How We Cut AWS Costs by 26% Without Touching a Single Line of Application Code

AWS bills have a way of quietly growing while nobody's paying attention. You spin up a server for a project, forget to turn it off. You provision storage for a migration, never clean it up. You pick an instance type based on a recommendation from two years ago and never revisit it. Before long you're paying significantly more than your workload actually requires.

We recently completed an infrastructure audit that reduced a client's monthly AWS spend by 26.6%. Here's exactly how we did it.

The Starting Point

The client was running a production environment on AWS that had grown organically over about three years. Nobody had sat down and reviewed the infrastructure holistically. The bill was growing, the team assumed it was just the cost of scaling, and nobody had time to dig into it.

Before touching anything, we audited the entire account: every running resource, every storage bucket, every data transfer pattern, and every service in use. This is always the first step. You cannot optimise what you haven't mapped.

What We Found

Over-provisioned compute. The most common issue in any AWS account. Instances that were sized for peak load from 18 months ago, running at 15-20% CPU utilisation on average. We right-sized these based on actual usage data from CloudWatch, not guesswork. Dropped to the appropriate instance family and size. Significant saving, zero performance impact.

Unattached and forgotten storage. EBS volumes that had been detached from terminated instances but never deleted. Old snapshots retained indefinitely with no lifecycle policy. S3 buckets from old projects storing data nobody accessed. Storage costs add up slowly and that's why they're easy to miss.

No reserved pricing. The client was paying on-demand rates for workloads that had been running continuously for over a year. On-demand is for unpredictable, short-lived workloads. For anything running 24/7, Reserved Instances or Savings Plans cut the cost by 30-40% on their own.

Data transfer costs hiding in plain sight. Traffic patterns that were routing data between regions unnecessarily. Small per-GB fees that compound into meaningful numbers at scale.

No resource tagging. Without tags, you can't attribute cost to teams, projects, or environments. You end up paying for things you can't trace back to anything useful, which means nothing ever gets cleaned up.

What We Did

We didn't rewrite any application code. We didn't change any deployment architecture. Every change was at the infrastructure layer.

We right-sized the compute instances based on actual CloudWatch metrics. We set up S3 lifecycle policies to automatically move infrequently accessed data to cheaper storage tiers and delete data that had exceeded its retention requirement. We deleted the orphaned EBS volumes and set up snapshot lifecycle policies going forward. We purchased Reserved Instances for the baseline workload and set up Savings Plans for the remainder. We implemented a tagging strategy so every resource could be attributed to a team and purpose. We moved the infrastructure into Terraform so all future changes would be tracked, reviewed, and reproducible.

The 26.6% reduction came from the combination of all of these. No single change got there alone.

What This Looks Like in Practice

On a £5,000/month AWS bill, a 26% reduction is £1,300 per month. That's £15,600 per year. For most SMEs, that's more than enough to cover the cost of proper infrastructure management for the entire year.

The real lesson isn't the percentage. It's that most AWS accounts running for more than 12 months have significant unnecessary spend that compounds quietly while the team focuses on building the product.

We offer a free AWS cost audit. No obligation. We'll go through your account, show you what we find, and give you an honest view of where the money is going and what can be done about it.

Book your free AWS cost audit here.