Connect Grafana to one real data source and build a dashboard for an incident question, such as why checkout latency rose. Add a clear baseline, useful units, and an alert that identifies an actionable condition. Give an on-call colleague the dashboard without explanation and see whether they can find the answer.
Check data-source permissions, alert ownership, and configuration backup. Attractive charts can obscure bad metrics or misleading aggregation. Define what each panel means and how operators should act before multiplying dashboards.