Observability is a product surface, not a job title
Operators do not read your dashboards. They are reading their own panic. Design accordingly.
The mistake
Most observability stacks are designed by the people who built them. The dashboards make sense to the engineers who already know what is going on. They are unreadable to anyone else, including the on-call engineer at 3am who is not the engineer who built the dashboard.
The shift
The teams I see doing observability well in 2026 treat the dashboard as a product surface. The operator is the user. The user is panicked. The user wants the answer to "is it broken, and if so, where, and if so, what do I do."
Three patterns that survive 3am
1. The single status page. One URL. Big text. Green / yellow / red. Linked to runbooks.
2. The Slack-first incident channel. All signal flows into a single channel. No dashboards. Just text.
3. The "what changed" panel. Top of every dashboard. Deploys, feature flags, config pushes in the last hour.
The pattern across all three is that the operator is not asked to interpret. They are told what is wrong.
Filed in Observability
Tags: Observability · SRE · Design