Seven decisions. Each one kills something specific from the map.

1. I built the state machine before drawing a single screen.

Every appointment runs on one spine: Booked → Reminded → Checked-in → In-queue → Next-up → In-consult → Completed → Follow-up-due, plus the states everyone pretends don't exist: Delayed · Emergency-bumped · Rescheduled · No-show · Cancelled. Get the spine right and every screen is just a rendering of it. That is also the scaling story: the states are universal across every clinic and hospital; only the skin changes.

2. One truth, two skins.

Each state renders differently depending on who's looking. Patient gets reassurance. Staff gets control.

Never dump operational noise on the patient. Never make staff dig for status. Same fact, two jobs.

3. I designed the disruption first, not last.

The delay screen is the hero of the product: live position, a plain-English reason ("Dr. Sharon was called to an emergency"), an honest ETA, three actions. The worst moment in the journey is where trust gets won, because it is the exact moment every competitor goes silent. The rule I held myself to: never go silent, always explain. A wait without a reason is the failure, not the wait.

4. Accountability cuts both ways.

Hospitals penalize late patients but never own their own delays. That one-sidedness came up unprompted in every conversation I had. So in Anvaya, a hospital-side delay is visible, explained, and met with options. It's not a UX nicety. It's a trust feature.

5. I gave the staff a board, not a register.

One screen: every checked-in patient, their location, their wait-timer, who's next for each doctor. Doctor finishes → one tap calls the next person to the right room. Emergency hits → one action broadcasts the reason to everyone affected and re-sequences the queue. The coordinator's "third of my day hunting people" is pure recoverable waste, and the dead time between consults is lost hospital capacity.

6. I split Care from Bookings.

The fix for the midway discovery. What the doctor prescribes became first-class objects with their own lifecycles, each permanently linked to the consult that created it. A test runs Ordered → Scheduled → Sample-given → Results-ready → Reviewed; a medicine runs Prescribed → Active → Refill-due → Completed. And the move that made it click: an ordered test can be scheduled into a booking by reference. It shows up on your day and on the staff queue like anything else, but it never loses the thread back to the consult that ordered it. Bookings answer "where do I go and when." The care plan answers "what did the doctor tell me to do, and where is each of those things now." Two questions, two objects. That thread is literally why the product is called Anvaya.

7. Reminders that don't give up, and a day with an order.

Confirmation → a day-before nudge with one-tap confirm/reschedule → a morning-of message with travel time and a prep checklist. The fasting instruction that one of my interviewees discovered at the counter now arrives the night before. And a multi-booking day stops being a meaningless stack. It becomes a sequenced itinerary, grouped by floor, with live status on each item.

Underneath it all, one system. Status tokens: one color-and-shape language for every state, identical on both surfaces. The patient appointment card (four states). The care-plan cards (lineage-stamped). The staff queue row (five states). And the disruption and empty states designed first-class, because that is where the value actually lives.