Cloud spend has a funny property: it grows with confidence. Teams that understand the infrastructure they've built tend to optimise it. Teams that don't tend to provision defensively — spinning up larger instances than they need, leaving resources running 24/7 that only need to run for 8 hours, and missing the discount options that could halve their bill.
In every cloud audit we've run, we find meaningful waste. Here's the framework we use.
Right-sizing is always the biggest win
Most workloads don't need the instance size they're running on. Cloud providers make it easy to provision a c5.2xlarge “just to be safe.” In practice, we consistently find instances running at under 20% average CPU utilisation. Right-sizing these to the next tier down typically saves 30–50% on compute alone.
The tool: CloudWatch, Datadog, or any APM tool showing 2-week average CPU and memory utilisation by instance. Sort by utilisation, start at the bottom.
Storage lifecycle policies are almost never configured
S3 and equivalent blob storage services are cheap per GB — until you have terabytes of data that hasn't been touched in two years. S3 Intelligent-Tiering and Glacier lifecycle policies move cold data to significantly cheaper storage tiers automatically. Most teams set up the bucket and never revisit the lifecycle configuration.
Reserved Instances and Savings Plans
If your baseline compute is predictable — and most production workloads have a predictable baseline — Reserved Instances or Savings Plans typically save 40–60% versus on-demand pricing. The fear of commitment is understandable, but 1-year commitments on core infrastructure almost always pay off.
Idle resources are harder to spot than you think
We regularly find: development databases left running over weekends, load balancers with no targets, Elastic IPs allocated but not attached, and snapshots from instances that were terminated 18 months ago. None of these show up as “waste” in the standard billing dashboard because each individual item is cheap. In aggregate, they add up.
The 90-minute quick win audit
Open your cloud provider's cost explorer. Filter to the last 30 days. Sort resources by cost, descending. For the top 20 items, ask: Is this correctly sized? Is it running when it needs to? Are we on the right pricing model? That exercise alone typically surfaces 15–20% savings that can be implemented in a week.