Bringing cloud spend under control without breaking production

Cloud bills grow the way technical debt does: gradually, defensibly, and with nobody clearly accountable. By the time the number reaches the board, the line items have become tribal knowledge. Reducing spend sustainably is less about discount negotiation than about making waste visible to the people who create it.

Attribute before you optimise

You cannot reduce what you cannot assign. Tag every resource with an owner, an environment and a product, and enforce the tags at provisioning time — not in a quarterly clean-up spreadsheet. Until cost has an owner per line, every optimisation conversation ends in a debate about whose team caused it.

The four recurring sources of waste

• Idle non-production capacity. Test and staging environments running at production scale around the clock for teams that use them eight hours a day.
• Over-provisioned defaults. Instance and database sizes copied from a template three years ago and never revisited against actual utilisation.
• Storage that never expires. Logs, snapshots and artifacts retained indefinitely because deletion requires a decision nobody owns.
• Cross-zone traffic. Chatty service boundaries that quietly turn every request into inter-zone data transfer.

Make cost a delivery metric

Put cost per transaction or cost per active tenant on the same dashboard as latency and error rate, and review it in the same meeting. Teams respond to visible numbers attached to their own service. A monthly finance report circulated to leadership does not change engineering behaviour; a graph in the sprint review does.

Automate the boring savings

Schedule non-production shutdowns, expire artifacts on a policy, right-size with utilisation data rather than guesswork, and reserve only what you can forecast with confidence. None of these are exciting, and together they typically account for the majority of achievable savings in the first two quarters.

Guard against the wrong cut

The fastest way to lose more money than you save is to cut observability, backups or redundancy. Set a floor for anything that protects revenue or recovery, and treat reductions below that floor as an architectural decision requiring sign-off, not a cost initiative.