14 medium, 11 low.
What was found→What was found28 findings — 3 high, 14 medium, 11 low- Upload was a bare spinner → progress bar, send in background, cancel
- $500 charged with no summary → charge summary + refund policy
- Photo quality left to the user → automatic check with retake
- "Tray" and "aligner" mixed → one word, 74 strings
- Cancelled visit had no undo → "Keep my visit" for 15 minutes
Copy audit
~450 strings reviewed. 90 strings and 32 labels changed in one pass; vocabulary locked.
What was found→What was found~450 strings reviewed, 90 changed, 32 button labels- Five identical "Continue" → "Next: Clinic", "Next: Time"…
- "Submit forms" → "Send to my care team"
- Errors made actionable: "Your bank declined this card. Try another card or contact your bank."
- Empty states say when content appears: "After your Aug 19 visit"
Projected outcome
This is a concept, so these are design targets, not results. How I got them: baselines are typical numbers for mobile health and subscription apps, since no client data was shared. Each target is tied to one specific change and sized by published benchmarks for that kind of change. Confidence is lower where clinical factors, not the UI, drive the number.
| Metric | Baseline (est.) | Target | Driven by |
|---|---|---|---|
| On-time weekly photos | ~60% | +12–18 pp | 7–8 taps → 4, status at every step, quality check |
| Payment conversion | ~55% | +5–9 pp | Charge summary and refund terms, no countdown |
| Photos rejected | ~15% | −5–8 pp | Automatic photo quality check |
| 30-day retention | ~30% | +6–10 pp | Review stamp, "Something feels off?", reply time |
| Status tickets | — | −25–35% | Status shown at every step |
Benchmarks: mobile health apps keep 25–35% of users at 30 days; transparent pricing lifts checkout 5–15%; remote monitoring with visible progress lifts adherence 10–20 pp.
North Star: on-time weekly photos — they predict both the clinical outcome and retention. Three A/B testsWhere the targets come fromIndustry benchmarks: mobile health apps (30-day retention 25–35%), checkout transparency (+5–15%), remote monitoring with visible progress (+10–20 pp adherence). The tests are how they would be proven.Test A — PaymentOld payment screen vs. charge summary + refund policy. Measures: paid within 7 days.Test B — Photo loop · main testOld vs. new weekly check-in. Measures: on-time photos over 4 weeks.Test C — ReassuranceWith vs. without review stamp and "Something feels off?". Measures: 30-day retention. would prove or disprove the forecast; if photos move less than 8 points, the constraint is motivation, not the UI.
Selected screens
The journey, screen by screen.
