Digital Agency

The agency ran infrastructure for eight client accounts, ranging from small WordPress deployments to a multi-region e-commerce platform. Two engineers handled everything. Each client environment was different — different cloud, different tooling, different alerting setup, or no alerting at all. Incidents were chaotic because there was no common mental model across accounts.

We designed a shared operations layer that worked across all eight environments without forcing them into a single stack. A Grafana instance with per-client dashboards gave the team a single pane of glass. Prometheus exporters ran on every host; log shipping was standardised to a shared Loki instance behind client-scoped label filtering.

We then wrote 24 runbooks covering the most common failure modes across all eight accounts — not generic templates but actual decision trees written with the engineers, tested against real recent incidents. The runbooks lived in a shared repository linked directly from PagerDuty alerts.

At the end of the engagement the team estimated they were spending roughly 30% less time on reactive firefighting. They had capacity to take on a ninth client account without adding headcount.