When I joined the Law Enforcement Cloud Platform at the Home Office, most teams had built their own monitoring and alerting. Each setup made sense to the team that built it. Across the platform, though, alerts went to different places, some went nowhere useful, and nobody could give a straight answer to “is the platform healthy right now?”

We also had a Splunk estate that was expensive to run, a legacy Grafana and Prometheus stack that few people wanted to maintain, and a monitoring team whose remit was too narrow to fix any of it.

The roadmap

A big-bang replacement was never an option. The platform supports live policing services, and teams were relying on what they had, however patchy. So I put together a 24-month roadmap that closed the most critical gaps first and modernised the rest in stages, and we delivered against it piece by piece.

One of the first changes was organisational. We brought Operational and Protective Monitoring into a single team, so one group owned visibility end to end.

One place for alerts

The biggest single change was a shared Alert Manager. It takes alerts from multiple sources, adds context, and routes them into ServiceNow, Jira or email depending on who needs to act.

The build took less effort than the adoption. I led the onboarding of every platform team and tenant, which meant a lot of conversations, a lot of dependency chasing, and keeping stakeholders lined up so we hit the date. More than ten teams and tenants now use it, and none of them maintain their own alerting any more.

Security findings as backlog items

We routed AWS Health events and Security Hub findings through Alert Manager into each team’s Jira backlog. A finding became a ticket with an owner, triaged in the same planning sessions as the rest of the team’s work.

For leadership, we built health and security dashboards and monthly KPI reports showing overall platform health, with scores and leaderboards by team. Leaderboards can sound gimmicky, but no team wants to sit at the bottom of one.

We also designed a Platform Health status page, so users have one place to check current availability, live incidents and past outages.

Retiring the old tooling

With alerting in place, we moved monitoring to a cloud-native CloudWatch approach. A service wrapper and templates let teams set up their own alarms and dashboards without waiting on us. That meant we could decommission Amazon Managed Grafana, Prometheus and the old Splunk heavy forwarders.

Every change went through IT Health Checks and Protective Monitoring testing before going live, and we fixed every critical and high security finding along the way. The platform now has one alerting service and one view of health, costs less to run, and has a stronger security posture.

What I’d do again

  • Change a live platform in stages. Fixing the worst gaps first builds the trust you need for the harder changes later.
  • Plan more time for adoption than for the build. A shared service does nothing until teams move onto it.
  • Push the information to people. Dashboards, monthly reports, alerts straight into Jira and leaderboards in programme calls put the data in front of people where they already are, so they don’t have to go looking for it.

If you’re dealing with the same sprawl on your own platform, I’m happy to compare notes. You can also read the full UK Home Office case study.