Client Background and Challenge
A major Midwestern US regional bank with approximately 40,000 employees serves retail banking, wealth management, commercial lending, treasury services, and insurance across 12 states. The organisation operates over 600 branches and a full-service digital banking platform processing millions of transactions daily.
AWS workloads include digital banking applications (mobile and online), real-time fraud detection and anti-money laundering (AML) systems, data analytics and customer intelligence platforms, core banking API layers, disaster recovery and business continuity infrastructure, and development and testing environments for a 1,200-person technology team.
Annual AWS spend reached approximately $6.3 million — a 180 percent increase over three years without corresponding cost management. The spend was distributed: EC2 compute (48 percent), RDS and DynamoDB database services (22 percent), S3 and EBS storage (14 percent), data transfer and networking (9 percent), and managed services including Lambda, SageMaker, and CloudWatch (7 percent).
The bank operates in a heavily regulated environment subject to OCC (Office of the Comptroller of the Currency), FFIEC, SOX, and PCI-DSS requirements.
"Regional banks migrating to AWS face a predictable cost trajectory: rapid consumption growth during migration (150 to 200 percent over three years is typical), followed by a plateau where the organisation realises it is significantly overspending because infrastructure was sized for migration speed rather than operational efficiency. The window between migration completion and EDP renewal is the optimal time to intervene — the bank has enough consumption history to identify waste, the EDP renewal creates negotiation leverage, and the consumption volume justifies meaningful discount improvements. We typically find 25 to 35 percent savings opportunities for banks at this stage of their cloud journey."
Key Challenges Identified
No Independent Review of Consumption Efficiency
The bank had never conducted an independent review of AWS consumption patterns. The technology team had prioritised speed of deployment over cost optimisation during initial cloud migration, creating a legacy of over-provisioning:
- Many EC2 instances sized for peak capacity but running at 15 to 25 percent average utilisation
- Development and testing environments running 24/7 despite being used only during business hours
- S3 storage accumulating without lifecycle policies (regulatory archives stored in expensive S3 Standard instead of S3 Glacier)
Poorly Structured Reserved Instance Portfolio
The existing Reserved Instance portfolio was 60 percent misaligned with actual usage. Most were purchased based on initial migration sizing rather than steady-state requirements, resulting in reservations for instance types and sizes no longer matching production workload.
Outdated Enterprise Discount Program
The AWS EDP negotiated three years prior at lower spend levels no longer reflected the bank's purchasing power. The bank's 180 percent consumption growth meant the original discount tiers were obsolete.
Lack of Cost Allocation and Accountability
AWS costs were tracked at account level but not attributed to specific business units, applications, or projects, making it impossible to hold stakeholders accountable for consumption.
Managing AWS costs in a regulated environment?
Five-Phase Engagement (10 weeks)
Forensic analysis using AWS Cost Explorer, Trusted Advisor, Compute Optimizer, and detailed billing reports identified critical inefficiencies:
- EC2 instances in development and testing (22 percent of total compute spend) running 24/7 but used only during business hours
- 35 percent of EC2 instances over-provisioned by 2 to 4 times relative to actual CPU, memory, and network utilisation
- RDS instances showing similar patterns, with several running on r5.4xlarge when r5.xlarge would accommodate actual query loads
- S3 storage totalled 1.8 petabytes, of which 1.1 PB (61 percent) was regulatory archive data not accessed in 12+ months but stored in S3 Standard
Total identified waste: $1.5 million annually.
- EC2 right-sizing: 142 instances recommended for downsizing, saving $680K annually
- Scheduling automation: Non-production environments scheduled for business-hours-only operation using AWS Instance Scheduler, saving $420K annually
- Storage tier migration: 1.1 PB of regulatory archive data migrated from S3 Standard to S3 Glacier Deep Archive, reducing storage costs from $25K/month to $1.1K/month — saving $290K annually
- RDS optimisation: 18 database instances right-sized, 6 underutilised read replicas consolidated, saving $180K annually
- Data transfer optimisation: VPC endpoint implementation and data transfer routing improvements, saving $130K annually
Comprehensive review of the existing Reserved Instance portfolio. Sold unused and mismatched Reserved Instances in the AWS marketplace. Replaced with a combination of Compute Savings Plans (providing flexibility across EC2, Fargate, Lambda) and targeted EC2 Instance Savings Plans for stable workloads. Banking-specific consideration: Reserved Instances and Savings Plans for critical financial systems (fraud detection, core banking API) maintained with higher coverage rates to ensure cost predictability and regulatory reporting accuracy.
The bank's existing EDP provided 8 percent discount — below market for its consumption volume. With $6.3 million in annual spend (projected to grow to $7.2 million+ over the next three-year term), the bank was in a strong position to negotiate significantly improved terms. Redress analysis demonstrated that comparable financial institutions at similar spend levels were achieving 16 to 22 percent discounts. Negotiated improved EDP tier of 18 percent discount on committed spend, with provisions for quarterly true-up adjustments as consumption patterns evolved. Formal SLA commitments for critical banking workloads documented, addressing OCC and FFIEC requirements.
Cost allocation tagging taxonomy applied across all AWS resources (business unit, application, environment, cost centre). AWS Budgets configured with appropriate alerting thresholds. Monthly FinOps review process established. Cost optimisation roadmap for the next 18 months documented.
Results and Outcomes
$6.2M Total Savings Over Three Years (33% Annual AWS Cost Reduction)
The engagement delivered significant financial impact through optimisation and improved contractual terms.
$1.5M in Eliminated Spending on Underutilised Resources
Direct elimination of waste through EC2 right-sizing, RDS optimisation, storage tier migration, and development environment scheduling.
42% EC2 Right-Sizing Reduction Across Non-Production Environments
Development and testing environments substantially reduced through scheduling and instance type optimisation while maintaining functionality for the 1,200-person technology team.
18% EDP Discount (Up From 8%)
Successfully renegotiated AWS enterprise discount programme with 10-point improvement over previous terms.
Regulatory Compliance Provisions Formally Documented
OCC, FFIEC, SOX, and PCI-DSS requirements explicitly addressed in renegotiated contract with formal SLA commitments for critical banking systems.
Lessons for Financial Services Cloud Buyers
1. Migrate First, Optimise Second — But Don't Wait Too Long
Most banks over-provision during migration to meet speed and reliability requirements. This is rational during migration but creates significant ongoing waste. The optimal time to conduct an independent review is 6 to 12 months after migration completion, when consumption patterns have stabilised.
2. Your EDP Is a Negotiation, Not a Discount Schedule
Banks often accept AWS's initial EDP offer as non-negotiable. At $5 million+ in annual spend, the difference between a first-offer EDP and a market-rate EDP can be 10 percentage points — representing $500K+ annually on a three-year term.
3. Regulated Environments Require Contract Specificity
Standard AWS terms do not adequately address OCC, FFIEC, SOX, or PCI-DSS requirements. Regulatory provisions — data residency commitments, notification requirements for service changes, audit rights, incident response procedures — need to be explicitly negotiated and documented in the Enterprise Agreement.
4. Cost Allocation Is a Prerequisite for Governance
Without granular attribution of AWS costs to business units, applications, and environments, it is impossible to hold stakeholders accountable for consumption or identify which workloads are driving cost growth.
This case demonstrates that even sophisticated financial services organisations can achieve substantial AWS cost reductions through comprehensive usage analysis, disciplined cost optimisation, and commercial renegotiation informed by market benchmarking. The $6.2 million in three-year savings represents improvement in both financial performance and operational efficiency while maintaining security and regulatory compliance requirements specific to banking operations.