Batch-update latency in bed-tracking logs

1. System failure identification

   Bed availability shown to placement staff is stale by the length of a batch cycle. Decisions are made against a picture of the hospital as it was, not as it is. Two placements are made against the same bed because both saw it free.

2. Data pipeline diagnostics

   The tracking system updates on a fixed schedule inherited from an overnight reporting design. Latency is a property of the integration and invisible in the interface, which presents every figure as current. The schedule was chosen for a reporting workload that no longer exists.

3. Predictive computational model

   Timestamp every capacity figure with its own age and treat staleness as a displayed attribute. Where feasible, move high-churn fields to event-driven updates and leave the rest batched. Selective real-time updates cost far less than converting the whole feed.

4. Workflow integration

   Show the age of the number beside the number. Users cannot compensate for a delay they cannot see, and an explicit timestamp costs nothing to read. Displayed staleness lets a human apply the judgement the system cannot.

Back to the case study database