Cloud costs have a way of compounding quietly. A development environment left running over a weekend, an oversized RDS instance provisioned for a peak that never arrived, reserved instances purchased for workloads that were later containerised — individually they look minor, but across a medium-sized AWS environment they routinely add up to tens of thousands of pounds per month in pure waste.

The Starting Point

Our client was a UK-based SaaS company running their platform on AWS with a monthly bill of approximately £38,000. They knew their costs were high but had no visibility into where the money was going. Their engineering team was focused on product delivery, not infrastructure economics.

Our first action was deploying AWS Cost Explorer with custom allocation tags and enabling Compute Optimizer across the account. Within 48 hours we had a clear picture of the waste.

What We Found

  • Idle and oversized EC2 instances: 23 instances with average CPU utilisation below 5% over 30 days. 11 were development environments that engineers had forgotten to stop.
  • Unattached EBS volumes: 47 volumes totalling 12TB, orphaned after instances were terminated without cleanup.
  • On-demand pricing for predictable workloads: Their production database cluster and application tier had been running on on-demand pricing for two years despite 24/7 steady-state usage — a perfect candidate for Reserved Instances.
  • Data transfer costs: Unnecessary cross-AZ traffic driven by an application architecture that had not been reviewed since initial deployment.
  • Unused Elastic IPs and Load Balancers: Resources from deprecated features and old deployments, still being billed.

What We Did

Week 1–2: Immediate wins — terminated idle instances, deleted unattached volumes, released unused Elastic IPs. Monthly saving: £4,200.

Week 3–4: Right-sizing. Using Compute Optimizer recommendations, we downsized 31 instances to appropriate sizes without any performance impact. We also moved 8 workloads to Graviton3 instances which delivered a 20% price/performance improvement. Monthly saving: £6,800.

Week 5–8: Reserved Instance and Savings Plan purchases for steady-state workloads. A 1-year convertible Reserved Instance commitment on the production tier alone saved £5,400/month versus on-demand.

Week 9–12: Architecture optimisation — restructuring cross-AZ traffic patterns and implementing S3 Intelligent-Tiering for infrequently accessed data.

The Result

Monthly bill reduced from £38,000 to £24,700 — a 35% reduction sustained over the following quarter. The engineering team now receives a weekly cost anomaly digest and we have implemented budget alerts to prevent future drift.

Total implementation cost: 3 weeks of engineer time. Payback period: 18 days.

If your cloud bill has grown faster than your business, book a free FinOps assessment — we will identify your top 5 savings opportunities at no cost.