Cloud infrastructure is usually tracked in a spreadsheet at first: a list of servers, what each is for, roughly what it costs. It works for a handful of resources and one person who provisioned all of them.
It stops working quickly, and the reason is structural rather than a failure of diligence.
Why manual tracking fails
Resources appear faster than anyone records them
The entire point of cloud is that anyone with access can provision in seconds. Every such action should produce a spreadsheet row and most do not — not through carelessness, but because the person was solving a problem, not maintaining an inventory.
The bill is retrospective
You find out what something cost after it has cost it. A test environment left running over a long weekend is discovered when the invoice arrives.
Costs are not where you think
The compute line is visible and expected. What surprises people is the rest: storage that accumulates because nothing deletes old snapshots, data transfer between regions or out to the internet, load balancers and IP addresses attached to nothing, and backups retained indefinitely by a default nobody chose.
A spreadsheet listing servers captures none of this.
Nobody owns anything
Six months in, there is a resource nobody recognises. It might be important. Nobody will turn it off, so it runs forever. Every cloud account accumulates these.
What to do instead
Tag at creation, and enforce it
Every resource tagged with an owner, an environment and a purpose. Tags applied later never get applied.
Where the platform supports it, make untagged creation fail. That sounds heavy-handed and is the only thing that works — a policy without enforcement produces tagged resources for about three weeks.
Use the provider's own cost tools
Every major platform includes cost analysis, budgets and alerts, at no extra charge and with better data than you can assemble manually. Set a budget alert on day one, at a threshold that would concern you.
Schedule non-production environments off
Development and test environments rarely need to run overnight or at weekends. Scheduled shutdown is straightforward and removes a large share of typical waste, because idle non-production capacity is where most of it sits.
Review monthly, with the owner present
Thirty minutes: largest line items, anything new, anything untagged, anything whose owner cannot explain it. Do it with the people who provision things, not only with finance — finance can see the number but not whether it was necessary.
Set a retention policy for snapshots and backups
These accumulate silently and are almost never reviewed. A stated retention period, enforced automatically, prevents the slow accumulation that turns storage into a major line over a couple of years.
Where a spreadsheet still helps
Not as an inventory — the platform knows better than you do. But as a record of decisions: why this instance size, why this region, what this environment is for, what the plan is for the thing everyone knows should be decommissioned.
That context does not live in any cloud console, and it is what makes next year's review possible.
About the author. Kislay Chandan has worked with schools since 2014. Rather than gathering requirements from the outside, he worked inside a school to see what actually goes wrong day to day, and built what he found into the software. He still works with Shemrock Glitters in Patna — most changes to the product start with something that came up there.
Gallery
