How We Reduced AWS Costs by 72% for a Financial Services Company
How we helped a financial services company slash AWS costs by 72%—without compromising performance.
Nitin Garg
Founder, HyperCode
5 min read · Tue Jan 07 2025
📉 Executive Summary
Between September and November 2025, our team led a comprehensive AWS cost optimization initiative for a financial services client. The result: a dramatic reduction in their monthly cloud expenditure from over $10,500 to just under $3,000 — a 72% savings. This transformation was achieved through workload migration, architectural changes, and performance tuning, all without compromising performance, availability, or scalability.
🏢 Client Background
This financial services client relies on AWS to host its core infrastructure. Their workloads include production-grade containerized services, MongoDB clusters, streaming pipelines using Kafka, real-time data processing, and scheduled automation tasks. Over time, costs had spiraled due to scale, fragmentation, and under-optimized resource allocations.
⚠️ Challenges Faced
When we first audited their infrastructure, we discovered:
- Extensive use of on-demand EC2 instances with inconsistent sizing.
- Lambda functions used for cron jobs, running almost continuously.
- MongoDB Atlas clusters (M80 and M20) with high hosting.
- Over-provisioned ElastiCache instances.
- A costly Business Support Plan with limited ROI.
- Lack of cost tagging, alerting, or consolidated reporting.
These factors combined to push their AWS bill to over $6,500/month, with an additional $4,000 spent on MongoDB Atlas.
🚀 Optimization Journey
1. Migrating Lambda to Kubernetes CronJobs
The client had over 40 scheduled jobs triggering Lambda invocations around the clock. By containerizing these workloads and running them as CronJobs on their existing EKS cluster, we eliminated over $1,600 in monthly Lambda costs and improved observability with centralized logging and metrics.
2. Migrating from MongoDB Atlas
Their M80 (production) and M20 (staging) Atlas clusters were expensive and introduced data transfer costs. Additionally, they were running a dedicated analytics node to expose MongoDB as SQL, which added significant cost overhead. We replaced this setup with a more efficient architecture by integrating Trino to query MongoDB directly and expose it as SQL. We also configured Apache Superset on top of Trino, enabling the client to build reports and dashboards using familiar SQL queries. We moved these workloads to a self-managed MongoDB setup on Kubernetes with automated backups and monitoring, saving more than $4,000 per month.
3. Rightsizing ElastiCache
We downgraded from r6g.large to t4g.medium instances and tuned TTLs and eviction policies. This reduced cache costs from $360 to $145 without impacting performance.
4. Optimizing EC2 and Storage
We right-sized EC2 workloads and adopted spot instances for steady services. Additionally, we applied lifecycle policies to S3 and used Intelligent-Tiering for storage optimization as part of our AWS cost optimization strategy.
5. Downgrading Support Plan
We replaced the costly Business Support Plan with the Developer tier, saving $466 per month.
We achieved 72% savings while improving transparency, performance monitoring, and infrastructure ownership.
📊 Results at a Glance
| Service/Area | September 2025 Cost | November 2025 Cost | Reduction |
|---|---|---|---|
| EC2 | $1,997.91 | $1,322.52 | ↓ 34% |
| Lambda | $1,607.34 | $0 | ↓ 100% |
| ElastiCache | $359.62 | $145.29 | ↓ 60% |
| MongoDB Atlas | ~$4000 | $0 | ↓ 100% |
| Support Plan | $466.74 | $0 | ↓ 100% |
| Data Transfer + ELB | ~$430 | ~$255 | ↓ 40% |
| Total Monthly Cloud Spend | ~$10,500 | ~$2,982 | ↓ 72% |
🎯 Client Impact
This transformation enabled the client to reallocate over $7,500 in monthly savings toward product development and innovation. With ownership of their MongoDB infrastructure, increased transparency in scheduling and monitoring, and improved tagging and reporting in place, they now maintain control over their cloud spend in real-time.
💡 Key Takeaways
- Kubernetes CronJobs are more cost-effective and observable for scheduled workloads.
- MongoDB Atlas, while convenient, may not be the best long-term financial choice for data-heavy apps.
- Cost savings require technical audits, not just financial oversight.
🔧 Next Steps
We’re continuing to support the client by evaluating Savings Plans, introducing cost anomaly detection, and reviewing Kafka usage for a potential Kinesis migration. Quarterly reviews are in place to ensure sustained cost control.
🌍 About Us
We help growth-stage businesses cut their AWS bills by 30–70% through cloud-native architectural best practices and engineering-led FinOps.
-------
💬 Want to reduce your AWS bill?
We offer a free, no-risk cloud cost audit. Contact us today.