Polling is dead. Users in 2026 expect dashboards that update the moment data changes — not 30 seconds later when the next polling interval fires. We have standardised on Supabase Realtime with Next.js for every dashboard project, and the developer experience is remarkably smooth. Supabase's Realtime engine uses PostgreSQL's built-in replication stream to broadcast row-level changes over WebSockets, which means any INSERT, UPDATE, or DELETE on your tables is pushed to subscribed clients within milliseconds.
The architecture is straightforward. Server components fetch the initial dataset and render the full dashboard on first load — no loading spinners, no skeleton screens. Then a thin client component wraps each data section and subscribes to the relevant Supabase Realtime channel. When a change arrives, the client component merges it into the local state using an optimistic update pattern. For a logistics client, this meant their operations team could see shipment status changes in real-time across a fleet of 200 vehicles without ever hitting a refresh button.
The tricky part is handling conflicting updates and stale subscriptions. If two users edit the same record simultaneously, the last write wins at the database level, but the UI needs to handle the intermediate states gracefully. We implement a version vector on each row — a simple integer column that increments on every update — and the client component only applies incoming changes if the version is higher than what it currently holds. This prevents visual glitches where a newer local edit is overwritten by an older server broadcast.
Performance at scale requires channel management. Subscribing to every table in your database is a recipe for WebSocket saturation. We use row-level security policies combined with filtered subscriptions — Supabase allows you to subscribe to changes that match specific filter criteria, so each user only receives updates relevant to their view. For a SaaS analytics dashboard with 500 concurrent users, this reduced WebSocket message volume by 95% compared to a naive broadcast approach, keeping the connection lightweight and responsive.