A 4× growth-stage HealthTech serving European customers had AWS spending climbing faster than revenue. We delivered 38% reduction without touching reliability, and turned cost optimization into a sustained engineering practice.
After 4× growth in 18 months, AWS spend had become the second-largest line item in the company's P&L, and was still climbing. The CTO needed savings without slowing delivery.
Rather than a 12-week analysis with a giant remediation list at the end, we sequenced the work so savings appeared on the bill within the first two weeks, and compounded from there.
Started with read-only AWS access via cross-account IAM role. Pulled 90 days of Cost Explorer data, ran Compute Optimizer recommendations, audited every Reserved Instance against actual usage, and inventoried idle resources.
Started with the lowest-risk, highest-impact items first. Idle resource cleanup, RI/Savings Plan rebalancing, and right-sizing the worst-offender EC2 instances. Each change went through dev → staging → 5% canary → full production.
Architectural changes that took longer to validate but compounded existing savings. EKS rightsizing, Graviton migration for eligible workloads, NAT Gateway consolidation. In parallel, set up the per-team cost visibility and weekly review cadence that turns one-time savings into sustained discipline.
The architecture diagrams below show the meaningful changes. Note what stayed the same: multi-AZ deployment, KMS encryption, multi-region backups, ISO 27001 evidence collection. We never traded reliability for cost.
Every saving below was validated against the AWS bill before the engagement closed. No projected savings, no estimated future benefits, only confirmed monthly cost reductions visible on the AWS invoice.
| Optimization Area | Method | Monthly Saving |
|---|---|---|
| Reserved Instance / Savings Plan rebalance | Sold mismatched RIs, bought Compute Savings Plans | $6,500 |
| EC2 right-sizing | 22 instances down-sized per Compute Optimizer | $4,100 |
| EKS rightsizing + Karpenter | Auto-scaling vs static peak provisioning | $3,000 |
| Graviton migration | 18 instances moved to ARM-based instances | $2,300 |
| NAT Gateway consolidation | 3 → 1 NAT GW + VPC endpoints for AWS services | $1,400 |
| S3 Intelligent-Tiering | 3 largest buckets migrated to tiered storage | $1,000 |
| Idle resource cleanup | Unattached EBS volumes, unused EIPs | $530 |
| RDS right-sizing | 3 Aurora instances down-sized | $510 |
| CloudWatch log retention | Retention rules + S3 archive policy | $380 |
| Other (small line items) | Spot instance migration on dev/staging, etc. | $610 |
| Total Validated Monthly Saving | $18,000 |
The $18k/month is the headline number, but the real outcome is the lasting practice. The team now runs cost reviews monthly, catches anomalies within 24 hours, and treats cloud cost as an engineering KPI.
Every saving was validated against the AWS bill before sign-off. No projected savings. The CFO confirmed the new run rate at the end of month 3.
Deployment frequency, lead time for changes, and MTTR all remained within their pre-engagement bands. Cost optimization didn't slow the team.
Per-team cost dashboards in Grafana, Cost Anomaly Detection routed to Slack, and a monthly cost review meeting between engineering and finance with a documented agenda.
ISO 27001:2022 evidence collection ran throughout. CloudTrail, multi-AZ, KMS, and Backup remained unchanged. No reduction in DR or audit capacity.
We expected to save 15-20%. We saved 38%, with no reliability impact. The dashboards alone changed how our engineers think about deployments, that's the part we didn't see coming.
Share read-only access to your AWS account before the call. We'll spend 30 minutes walking through your Cost Explorer and identify at least one optimization worth $X/month, yours to keep whether you engage us or not.
★ AWS Advanced Tier Services Partner · ISO 27001:2022 · ISO 9001:2015 · 5× AWS-Certified Founder
