Dashboards are useful only when they answer the first questions during an incident.
Useful observability is opinionated
Which customer is affected? Which dependency changed? Which queue is backing up? Which deployment introduced the shift?
Good observability connects traces, logs, metrics, and deployment events so the team can move from symptom to ownership quickly.
The operational win
The goal is not to collect every possible signal. The goal is to make the important path obvious:
- Request ID
- Tenant or customer
- Service boundary
- Kafka topic or partition
- Database latency
- Error classification
When this is done well, incident response becomes less dependent on memory and more dependent on a system that explains itself.