At a glance
Share of calls to the Sendle API that returned an error (HTTP 4xx or 5xx) over the trailing 24 hours. This is the connector-health pulse behind despatch: when the error rate climbs, label creation, order booking and tracking polls start failing, and parcels stop moving. The alert fires above 2 percent, which for a busy despatch desk is the difference between a clean run and a queue of stuck orders.
What it tracks
The card reads the response status of every request the connector makes to the Sendle API (POST /api/orders for booking and labels, GET /api/orders/{id}/tracking for scan polls, GET /api/me for auth checks) and expresses errors as a percentage of total calls in the window. The time_window is 24H, so it is a rolling-day view that catches incidents fast rather than smoothing them away. The alert trips at >2%. A 5xx spike points to a Sendle-side outage (cross-check Sendle Tracking API Unavailable / 5xx); a 401 / 403 cluster points to an invalidated or rotated key (cross-check Days to Token Expiry); a 422 cluster usually means malformed booking payloads (bad postcodes, missing dimensions) rather than a Sendle fault. The most visible downstream symptom is a drop on Label Generation Success.