Online Booking
Availy — the patient-facing booking funnel of a multi-clinic network, and now a product sold to practices beyond it. Five years of design work, measured in bookings.
Measured in PostHog across the network's own clinics since the funnel was instrumented in late summer 2026. Earlier analyses come from the network's 2021 funnel analytics.
Every completed booking is clinic revenue. Every abandoned one is a phone call — or a patient somewhere else.
Patient21 runs a network of dental and general-practice clinics across Germany. Availy is their front door: the surface where a website visitor becomes a booked patient. It has to answer three questions before anything else — which clinic, which treatment, which time — and it has to do it fast, because the alternative is one tap away.
The market leaves no room for a mediocre funnel: in our national research, 51% of patients booked online — and 84% of those who booked by phone knew online booking existed but chose a human anyway. A booking flow doesn't just have to function. It has to earn more trust than a receptionist's voice.
I have designed Availy for five years, from its first version through a full overhaul to the funnel that today runs across the network’s clinics — and has become a product in its own right, now sold to practices outside the network.
We measured before we redesigned — and killed a comfortable excuse on the way.
When we took the funnel apart step by step, the picture was uncomfortable and useful:
Session recordings explained part of it: patients didn't understand that the collapsed day-bars were hiding free slots. They saw days, not appointments — and left. The occupancy analysis mattered just as much: it removed the easiest excuse ("the clinics are simply full") and put the responsibility back on design, where the evidence said it belonged.
Show the room before the form. Trust is built before the funnel begins.
Patients hesitate when they can't picture where they're going. A short photo strip — reception, treatment rooms, the team — does more for completion than any reassurance copy.
Treatments named the way patients describe them — and a first screen that kept earning its simplicity.
The treatment list is where every booking starts, and it went through three deliberate generations. The first version used photography — friendly at a glance, but it tested badly: a usability test of photos against icons showed icons winning clearly; the images added visual noise without helping patients choose. The second generation shipped the icons, grouped treatments the way patients describe them — not the way the system stores them — and gave every row a plain-language description.
Then we retired the icons too. With honest groupings, a duration on every treatment, and descriptions that actually answer "is this for me?", the pictograms were repeating what the words already said. What survived is exactly what patients scan for: the name, the time it takes, and what's included. Simplification isn't a style — it's what's left when every element has to justify its row.
The smallest possible question, asked once. Slots surface immediately.
Two dropdowns — patient status and insurance — and we already show available times for tomorrow. The patient sees a real time before they enter a single character.
Account creation is pushed to the very end of the flow — never as a gate. Patients can browse, choose a slot, and only then commit to an account if they want one. The dark pattern of "register to see times" is gone. We also dropped questions the business never read — insurance data was collected for years and used nowhere, so it left the critical path.
The slot picker was the 46%. It got the most deliberate design in the product.
Slots are visible by default — no collapsed bars, no expand icons to discover. On mobile, the calendar shows three days instead of the original five: a full Monday-to-Friday grid with every slot was simply too much information at once, and reducing the visible window let individual times be seen and compared without scanning fatigue.
One screen, three views — a week, a city, a person. The picker scales from a single practitioner's calendar to nearby clinics to a roster of available doctors with rating, languages and specialties. Same component, different lens — the patient picks the dimension that matters most to them.
Not every reassurance reassures. We added the clinic's phone number to the slot step as a safety net for hesitant patients — and conversion in the pilot clinic dropped from 30% to 21%: people abandoned the funnel to call instead. We removed it and kept the lesson: an escape hatch placed at the moment of decision is an invitation to escape.
Confirmation as a hand-off, not an end-card.
The confirmation screen does three jobs at once: confirms the booking with everything the patient needs (treatment, doctor, address, time), offers a one-tap calendar add, and bridges into the digital anamnesis — saving 15 minutes in the waiting room. The flow doesn't end at "booked"; it sets up the visit.
Post-booking communication was redesigned end-to-end: immediate in-app confirmation, SMS + email with full clinic address and directions link, a 24-hour reminder with one-tap reschedule, and appointment-specific preparation instructions. The reminder-with-reschedule proved the most valuable addition — patients who would have silently no-showed moved their appointment instead.
Two clinics first. Then the network.
The overhaul shipped as a staged test in two clinics before rolling out across the network — measure, then scale. The success metrics were business metrics from the start: more bookings self-served online, without the no-show rate creeping up.
27% of visitors become booked patients — in a median of 58 seconds. And mobile finally caught up.
In 2026 the funnel was instrumented in PostHog, giving the product its first reliable end-to-end measurement. Today, 27% of visitors who open the booking screen complete a booking — 12,300+ visitors and 3,300+ completed appointments in the measured period — with a median time-to-book of 58 seconds. Booking a doctor's appointment in under a minute is a designed outcome, not an accident.
The number I watch most closely: desktop converts at 29.8%, mobile at 26.6%. In 2021, desktop converted at twice the mobile rate. The two eras were measured with different tools, so I don't claim a delta — but the gap collapsing from 2× to three points is the story of a mobile-first rebuild that worked, on the majority of the traffic.
The leak that remains is slot selection — and part of it is no longer an interface problem. When the next free appointment for a new patient is three weeks out, no UI saves that booking. Knowing where design problems end and capacity problems begin is what keeps a funnel honest.
Sources: network funnel analytics 2021 · PostHog instrumentation 2026
Three screens are designed for the next round of testing. None of them have earned a number yet.
The funnel that runs today is the one measured above. These are the next proposals — drawn, reviewed internally, and waiting on a test. I am keeping them in the case study rather than out of it, because a flow that is finished is a flow nobody is still learning from.
