Modern stacks are full of powerful components. They are also full of ways for a request to disappear between services. Observability is how a team keeps reality connected to its mental model.
Three questions
Every production system should help answer: what is slow, what is failing, and who is affected? Logs, metrics, and traces are useful only when they map to those questions.
Prefer signals over noise
More dashboards do not automatically create more understanding. Give each important user journey a small set of service-level indicators, then link alerts to an action someone can take.
Boring systems have clear ownership, graceful failure, and few surprises. They leave engineers enough attention to improve the product instead of merely explaining yesterday.