In short
The pattern is common: the shop was set up for a launch, traffic settled, and the setup never changed. The bill is steady and nobody has read it line by line.
Who: A fictional shop with one web store, a staging copy and a nightly export job, run by two people. Illustrative scenario: the numbers show what a typical engagement of this kind looks like, not a measured client result.
Where it started
- Three virtual machines run all day: the store, a staging copy and one that runs a nightly export for ten minutes.
- The managed database has a read replica that nothing reads from and storage several times larger than the data.
- Daily disk snapshots have been kept since the first day with no retention rule.
What was done
- Read the bill line by line with the owners and tagged every resource with a purpose.
- Moved the nightly export to a scheduled function that runs, finishes and stops.
- Scheduled the staging copy to switch off outside working hours and shrank it.
- Removed the unused replica after a decision from the owners, right-sized storage from measured use, and added a snapshot retention rule.
- Set budget alerts and a monthly report.
What changed
- The bill follows use: it drops at night and at weekends. The owners can read it, because every line has a purpose.
- Nothing visible changed for customers; the store, the backups and the recovery plan are intact.
The numbers
| What | Value | How it was measured |
|---|---|---|
| Savings | Depends on how much was idle | illustrative; use the calculator for a range from your own numbers |
Source: Illustrative scenario built from the problems listed on the cloud FinOps page. No client, no measured figures.



