Cloud cost optimization is one of the highest-ROI engineering initiatives available to enterprise organizations. We helped a Fortune 100 client reduce their annual AWS spend by $4M — approximately 38% of their total cloud bill — in 6 months.
Here's the playbook we used.
Phase 1: Visibility (weeks 1–3)
You can't optimize what you can't see. We started by deploying AWS Cost Explorer with resource-level tagging across all 380+ workloads. Within two weeks, we had identified that 40% of spend was coming from 8% of resources — the classic Pareto distribution.
Phase 2: Rightsizing (weeks 3–8)
EC2 rightsizing is the fastest win. Using CloudWatch metrics, we identified instances running below 10% CPU utilization for 30+ consecutive days. A combination of automated rightsizing scripts and manual review for critical workloads reduced compute spend by 22%.
Phase 3: Reserved Instances and Savings Plans (weeks 8–12)
After rightsizing, we committed to 1-year Savings Plans for the stable compute baseline, and Convertible RIs for the top 20 database instances. This alone saved $1.2M annually on a normalized basis.
Phase 4: Architecture changes (weeks 12–24)
The largest savings came from architectural changes: migrating batch workloads to Spot Instances, moving infrequently-accessed S3 data to Glacier, and implementing aggressive CloudFront caching to reduce data transfer costs.
The key lesson: FinOps isn't a one-time project. It requires continuous monitoring, anomaly detection, and quarterly review cycles embedded into your engineering culture.