Observability as a platform at Nationwide

The brief

More than forty engineering teams needed logs, metrics, traces and dashboards without depending on a central team for every change. The existing support model made the platform team a queue, while inconsistent implementations made operational evidence harder to compare.

The work treated observability as a product with users, interfaces and a support model rather than a collection of installed tools.

A common platform, not one fixed dashboard

The platform combined enterprise Kubernetes with Elasticsearch and the Grafana observability stack: Loki for logs, Mimir for metrics and Tempo for traces. Shared foundations handled operation and integration; teams retained the ability to express the views and alerts their services required.

Templates and self-service paths replaced ticket-driven dashboard creation. Consistent labels, data-source conventions and access patterns gave teams a known route from telemetry to investigation without forcing every service into an identical shape.

KEDA connected workload scaling to operational signals where that model was appropriate, turning telemetry into part of runtime behaviour rather than a passive display layer.

Operating model

Leading a twelve-person team made adoption and support as important as installation. The platform was documented, vended and improved around repeated team needs. Common requests became product capabilities; unusual requests remained visible instead of disappearing into one-off configuration.

Result

The platform served more than forty teams and reduced dashboard-related support tickets by 90%. Engineers gained a coherent route across logs, metrics and traces, while the platform team recovered time previously spent reproducing routine dashboard work.