What a 34% Cloud Bill Reduction Actually Looks Like

Cloud cost audits have a reputation for being tedious. In practice, the majority of the savings usually come from a small number of categories, and they're rarely a surprise once you see the bill sorted by service instead of by team.

On a recent logistics client engagement, orphaned resources — snapshots, unattached volumes, and load balancers nobody remembered provisioning — accounted for about 9% of the eventual savings. Nobody's fault, just the natural residue of a fast-growing infrastructure.

The larger win, at roughly 15% of the total, came from right-sizing compute. Instances had been provisioned for anticipated peak load that never actually materialized at that scale, and were running 24/7 at a fraction of their capacity.

The rest came from architectural changes: moving a batch job from constantly-running instances to scheduled serverless functions, and switching a rarely-accessed storage tier to a cheaper class. Neither required new infrastructure, just matching the infrastructure to actual usage patterns.

The pattern holds across most audits we run: look at what's idle, look at what's oversized, and look at what's running continuously that doesn't need to be. In that order, almost every time.