Failure alerts (/workflows/failure-predictions) surfaces workflows at elevated risk of failing based on recent run patterns, connector health, and configuration signals.
Open failure predictions
- Navigate from BUILD → Workflows or directly to
/workflows/failure-predictions - Review the ranked list — workflow name, risk score, contributing factors
- Open a workflow to inspect failing nodes or dependencies
- Cross-check Runs for recent errors matching the prediction
What influences predictions
Signals may include:
- Recent failure rate vs historical baseline
- Connector sync errors on dependencies
- Schema or API changes on integrated systems
- Timeout or retry exhaustion patterns
- Missing approvals on write steps
Predictions are advisory — verify before changing production workflows.
Triage workflow
- Confirm — open latest failed Runs for the workflow
- Stabilize — fix connector auth, adjust timeouts, add error branches
- Test — run in a staging Environment when available
- Monitor — watch Metrics and failure list for 24–48h after fix
Link to Schedules
Scheduled workflows with rising failure risk may need pause until fixed — see Schedules.