Skip the blank-canvas paralysis. Clone a starter layout, swap two queries, and ship a dashboard the team will actually open during an incident.

The default Helix workspace ships with a Service Overview template. Most teams either ignore it or rebuild it from scratch. Neither is necessary, clone it and edit two panels.

1. Clone, do not create

From any service page, use Dashboards, New from template, Service Overview. This gives four panels pre-wired: request rate, error rate, p95 latency, and saturation (CPU/memory). That is the RED-plus-USE hybrid most on-call engineers reach for first.

2. Point it at a service

Every panel on the template uses a dashboard variable, $service, instead of a hard-coded name. Set it once at the top of the dashboard and all four panels update.

from traces
| where service.name == "$service"
| summarize p95(duration) by bin(timestamp, 1m)

3. Add the one metric that actually matters

Templates give the generic four. Every service has one metric that matters more than those, queue depth, cache hit rate, a business counter. Add a fifth panel for it.

from metrics
| where metric.name == "checkout.cart_abandon_rate"
| where service.name == "$service"
| summarize avg(value) by bin(timestamp, 5m)

4. Pin it where on-call will find it

A dashboard nobody opens during an incident did not save anyone any time. Attach it to the alert rules for the service under Alert, Linked dashboard, so it is one click from the page, not a search.

That is the whole workflow: clone, retarget the variable, add one custom panel, link it to alerts. Five minutes, assuming nobody gets distracted by panel colors, which, refreshingly, Helix does not let anyone customize on the starter template anyway.

5. Delete it if nobody looks at it

Six months from now, half the dashboards created this way will have zero views in the last thirty days. That is fine, and worth checking for on a quarterly basis: a stale dashboard pointed at a service that got merged into another one is worse than no dashboard, because it answers a question with an answer that stopped being true months ago.

  • Clone the Service Overview template instead of building from a blank canvas.
  • Use a single $service dashboard variable so every panel updates from one field.
  • Add exactly one custom panel for the metric unique to the service.
  • Link the dashboard to alert rules so it is reachable from the page itself.