How We Reduced AWS Costs by 62% for a B2B SaaS Platform
The Situation
A Series B SaaS company came to us with a straightforward problem: their AWS bill had grown from $18k/month to $51k/month over 18 months, while their customer count had grown only 3×. The cost curve had decoupled from growth, and they needed to understand why — and fix it — before their next fundraise.
62% cost reduction · $380k annual savings · 6 weeks · 0 app code changes
Discovery: What Was Actually Happening
Week one was pure discovery. We tagged every resource, built an Athena-based cost attribution model, and identified the top cost drivers by service and team. Three patterns dominated:
- Over-provisioned RDS: Six db.r6g.2xlarge production databases running at 8% average CPU. Two weeks of CloudWatch data confirmed they should have been db.r6g.large instances.
- NAT Gateway egress: $12k/month in NAT data processing fees — primarily S3 and DynamoDB traffic that should have been routing through free VPC Gateway Endpoints.
- Orphaned resources: 340 unattached EBS volumes, 2.1 TB of old snapshots, and 14 unused Elastic IPs accumulated over three years of rapid growth.
The Fix: Sequencing Matters
We prioritised by impact and risk. NAT Gateway routing changes were zero-risk and delivered immediate results. RDS rightsizing required a maintenance window but carried no application risk — we validated using historical metrics before touching anything. Reserved Instance purchases came last, after rightsizing was complete and stable. Buying RIs before rightsizing locks you into the wrong instance sizes.
Key Learnings
- Always rightsize before purchasing Reserved Instances — in the wrong order, RIs lock in waste permanently
- VPC Gateway Endpoints for S3 and DynamoDB are free and take 10 minutes to configure — there is no reason not to use them
- Tagging discipline matters more than any single optimisation; without it, attribution is impossible and decisions are guesses
- Schedule a monthly orphan sweep — resources accumulate faster than intuition suggests in fast-moving engineering teams
Summarize this post with:
Ready to put this into production?
Our engineers have deployed these architectures across 100+ client engagements — from AWS migrations to Kubernetes clusters to AI infrastructure. We turn complex cloud challenges into measurable outcomes.