A three-week proof that stopped an R800k build

A Cape Town healthtech running clinic-management software wanted to add a patient-facing booking product. The plan on the table was a nine-month, roughly R800k full build, greenlit on a board deck, before anyone had tested whether patients would actually self-book at all.

The entire business case rested on one unproven assumption: that patients would switch from phoning the clinic to booking themselves online. Everything downstream (the R800k, the roadmap, the hiring plan) treated that assumption as settled fact. Nobody had put a real booking flow in front of a real patient.

We built a proof of concept in three weeks: a genuinely working self-booking flow wired to a handful of pilot clinics, not a clickable mock-up. I ran product and design; my engineering network built it. Then we put it in front of real patients and watched what they actually did.

The evidence reframed the whole thing. Patients booked happily, but only for routine repeat visits. The complex first-appointment flow the nine-month build was centred on went almost untouched. We killed the big plan, scoped a focused MVP around the repeat-booking workflow patients actually used, and built and shipped that instead.