40% Cloud Cost Reduction Without Sacrificing Performance
Led a FinOps initiative that reduced AWS monthly spend from $180K to $108K — a 40% reduction — without degrading performance or availability. Established ongoing cost governance through tagging, rightsizing, and reserved instance strategy.
Challenge
AWS monthly bill had grown to $180K with no clear ownership or rightsizing strategy, and finance was demanding accountability for runaway cloud costs.
Solution
Implemented a FinOps framework — resource tagging for cost allocation, rightsizing audit, reserved instance strategy, and idle resource cleanup across all environments.
Result
Monthly spend dropped from $180K to $108K (40% reduction), all performance SLAs maintained.
The Problem
I was managing delivery for a growing e-commerce platform built on AWS when the CFO raised a red flag. The monthly cloud bill had crept from $90K to $180K over 18 months, and nobody could clearly explain why. There was no tagging strategy, so costs could not be attributed to specific teams or services. Engineers spun up resources for testing and never tore them down. Production instances were oversized because "we might need the headroom." Reserved instances had expired and were never renewed, so everything ran on-demand pricing.
The finance team wanted a 30% reduction. Engineering leadership was nervous that any cuts would impact performance or uptime. I was asked to lead the initiative and find a path that satisfied both sides.
What I Did
I started by establishing visibility. Working with the DevOps team, I implemented a mandatory tagging policy — every resource had to be tagged with team, environment, service, and cost centre. For existing untagged resources, we ran a two-week cleanup sprint to retroactively tag everything. This alone gave us the first clear picture of where money was going.
Next, I led a rightsizing audit. Using AWS Cost Explorer and third-party tooling, we analysed utilisation data across all EC2 instances, RDS databases, and ECS clusters. The findings were striking — nearly 35% of compute instances were running below 20% average CPU utilisation. We downsized these in staged rollouts, monitoring performance metrics at each step.
I then worked with finance and engineering to build a reserved instance and savings plan strategy. For stable, predictable workloads, we purchased 1-year reserved instances. For variable workloads, we moved to spot instances with proper fallback configurations.
Finally, I implemented automated cleanup policies. Non-production environments were scheduled to shut down outside business hours. Unattached EBS volumes, old snapshots, and orphaned load balancers were identified and removed. I set up weekly cost reports sent to each team lead, creating accountability at the team level for the first time.
The Outcome
Monthly AWS spend dropped from $180K to $108K — a 40% reduction that exceeded the CFO's 30% target. All performance SLAs were maintained throughout the process; not a single incident was attributable to the cost optimisation work. The tagging and governance framework I established became a permanent part of the organisation's cloud operations, with quarterly cost reviews built into the delivery cadence. Engineering teams reported that the visibility actually helped them make better architectural decisions, not just cheaper ones.