mh studio Book a free call
← All work
HealthcareBookingConcept project

carely

Booking a doctor in under a minute, and a calmer day for the clinic. A patient app, a clinic web app and a website, all working.

How booking a doctor feelsBefore CarelyWith Carely
Find
phone tag → live slots
Book
9 min → 50 s target
Prepare
no reminder → 48 h and 2 h
Visit
paper forms → check-in on arrival
Follow up
lost notes → all in the app
Illustrative journey for a concept product. Times are targets, not measurements.
Role
Lead product designer, solo
Team
Solo concept. Real build: PM, 2 to 3 engineers, GP advisor
Deliverables
Patient app, clinic app, website
Benchmarks
Doctolib, Zocdoc, One Medical
i

Symptoms

The brief, the people and the limits
The brief

Booking a doctor shouldn’t take a phone call.

Carely is a concept brief: a booking platform that connects patients with local clinics, written as if for a clinic group in Berlin. The problem it targets is familiar: patients wait on hold to book, forget visits and fill the same forms every time, while front desks lose slots to no-shows and spend their mornings on the phone.

Assumed baseline
9 min
to book by phone, including hold time
Assumed baseline
1 in 7
appointments missed without notice
Assumed baseline
23 h
of phone time per clinic each week

These baseline numbers are assumptions written into the brief, not measurements. In a real engagement I would replace them in week 1 with the clinic’s own telephony log and calendar export.

What I owned

Solo concept, end to end.

I owned
  • Problem framing and the research plan
  • Benchmark teardown of Doctolib, Zocdoc and One Medical public flows
  • Information architecture and flows for all three surfaces
  • All UI, the design system and the interactive prototypes, built in code
  • Success metrics and the measurement plan
In a real engagement I would share
  • Metrics and scope with a PM and the clinic operations lead
  • Clinical safety wording with a GP advisor
  • Data protection with the clinic’s DPO
  • Calendar integration limits with engineering
Not mine here: backend, the clinic system integration, legal sign-off.

Planned timeline for this concept

About 7 weeks, split into five phases. This is how I would plan the work as a real engagement, not a record of time spent.

  1. Week 1
    Discover
    Telephony log and calendar baseline, interview plan, personas
  2. Weeks 2 to 3
    Design
    Booking flow, reminders and the reception workflow
  3. Weeks 4 to 5
    Prototype
    Working patient app, clinic web app and website
  4. Week 6
    Build
    Design system, accessibility pass and developer handoff
  5. Week 7
    Test and launch
    Booking tests with patients over 60, then a two-clinic pilot plan
Constraints

Honest limits, set up front.

This is a solo concept project. No patients, doctors or front desks were interviewed or tested, so every claim about behaviour below is an assumption I would validate first. I set these constraints to keep the work honest:

Health data is special-category data
GDPR Article 9 applies. Collect the minimum, keep it in the EU and show each role only what it needs.
Booking is not triage
Carely never gives medical advice. Anything that sounds urgent routes to emergency numbers first.
The clinic calendar is the source of truth
Slots come from the clinic’s existing practice system. If the integration lags, the patient must never see a slot that is gone.
Every age, every confidence level
Patients over 65, carers and parents book most often. Phone booking stays available, always.
Who it’s for

Three sides of the same appointment.

Persona voice

“I just want to see someone this week without calling three clinics.”

Emma, 31 · patient
Persona voice

“Half my morning is phones and paper. I want to look after the people in the room.”

Julia, 44 · front desk lead
Persona voice

“I book for myself and for my mother. Small text and fiddly buttons make me give up and call.”

Leo, 62 · patient and carer

Personas written by me to anchor decisions, not drawn from interviews.

Julia’s morning

08:00 to 10:00 at the desk.

The reception workflow is the real product for the clinic. This persona scenario is how I checked that every clinic screen earns its place.

TimeBefore (assumed)With Carely (as designed)
08:00Phones ring before the doors open. Calls go unanswered while patients queue at the desk.Most routine bookings arrive online overnight. The phone is left for people who prefer it.
08:30A walk-in arrives. Julia flips through the paper diary to find a gap.One tap on “Walk-in” finds the next free slot and books it.
09:10Dr. Haddad is running late. Nobody tells the waiting room.“Running late” on the doctor column notifies everyone still waiting for her.
09:30A patient does not show. The slot stays empty.The 2 h reminder already released it, and the waitlist filled it.
ii

Diagnosis

What I’d expect research to confirm
Research plan

Assumptions first, then tests.

Hypothesis
Time decides
Many patients would pick the earliest slot over their usual doctor.
How I’d test it: Card sort in interviews: slot-first vs doctor-first results.
Hypothesis
Phones are the bottleneck
Most booking calls land in the first hours of the day, when the desk is busiest.
How I’d test it: Hour-by-hour telephony log from two clinics.
Hypothesis
Most no-shows are forgotten visits
A reminder with an easy way out turns silent no-shows into freed slots.
How I’d test it: Compare no-shows before and after reminders, by visit type.
Research plan (not run, concept project)
  • 10 to 12 patient interviews, including 4 people over 60 and 2 carers
  • 2 to 4 front-desk shadowing days across two clinics
  • A short survey of about 300 patients on how they book today
  • Telephony and calendar data from the pilot clinics as the baseline
What I did do
  • Desk research on public booking flows of Doctolib, Zocdoc and One Medical
  • A teardown of their search, slot picking, reminders and cancellation rules
  • Built the three surfaces as working prototypes to test the flows end to end

Riskiest assumption: patients trust live slots enough to stop calling. Cheapest test: 5 unmoderated sessions booking a family doctor, measuring time to book and whether anyone looks for a phone number.

iii

Treatment

What I designed
Clinic web app

A day you can see at a glance.

The clinic app is a resource calendar with a column per doctor, a live now-line and a side drawer for the selected patient. Try it below.

clinic.carely.health
carely Riverside Family Clinic
Wed 18 Nov
LH
Dr. Lina Haddad
7 visits
MR
Dr. Marco Rossi
5 visits
SK
Dr. Sara Kim
5 visits
08:00
09:00
10:00
11:00
12:00
13:00
14:00
15:00
16:00
17:00
  • Click an appointment, then check the patient in
  • Click Daniel Okoye and send a rebook link
  • Press Late on a doctor column
  • Add a walk-in from the header
Open full screen ↗
Patient app

From search to slot in under a minute.

Good morning
Hi Emma how are you?
NEXT VISIT · TODAY
LH
Dr. Lina Haddad
Follow-up, blood test
📅 18 Nov, 14:00📍 0.8 km away
What do you need?
🩺
Flu shots are open
Free with your insurance. 10 minutes, no wait.

My assumption: the most stressful part of booking is the unknown, meaning when, where and how much. Every screen answers those three things first.

  • Tap a specialty, then a time slot
  • Pick video or in person and confirm
  • See the emergency note on Find
  • Cancel or reschedule in Visits
Website

Search first, everything else second.

The homepage is a search box with real availability right below it, plus a map that highlights the doctor you hover. Booking happens without leaving the page, and the emergency note sits right under the search.

See a doctor this week, not next month.
Family doctorKreuzbergSearch
Open the full website ↗
The no-show loop

Remind, release, refill.

A missed visit is usually a forgotten one. Two reminders, each with a one-tap “Can’t make it”, turn silent no-shows into freed slots that the waitlist fills.

1Two reminders
48 hours before
Dr. Lina Haddad, Thu 19 Nov, 09:15
ConfirmCan’t make it
2 hours before
See you at 09:15. Riverside Family Clinic, 0.8 km
ConfirmCan’t make it
Push or SMS, the patient’s choice. The 2 h reminder only goes out if the first was not confirmed.
2Freed slot to the waitlist
Waitlist offer
A 09:15 slot with Dr. Haddad just opened. Want it?
Take this slot
Held for 15 min, then offered to the next person
One person at a time, so nobody races for the same slot. The front desk does nothing.
3No-show in the clinic app
Daniel OkoyeNo-show
10:30 · Acne treatment review
Send rebook link
Logged for the no-show report and followed up with one tap, no phone call. Try it on Daniel in the clinic app above.

I chose reminders over no-show fees. Fees punish the people who forget most, often older or busy patients, and push them back to the phone.

Safety and privacy

Booking is not triage.

Patient safety
Urgent symptoms route out, not in
Symptom search and the “New symptom” reason always show emergency routes first: Chest pain, trouble breathing or signs of a stroke? Call 112 now. Practice closed? Call 116117. On the clinic side, reasons like fever in a toddler get a “Check today” flag so the desk can ask a nurse to call. Try Zara Ali in the clinic app.
Privacy and GDPR
Each role sees what it needs
Front desk sees name, time, status and the type of visit. The reason for visit, history and prescriptions are for clinicians only, and every record view is logged. Patients agree to share their reason with that clinic at booking, can see who has access, and can export or delete their data from their profile. Data stays in the EU, encrypted.

Concept design. Before launch, the wording would be reviewed with a clinical advisor and the clinic’s data protection officer. In the prototype the drawer shows the clinician view.

Accessibility

Designed for Leo first.

The people who book most often are older patients and carers, so accessibility is the main brief, not a checklist at the end.

Readable at any size
Layouts are built to hold at 200% text size, and the clinic app’s smallest labels are next in line to move up to 14px. I darkened the brand teal from #12A07A (3.3:1 with white) to #0C805F (4.9:1) so every button passes WCAG AA.
Clear without colour or sight
Every time slot is a real button with a full label, like “Thursday 19 November, 09:15, Dr. Haddad”. Status always carries text, never colour alone. Touch targets are at least 44px.
A human fallback
Each clinic page shows its phone number and hours. Booking online is an extra option, never the only one.
Booking for someone else
Family members are saved once in the profile and chosen per visit. For children, a parent or guardian books and gets the reminders.

Not yet verified with users. Next step: keyboard-only and screen reader passes, then sessions with patients over 65 using their own phones.

Edge states

When the day does not go to plan.

No slots this week
Show the next available date, a nearby clinic with the same specialty, and “Join the waitlist”. Never a blank page.
Doctor cancels
Every affected patient gets a message with three alternative slots, same doctor first, and rebooks in one tap. The desk sees who has not answered.
Patient running late
A “Running late” button in the visit tells the desk. Within 10 minutes the visit stands; after that, the patient is offered a later slot the same day.
Doctor running late
“Late” on the doctor column notifies everyone still waiting for that doctor with a new estimate. Try it in the clinic app.
Key decisions

And what each one cost.

1
Show the slot first, not the doctor

Results lead with three days of real slots, so booking can start from the list.

Rejected: a doctor directory with profiles first. It is how most listings work, but it adds two steps before anyone sees a time.

What I gave up: continuity of care. So returning patients see their usual doctor pinned first, and the earliest slot second.

2
Columns per doctor, not per room

Front desks think “is Dr. Haddad running late?”. Doctor columns with a live now-line answer that at a glance.

Rejected: a room-based grid and a plain list view. Rooms matter to managers, not to the person at the desk.

Trade-off: room clashes are flagged, not shown as a layout.

3
One button that moves the visit forward

Check in, call into room, finish. The main action changes with the status, so staff never hunt for the next step.

Rejected: a toolbar with every action visible. Faster for experts, but it causes mis-clicks under pressure.

Trade-off: rarer actions sit one level down.

4
Booking is not triage

Any symptom search shows emergency routes first. Carely helps people get a slot, not a diagnosis.

Rejected: an AI symptom checker in v1. It raises clinical risk and medical-device questions a booking product should not take on early.

Trade-off: some patients book the wrong specialty, so the clinic can move them.

5
Calm by default

Soft mint, serif headlines and lots of space. Health is stressful enough; the interface should lower the heart rate.

Rejected: a bright consumer style with badges and streaks. Fun for fitness, wrong for someone worried about a symptom.

Trade-off: urgent states need stronger contrast to stand out, so red is reserved for safety only.

Design system

Calm and clinical.

Teal
#0C805F
Deep teal
#0B7A5C
Ink
#0F2E2A
Mint
#CFF3E6
Sky
#DCEBFF
Peach
#FFE3D3
Lavender
#E8E2FF
Sand
#FFF6E0
Aa Aa
Fraunces · warm headlines
Aa
Figtree · clear interface text

Accessibility rules in the system: Teal #0C805F and Deep teal #0B7A5C are the only fills that carry white text. Mint, Sky, Peach and Lavender are backgrounds only. Red is reserved for safety and no-shows.

iv

Recovery

Target outcomes and how I’d measure them
Target outcomes

Success metrics for a pilot.

Target
50 s
to book online, from 9 minutes by phone
Target
−38%
no-shows with two reminders
Target
11 h
phone time saved per clinic a week
Target
4.5★
patient rating of the booking flow

Success metrics I would set for a 3-month pilot with six clinics. These are targets, not results: no pilot was run.

How I’d measure it
  • Booking time: analytics events from search to confirmation
  • No-shows: the clinic calendar against a 3-month baseline
  • Phone time: the clinic’s telephony log, same weeks last year
  • Rating: one question after the visit, not after booking
Guardrails that must not get worse
  • Time patients wait in the clinic
  • Slot utilisation per doctor
  • Share of bookings by phone for patients over 65 (should fall by choice, not by force)
  • Urgent cases booked as routine visits
v

Aftercare

Scope cuts and reflection
What I left out of v1

Cut on purpose.

Insurance-specific billing
Every insurer and clinic differs. v1 shows only whether a visit is covered.
In-app chat with doctors
It creates an unpaid inbox for clinicians and a safety risk if messages go unread.
AI symptom checker
Clinical and regulatory risk. Booking is not triage.
Repeat prescriptions
Valuable, but it depends on e-prescription integrations per clinic system.
Multi-clinic admin for groups
One clinic has to work well first.
Reflection

What I’d do differently.

What I’d change
Taking this further, I would start from the clinic side. The front desk decides whether a clinic adopts Carely at all, so Julia’s morning should shape more of the product than Emma’s search.
The hardest part
Safety copy that helps without scaring people or giving medical advice. The rule I settled on: say who to call in one short line, before anything else on the screen.
What I’d test first
Whether patients trust live slots enough to stop calling, and whether two reminders feel helpful or nagging.
Open questions
  • Reminder fatigue: is one reminder enough for regular patients?
  • The practice system integration is a hard dependency. How stale can a slot be before trust breaks?
  • Walk-ins: what share of the day should stay unbookable online?
Next project
Maison

An editorial fashion and home store.

View case study →