How do you optimise cloud costs?

Direct answer

What a strong answer covers

How it works

Cost work follows the same loop as performance work: attribute, find the biggest line, fix it, measure. Tags and account structure make every dollar attributable to a team and a service. The usual findings are idle resources such as unattached volumes, old snapshots, and forgotten environments; instances sized for a peak that never comes, corrected from utilisation data; workloads that could autoscale or run on spot; storage kept in the hot tier forever; and steady-state compute that would be cheaper under a commitment. The metric that keeps the work honest is unit cost, such as cost per thousand requests or per customer, tracked over time.

Example

A monthly review found 18 percent of spend in unattached volumes, snapshots older than a year, and a staging environment running 24/7; another 22 percent in instances averaging 12 percent CPU. Cleanup, right-sizing, an off-hours schedule for staging, and a one-year commitment for the steady database tier cut the bill by 35 percent, and cost per thousand requests became a panel on the platform dashboard.

Common mistakes

• Buying reserved capacity for a workload that is about to be re-architected.
• Right-sizing from a single day of metrics and missing the monthly peak.
• Cutting costs with no attribution so nobody notices when they creep back.

A useful answer structure

Answer from your own work.

Do not memorise a script: use only examples you can explain honestly in follow-up questions.