Your team built a dashboard. It has 40 panels. Nobody has looked at it since the demo. That is not a monitoring system — that is a museum, and your metrics are the exhibits.
Here is how to build a dashboard people actually open every morning.
Rule 1: One screen, five numbers, tops
The test: can a new hire tell in 10 seconds whether today is a normal day or an incident day? If your dashboard needs scrolling, it fails. Five leading indicators beat forty lagging ones:
- Are customers getting replies? (first-response time — hours)
- Are orders/jobs completing? (success rate — %)
- Is anything silently broken? (failed-job count — integer)
- How close are we to capacity? (top-3 resource runway — days)
- Did yesterday's fix hold? (regression check — yes/no)
Everything else can live in a drill-down link. Nobody clicks it. That is fine — it exists so the owner knows it's one click away.
Rule 2: Every panel has a threshold painted on it
A number without a "bad" zone is decoration. Paint the threshold directly on the panel:
- Green: within normal range
- Amber: known-risk range, needs a reason if crossed
- Red: page someone
Ambiguity is the enemy. "Users: 1,204" means nothing. "Users: 1,204 (red < 900)" means your team knows whether to celebrate or investigate.
Rule 3: The dashboard opens, it does not get opened
Push it. Post a daily screenshot to the team channel at the same time every morning, generated automatically. A dashboard someone has to remember to visit is a diary. A dashboard that arrives is a ritual. Rituals survive busy weeks; intentions do not.
Rule 4: Kill panels on a schedule
Once a quarter, delete every panel nobody has referenced in a discussion, incident, or decision. Dead panels train people that dashboards are wallpaper. Deletion is the feature that keeps the living panels alive.
Rule 5: Write the one-line story under the chart
Every chart gets a one-line caption that states the current interpretation, updated by whoever touched the data last: "Response time up since Monday deploy #442 — watching until Friday." Charts don't carry context. Stories do. Future-you at 2am during an incident will thank past-you for the caption.
The 45-minute starter version
If you have no dashboard at all, here is the honest first version:
- Pick your 5 numbers (30 min, use the template list above)
- Put them in a shared doc with thresholds and owners (10 min)
- Set one recurring morning post with yesterday's values (5 min)
That is it. A doc that arrives daily beats a Grafana instance nobody opens. Graduate to real tooling when the doc starts hurting — not before.
I package the dashboard layout, threshold templates, and the morning-ritual checklist as ready-to-use files in my small-ops store: the Automation Starter Pack (full stack: monitoring, backups, handoffs) and the free-ish starter Ops Starter Kit.
If you tell me your stack in the comments, I'll tell you which of the five numbers your team is most likely measuring wrong.
Top comments (0)