Case study · Anonymised
AWS cost audits: up to 55% lower spend
The problem
Three clients were overspending on AWS. One had a large idle EC2 inventory, one ran oversized RDS instances at low utilisation, and one had no Savings Plans despite predictable workloads. None could see cost by team or service.
Constraints
Each audit had two weeks. Production performance could not be put at risk to save money, and long-term commitments had to match real workload shape rather than a guess.
The decision
Start with cost allocation and utilisation data, remove idle resources, right-size with a safety margin, commit only stable workloads to Savings Plans or Reserved Instances, and add ownership guardrails so savings would not erode.
Where the money actually was
The largest savings came from the simplest findings: idle EC2 instances nobody remembered launching, RDS instances still sized for a load test from months earlier, and unattached EBS volumes. Python scripts flagged EC2 below 5% CPU, RDS below 8% utilisation, unattached volumes and unused Elastic IPs.
Options considered
- Savings Plans or Reserved Instances. Savings Plans for compute, because they flex across instance families. Reserved Instances only for stable RDS workloads.
- Spot or On-Demand for batch. Fargate Spot with graceful interruption handling gave about 70% savings on batch, with interruptions below 2%.
- Aggressive or conservative right-sizing. Sizing used 30-day P95 metrics plus 20% headroom, so cost cuts did not create performance incidents.
Making the savings stick
S3 lifecycle policies moved data to Infrequent Access at 30 days and Glacier at 90. Lambda and EventBridge shut down development environments outside working hours. Budgets alerted owners at 80% and 100%. QuickSight gave finance and engineering the same view of cost by team, service and environment, replacing a monthly PDF.
What went wrong
One client was close to committing to a three-year All Upfront plan on the wrong instance family. Modelling the scenarios against real usage caught it before an expensive multi-year mistake was locked in.
Results
| Measure | Before | After | When |
|---|---|---|---|
| Idle EC2 instances | 47 | 0 | 2024 engagements |
| Savings Plans coverage | 0% | 78% of eligible compute | 2024 engagements |
| Combined AWS spend | before the audits | up to 55% lower | After the changes |
Figures are as reported by Rahul and recorded in the InfraTales proof register. Measured on the client's systems; not a promise of the same result elsewhere.
When this approach would not fit
A two-week audit works when spend is large enough that idle resources and missing commitments are material. For a small or very new AWS account, a tagging baseline and budget alerts are usually enough until the workload stabilises. Committing to Savings Plans before usage is predictable risks paying for capacity you will not use.
Verified record
- Engagement
- Across two-week AWS cost audits for three clients in fintech, e-commerce and SaaS, Rahul removed 47 idle EC2 instances, raised Savings Plans coverage from 0% to 78% of eligible compute, and cut combined AWS spend by up to 55%.
- Savings
- Rahul's cost and reliability audits have removed more than ₹50 lakh (about $60,000) a year in AWS waste for clients.
- Typical range
- Documented AWS cost reductions from these audits are typically 30-40%, and up to 55% on the strongest engagements.