Cloud Cost Optimisation

Where the money usually goes

In our experience the largest items are rarely exotic: oversized database instances chosen during a launch and never revisited, non-production environments running twenty-four hours a day, snapshots and logs with no expiry policy, and cross-zone traffic nobody realised was billable.

None of that requires re-architecting. It requires someone to look, with data.

How we approach it

We measure before changing anything — usage percentiles over weeks, not a snapshot — because right-sizing off a quiet Sunday is how you cause an incident and lose the team’s trust in the whole exercise.

Then we work in order of saving-per-risk: lifecycle policies and idle environments first (no user impact), right-sizing next (measured, with headroom), commitments last, once the baseline is stable enough to commit to.

The part people skip

Attribution. Until each team can see what it spends, cost is nobody’s job. Labelling resources properly and putting a per-team figure in front of engineers changes behaviour more than any single technical fix — and it keeps the savings from quietly eroding over the following year.