Cloud cost is an architecture decision, not a purchase
Your Azure invoice is the consequence of technical choices made months earlier: the oversized VM someone provisioned 'to be safe', premium disks in a test environment, backups retained indefinitely. Renegotiating the contract without revisiting those choices achieves little.
The good news is that most waste concentrates in a handful of resources, and fixing it does not require stopping anything.
The four quick wins
Right-sizing: machines sized by guesswork usually run below 20% CPU. Reducing size based on 30 days of real metrics is the fastest and lowest-risk win.
Reservations and savings plans: workloads that run permanently should not be paid on demand. One- or three-year commitments substantially reduce the price of the same machine with no technical change.
Scheduled shutdown: development, staging and training environments rarely need to exist outside business hours. Scheduling start and stop cuts close to two thirds of their billed time.
Orphan cleanup: unattached disks, reserved unused public IPs, old snapshots and leftovers from closed projects appear in every environment older than a year.
After the quick win comes governance
One-off savings creep back within months without process. Three controls sustain the result: mandatory tagging by owner and cost center, budgets with alerts per subscription, and a monthly consumption review with the technical owners.
Without tagging nobody knows who owns a resource — and what has no owner gets optimized by no one.
How we do it
We analyze 30 days of consumption, rank resources by opportunity and present a plan in waves: first what saves money with no risk, then what needs a change window. You approve wave by wave.
We also implement tagging, budgets and alerts so the savings hold in the following months.