cloudcostdown.euAWS Cost Optimization · Case Studies
AWS Cost Optimization: Real Results, Real Numbers
Three verified CloudCostDown client results - plus published industry benchmarks - showing what AWS cost optimization actually delivers for Series A–C startups. No ranges, no estimates. Before and after.
Our Results
These are CloudCostDown’s own client results. Company names are anonymized - the numbers are not.
Series B Fintech · Our Result
Series B Payments Platform
€127K/year
annual savings
Situation
Payment processing platform at €28K/month AWS spend. No Reserved Instances, oversized RDS fleet, heavy NAT Gateway traffic to S3 and DynamoDB from private subnets.
Key findings
RDS running at 9% avg CPU. NAT Gateway processing 60TB/month - almost entirely S3 and DynamoDB API calls that could route via free Gateway Endpoints. EC2 fleet at 11% average utilization.
Changes made
S3 and DynamoDB Gateway Endpoints (free), RDS rightsizing db.r5.2xlarge → db.r6g.large (Graviton), EC2 rightsizing + 1-year Compute Savings Plan. No downtime. All changes implemented via IaC pull requests on retainer.
Healthtech SaaS · Our Result
Series A Healthtech Platform
€8,400/mo
monthly savings
Situation
HIPAA-adjacent healthtech at €22K/month AWS spend. Dev and staging environments running 24/7. ECS Fargate workers on On-Demand. CloudWatch ingesting 800GB/month of debug logs.
Key findings
Non-prod environments idle 14h/day and all weekend. Fargate background workers fully fault-tolerant - perfect Spot candidates. CloudWatch log retention set to 'Never expire' across 140 log groups.
Changes made
AWS Instance Scheduler for non-prod (saved €2,100/mo), Fargate Spot for workers (saved €3,800/mo), CloudWatch retention + log filtering (saved €1,200/mo), gp2→gp3 EBS migration (saved €1,300/mo).
Series C Marketplace · Our Result
Series C Two-Sided Marketplace
35%
AWS bill reduction
Situation
High-traffic marketplace at €85K/month AWS spend. Rapid growth had left the architecture over-provisioned across the board. Scaling events driving expensive cross-AZ data transfer.
Key findings
EKS node groups overprovisioned by 2–3×. Cross-AZ traffic generating €12K/month in data transfer charges. RDS IOPS provisioned at 10× actual usage. No Savings Plans coverage on steady-state compute.
Changes made
Karpenter for bin-packing + Spot node groups, cross-AZ traffic consolidation, RDS IOPS rightsize + Reserved Instances, 3-year Compute Savings Plan on baseline. Results achieved in 6 weeks.
Why the numbers vary
Savings depend on how much waste was accumulated before the audit. Companies that grew fast without FinOps controls typically have more - which is why we see larger savings at Series B and C. The floor is €15K/year (guaranteed) but the ceiling is determined by how long the waste has been compounding.
Published Industry Results
These are published case studies from other AWS consulting firms - not CloudCostDown’s work. They’re included to show what’s achievable across different company types and starting conditions.
Series A SaaS startup · via ZeonEdge ↗
74%
bill reduction ($47K → $12K/mo)
Primary savings drivers
- EC2 rightsizing (m5.xlarge → t3.medium): -$3,840/mo
- RDS rightsizing to Graviton (db.r5.2xlarge → db.r6g.xlarge): -$2,400/mo
- Compute Savings Plans (36% discount): -$8,600/mo
- NAT Gateway → VPC Gateway Endpoints for S3/DynamoDB: -$3,200/mo
- Spot Instances for background workers: -$2,800/mo
- Dev/staging environment scheduling: -$2,800/mo
US-based Series A SaaS/AI startup · via LeanOps ↗
88%
bill reduction ($13K → $1.5K/mo)
Primary savings drivers
- Multi-tenant architecture migration (eliminated per-customer cluster overhead)
- Dynamic auto-scaling replacing always-on over-provisioning
- Docker image reduction 278MB → 65MB, startup time 30s → <10s
- Full IaC automation removing manual drift
- Additional $1,000+/mo saved in 3rd-party SaaS subscriptions
Fast-scaling FinTech startup · via SquareOps ↗
30%
reduction from peak ($44K → $31K/mo in 3 months)
Primary savings drivers
- RDS Reserved Instance coverage for steady-state workloads
- EC2 and EKS rightsizing + autoscaling tuning
- NAT Gateway optimization via VPC Endpoints
- CloudWatch log retention and ingestion controls
- Cost tagging + ownership model (FinOps governance)
Typical Savings by Technique
These ranges come from the results above plus the wider pattern across 50+ AWS accounts. Your actual savings depend on how much waste is present - the only way to know is to measure.
| Technique | Typical savings |
|---|---|
| EC2 rightsizing | 20–40% compute savings |
| NAT → VPC Gateway Endpoints | $800–3,200/month eliminated |
| Compute Savings Plans (1-year) | 36–66% off steady-state compute |
| Spot Instances (fault-tolerant workers) | 60–90% compute discount |
| Graviton migration | 20–40% better price/performance |
| Dev/staging environment scheduling | 55–65% non-prod compute savings |
| RDS rightsizing + Graviton | 30–50% database savings |
| CloudWatch log reduction | 60–80% observability cost cut |
The compounding problem
These aren’t one-time gains - most are permanent reductions to the monthly baseline. A startup that finds €10K/month in waste and waits 6 months to act has burned €60K. That’s why the only bad audit is the one you don’t do.
Who Does the Work
Adam Pavlát
AWS Cost Optimization Consultant
Every engagement is run by me directly - not delegated to a junior analyst. Before CloudCostDown I was a cloud cost engineer at Disney+ Hotstar, managing infrastructure at tens of millions of daily users, and at Oddin.gg. I’ve seen what happens when cloud spend scales without controls across hundreds of AWS accounts at enterprise and startup scale.
The pattern recognition that makes these results possible comes from seeing the same waste profiles across enough accounts to know exactly where to look - and which changes move the number.
Common Questions
Are these numbers real or estimated?
Our three results are from actual client engagements - the figures come from comparing the AWS bill before and after implementation, measured 60–90 days post-change. Company names are anonymized by agreement. The industry benchmarks are from published case studies, linked to their original sources.
What's a realistic expectation for my startup?
The honest answer: it depends on how much waste has accumulated. Startups that grew fast without FinOps controls tend to have more. The floor we guarantee is €15K/year (or the audit is free). The actual figure comes from the audit itself - we don't estimate before seeing the data.
How long does it take to see the savings?
The audit delivers a detailed report and prioritised savings roadmap in 1 week. Implementation happens on the Managed FinOps Retainer as IaC pull requests (Terraform or CDK). Quick wins like NAT Gateway endpoints and Savings Plans can be live within days of a retainer starting. Rightsizing typically completes within 30–60 days. The retainer is where the bill actually drops.
Do the savings compound over time?
Monthly baseline reductions are permanent - they compound positively. The risk is drift: without ongoing monitoring, new resources get provisioned at default sizes, Savings Plans expire without renewal, and costs creep back up. That's what the Managed FinOps Retainer is for.