Cutting your cloud bill: a five-step audit plan

A cloud bill rarely grows because of one big mistake. It grows through hundreds of small defaults: instances sized for a launch that never came, logs nobody reads, storage that never expires. An audit finds them in a fixed order.

The five steps

1. Get the data. Export billing with resource-level detail and tag coverage. Untagged spend is the first finding.
2. Map spend to owners. Every euro should belong to a team, a product or an environment. Shared costs get a split rule.
3. Right-size. Compare requested CPU and memory with actual use over at least two weeks. Over-provisioned nodes and databases are usually the biggest item.
4. Fix storage and data transfer. Lifecycle rules for logs and backups, cheaper storage classes for cold data, and a look at cross-region and egress traffic.
5. Commit to the baseline. Once the stable part of the load is known, cover it with commitments or reservations, and leave the spiky part on demand.

What not to do

• Do not buy commitments before right-sizing. You lock in the waste.
• Do not cut monitoring and backups to save money. Make them cheaper instead.
• Do not optimize once. Without budgets and alerts, the bill drifts back within a few quarters.

The result of an audit is a ranked list: each change with its expected saving, its effort and its risk. Most teams can implement the top of the list in a few weeks.