A scheduling platform that lets patients rebook themselves

How Apoliums approached it.

Built a booking platform with automated WhatsApp reminders and a reschedule flow patients could self-serve.

1. Model the slot, not the appointment

   Availability is generated from a doctor's recurring rota plus exceptions, and every slot row carries its own unique constraint on doctor and start time. A double booking becomes a database error rather than a support ticket.

2. Make the reminder a queued job, not a cron sweep

   Each booking enqueues its own reminder at confirmation time with an idempotency key. Rescheduling cancels the pending job and enqueues a new one, so a patient who moves an appointment three times still receives exactly one reminder.

3. Give the patient a signed link instead of an account

   Reschedule and cancel links are short-lived signed URLs tied to a single booking. Patients change their own appointments without a password, and the clinic never stores another credential.

4. Record the outcome of every slot

   Slots close as attended, cancelled or missed, with a timestamp and the actor who closed them. That single column is what turned no-shows from an anecdote into something the clinic could report on.

What the system runs on.

• Next.js
• Node.js
• PostgreSQL
• Twilio

Front end

Next.js App Router. The staff diary is a client-rendered week view; the patient reschedule page is a server-rendered route behind a signed token.

API

Node.js service with all booking writes wrapped in a single transaction, slot claim, booking row, and outbox entry commit together or not at all.

Data

PostgreSQL. Unique index on (doctor_id, starts_at) for slot integrity, plus an append-only booking_events table that carries the audit trail.

Messaging

Twilio for WhatsApp and SMS, driven from a transactional outbox so a message is never sent for a booking that rolled back.

Jobs

A queue worker drains the outbox, applies per-recipient rate limits, and retries with exponential backoff on provider errors.

What changed after launch.

Fewer missed appointments after automated reminders shipped.

• The diary became the single source of truth: rota changes and bookings now live in one system rather than two documents.

• Double bookings stopped being possible at the database level rather than being caught by staff.

• Cancellations and no-shows are now separate, timestamped outcomes, which is what made reminder performance measurable at all.

• Reschedules move to the patient, so the front desk handles the exceptions instead of every routine change.