A FinOps review that starts from your own Cost Analysis export, not from generic advice about turning things off.
Start from the bill, not from advice
Generic cost advice — right-size your VMs, use reserved instances, delete unused disks — is correct and almost useless, because it does not tell you which of those applies to you or what it is worth.
Every review we run starts with your actual Cost Analysis export, grouped by resource group, service and tag. Within an hour that usually tells you that a small number of resources account for most of the spend, and that a long tail of forgotten ones account for more than anyone expected.
“Most cloud waste is not expensive mistakes. It is cheap mistakes nobody has looked at in two years.”
The recurring offenders
Orphaned managed disks left behind when VMs were deleted. Non-production environments running at production size, around the clock, including weekends. Over-provisioned App Service plans sized for a launch-day spike that never recurred. Premium storage tiers on data nobody has read in a year. Log Analytics ingesting verbose debug logs at full retention.
None of these are architectural errors. They are the residue of normal delivery, and they persist because nobody owns looking for them.
Tag first, or you are guessing
If resources are not tagged by owner, environment and service, you cannot tell what is safe to change, and every proposed saving turns into a long email thread.
Enforce tagging with Azure Policy so untagged resources cannot be created, then backfill the existing estate. This is dull and it is the prerequisite for everything else: once spend is attributable, teams start managing their own, which is more durable than any one-off cleanup.
Commitment discounts, carefully
Reserved instances and savings plans offer real discounts for committing to a term. The failure mode is committing to your current shape rather than your steady-state shape — locking in over-provisioned sizing for a year, or reserving capacity for a workload you are about to re-architect.
Right-size first, run at the new size long enough to be confident, then commit to the baseline only. Keep the variable portion on demand.
Make it stick
A one-off cleanup decays. Within a few months the same patterns reappear, because nothing changed about how resources are created.
Put a budget with alerting on every subscription, add anomaly detection so an unexpected jump is noticed in days rather than at month end, and review the top movers on a schedule. The goal is not a heroic reduction once; it is spend that stays proportional to what the business is actually doing.
Where ITLabz can help
If you’re tackling problems like these, our engineers have built them in production across lending, fintech, logistics, manufacturing and retail. We’d be glad to compare notes.