FinOps in Practice: Cutting Cloud Bills by 35% Without Touching Performance

Cloud bills rarely grow because of one bad decision — they grow because dozens of reasonable-at-the-time decisions never get revisited. For one client running a consumer platform on AWS, that accumulation added up to £1.1M a year. A structured FinOps audit brought it down by 35% — roughly £400k annually — without a single user-facing performance regression.

Where the Money Actually Goes

Before touching anything, we mapped spend against usage: which instances were provisioned for a peak load that happened twice a year, which services had been resized upward during an incident and never resized back down, and which storage volumes hadn't been read from in months. The audit alone surfaced most of the eventual savings before a single configuration change.

The Three Levers That Moved the Needle

• Right-sizing computeMatching instance types to actual CPU/memory utilisation rather than historical guesswork — most fleets run 30-40% oversized once measured properly.

• Spot and reserved capacityMoving fault-tolerant batch and background workloads to spot instances, and committing predictable baseline load to reserved capacity, at a fraction of on-demand pricing.

• Autoscaling policy tuningReplacing static scaling thresholds with policies driven by actual request-queue depth, eliminating both over-provisioning during quiet periods and scrambling during real spikes.

The Audit Process We Use

• Baseline the current cost and utilisation across every service, tagged by team and environment.

• Tag and allocate spend so cost ownership is visible to the teams actually generating it.

• Identify waste — idle resources, oversized instances, and orphaned storage — and quantify the savings before changing anything.

• Implement guardrails — budget alerts and tagging policies — so savings don't silently erode over the following quarters.

Results After 90 Days

Keeping Savings Sticky

The hardest part of FinOps isn't the first round of cuts — it's making sure spend doesn't creep back up six months later. Budget alerts tied to team-level tags, and a quarterly review baked into the engineering calendar, are what made this client's savings durable rather than a one-time event.