A good handoff answers three questions before anyone asks them. Build the dashboard around those three questions instead of a generic service overview.

Most handoff dashboards are just the standard service overview reused for a different purpose, and it shows, half the panels answer questions nobody asks during a handoff. A purpose-built one starts from the three questions an incoming on-call engineer actually needs answered, then builds exactly one panel per question.

1. What is currently open

The first question is never about historical trends, it is about anything currently unresolved. Pull open alerts directly rather than making the incoming engineer search a separate paging tool.

from alerts
| where status == "firing"
| where service.name in ("checkout-api", "payments-api", "inventory-service")
| project alert_name, fired_at, severity, runbook_url

2. What changed in the last shift

A deploy, a config change, or a feature flag flip in the last eight hours is the single most common root cause of anything that happens in the next eight. Surface it explicitly rather than assuming the incoming engineer will think to check.

from deployments
| where service.name in ("checkout-api", "payments-api", "inventory-service")
| where deployed_at > now() - 8h
| project service.name, version, deployed_at, deployed_by

3. What is trending toward a problem

Not every risk fires an alert yet. A panel showing SLO burn rate and cardinality growth over the last shift catches the thing that will page the next person if left unaddressed, giving the incoming engineer a chance to get ahead of it.

from metrics
| where metric.name == "slo.burn_rate"
| where service.name in ("checkout-api", "payments-api")
| summarize avg(value) by service.name, bin(timestamp, 1h)

4. Keep it to three panels, resist adding more

The temptation with any handoff dashboard is to keep adding panels for completeness. Resist it, a handoff that takes ten minutes to review does not happen consistently, three focused panels that get read every single time beat twelve comprehensive ones that get skimmed once and then ignored.

  • Lead with currently open alerts, pulled directly, not a link to a separate tool.
  • Surface every deploy and config change from the last shift explicitly.
  • Add a trending-toward-trouble panel for burn rate and cardinality, not just current state.
  • Cap it at three panels, a handoff dashboard that takes ten minutes will stop getting read.