Anushka Gupta

Queue Cure: Zero-Refresh Clinic Queue

76% of India's 1.5 million clinics still run on paper token slips and shouting. Patients wait 2–3 hours with zero visibility into when they'll be called; they can't step out for water, a call, or the washroom without losing their turn. Receptionists juggle the queue from memory and rewrite slips on every change. Doctors have no dashboard of who's next; the result: anxious patients, overworked staff, and missed tokens. Clinics needed a live, zero-refresh queue that both sides can trust in real time. Process Started by mapping the two real users: the receptionist (fast, mistake-proof input) and the waiting patient (glanceable, anxiety-killing info). Picked Supabase Postgres + Realtime so both screens stay in sync over WebSockets no polling, no refresh button. Moved token assignment and "call next" into SECURITY DEFINER SQL functions with FOR UPDATE SKIP LOCKED so two receptionists tapping at the same instant can never issue duplicate tokens or double-advance the queue. Wait time is computed live as (tokens_ahead × avg_consult_minutes) — never hardcoded; the avg is editable and pushed to every screen in <1s. Tried optimistic UI first, dropped it: the real-time echo was already fast enough, and a single source of truth was safer. Results Shipped a working 2-screen system in one sprint. Token sync latency measured at ~400ms between receptionist action and waiting-room update — zero manual refresh needed. Receptionist flow reduced to 2 clicks (type name → Enter, then "Call Next"). Wait estimates update live the moment avg consult time is tuned. Concurrency test: 2 receptionists tapping "Call Next" within the same second produced 0 duplicate "now serving" states thanks to row-level locking. Patients on the waiting-room screen can finally step away without losing their place. Reflection I'd compute avg consult time dynamically from the rolling median of completed_at − called_at instead of asking the receptionist to set it; the data is already in the table. I'd also add an SMS/WhatsApp ping when a patient is 2 tokens away, so they can leave the clinic entirely. On the engineering side, I leaned on Supabase Realtime for both screens; for a 1.5M-clinic rollout, I'd add a presence channel + heartbeat so the "live/offline" badge reflects true connectivity, not just subscription status. And I'd add a "skip & recall later" action for no-shows.