First: see and alert
1. Set budget alerts on every scope
An alert at 80% and 100% of budget, sent to the application owner. It is free, takes minutes to set up, and is what prevents incidents costing tens of thousands of euros.
2. Tie every resource to an application
One tag is often enough, linked to your CMDB. Without it, part of the bill belongs to nobody, so nobody optimizes it. See our tagging guide.
3. Analyze the top 10 services
Ten services usually make up most of the bill. A precise analysis of each often reveals quick wins: sometimes a 5-minute change is worth hundreds of thousands of euros a year.
The quick wins
4. Delete what is no longer used
Unattached disks, old snapshots, orphaned public IPs, forgotten environments: cleaning requires no application change.
5. Stop non-prod when it isn't used
Automatic Stop every evening with Start on demand saves about 90% on development and test VMs and VDIs. A fixed schedule caps at 50-60%. See our Stop/Start guide.
6. Rightsize on real usage
A VM whose CPU and RAM stay under 50% at P95 over 30 days can go down one size, saving about 50% on compute.
Pay less for what remains
7. Reserve the prod that runs all the time
Reserved Instances or savings plans: about 50% off a production VM. Always after steps 4 to 6. See our commitment strategy.
8. Reuse your licenses
Azure Hybrid Benefit on Windows Server and SQL Server, if you already own the licenses with Software Assurance.
The items that drift silently
9. Storage and backup
Put data on the right tier (Hot, Cool, Cold, Archive), review backup retention, avoid geo-redundancy in non-prod.
10. Logs and monitoring
Collect only useful logs, adjust retention and the commitment tier: the bill can be divided by four. See our Log Analytics and Sentinel guide.
And to make it last
These ten actions bring the bill down. To keep it from rising again, you need governance that makes FinOps by design: standards, building blocks, monthly follow-up. See our governance page and, to move fast, our FinOps audit.