Identifying and Eliminating Waste

Studies consistently show that organisations waste 25–35% of their cloud spend on resources that are idle, over-provisioned, or forgotten. In Azure, common waste sources include VMs running around the clock with near-zero CPU usage, unattached managed disks left over after VM deletion, orphaned public IP addresses, and unused App Service plans. Identifying and eliminating this waste is the first and highest-ROI step in any FinOps practice.

Azure Advisor continuously analyses your resource utilisation and generates Cost recommendations automatically. Common recommendations include: right-size or shut down under-utilised VMs (below 5% CPU or 2% memory utilisation for 7 days), purchase Reserved Instances for consistently running VMs, delete unattached public IP addresses, and remove idle Application Gateway instances. Each recommendation shows the estimated monthly savings so you can prioritise by impact.

Azure Monitor collects CPU percentage and network in/out metrics for every VM. A VM consistently below 5% CPU is a strong candidate for right-sizing or shutdown. Use the Virtual Machine Insights feature in Azure Monitor to see a fleet-wide view of CPU and memory utilisation over 24 hours, 7 days, or 30 days. The Advisor recommendation for under-utilised VMs links directly to the performance charts so you can confirm the pattern before acting.

When a VM is deleted without choosing to delete its disks, the managed disks remain and keep accruing charges. In the Azure portal, filter Disks by Disk state: Unattached to find these orphaned disks. Similarly, old snapshots taken for backup or migration purposes are often forgotten. Snapshot storage is cheaper than managed disk storage, but a year of forgotten snapshots across a large subscription can add up to thousands of dollars annually.

Public IP addresses in Azure are billed even when not associated with a running resource. When VMs or load balancers are deleted, their public IPs and network interfaces may remain. Filter Public IP Addresses by Associated to: None in the portal and delete those you no longer need. Similarly, orphaned network interfaces that are not attached to any VM or service can be safely removed. These are small individual charges but add up across hundreds of resources.

Zombie resources are long-lived staging, testing, or proof-of-concept environments that were provisioned and then forgotten. Unlike idle production VMs, these may have moderate utilisation during business hours but serve no active project. Implement a resource tagging policy that requires Environment and Project tags on every resource, then schedule automated alerts for resources where Environment=Dev or Environment=Test that are more than 90 days old.

Azure VMs support an Auto-Shutdown feature that automatically stops a VM at a configured time each day. Enable this on all dev and test VMs to ensure they do not run overnight or on weekends. For example, a dev VM running 24/7 costs the same as 24 hours, but one configured to auto-shutdown at 6 PM and auto-start at 8 AM on weekdays only uses 50 hours per week instead of 168 — saving approximately 70% on compute cost.

Azure Cost Management includes an Anomaly Detection feature that uses machine learning to identify unexpected spikes in spending. When a subscription or resource group's cost deviates significantly from predicted patterns — for example, someone accidentally deploys a GPU VM or forgets to delete an ExpressRoute gateway — Cost Management sends an anomaly alert via email. Setting up anomaly alerts is a quick win that catches runaway costs before they appear on the monthly bill.

Not all waste is compute-related. Azure SQL Databases provisioned on the Business Critical tier but running simple read-heavy queries may perform identically on the General Purpose tier at half the cost. Blob Storage data that is rarely accessed should be moved from the Hot tier to the Cool or Archive tier using lifecycle management policies. Azure Advisor surfaces database and storage rightsizing recommendations alongside compute ones.

Treat cost optimisation as an ongoing engineering practice, not a one-time event. Create a FinOps backlog in Azure DevOps or Jira with items generated weekly from Advisor recommendations and cost anomaly alerts. Assign each waste item to the owning team and set a target resolution date. Track savings realised vs. savings identified to measure FinOps maturity. Schedule a monthly cost review with engineering leads and finance to keep waste elimination top of mind across the organisation.

Prevention is more cost-effective than remediation. Use Azure Policy to deny the creation of expensive VM SKUs (like GPU or M-series) without approval, require Environment and ExpiryDate tags on all resources, and automatically enable auto-shutdown on newly created VMs in dev subscriptions. Combine policies with budget alerts set at 80% and 100% of the monthly allocation so teams are notified before overspending occurs.

Test your understanding of Microsoft Azure Fundamentals (AZ-900) concepts from this lesson.

In this lesson you learned: Azure Advisor surfaces actionable cost recommendations for idle VMs, unattached disks, and over-provisioned services, auto-shutdown for dev VMs can cut compute spend by 70%, and governance policies with tagging prevent future waste from accumulating. Next up we look at right-sizing techniques and autoscaling strategies.