Cloud billing is opaque by design. Providers profit from your confusion. But with an hour of detective work, you can find where your money is going and stop the bleeding.
The Cloud Bill Investigation Framework
Step 1: Get the raw data
Export 90 days of billing data. You need daily costs by service, resource-level costs, tagged costs, and data transfer costs. Data transfer is usually the hidden killer.
Step 2: Find the top 5 cost centers
Sort descending. The top 5 resources typically eat 80% of the bill. Everything else is noise for this pass.
Step 3: Classify each cost center
- Owned & needed — keep, but check rightsizing
- Orphaned — created for a test project, never deleted. This is where most money hides.
- Duplicated — staging running the same instance size as prod
- Zombie storage — snapshots, logs, and volumes nobody has opened in 60 days
Step 4: Ask the three questions
- If I deleted this today, who would notice? (If nobody: delete, don't "pause")
- Does this resource's size match its actual 30-day p95 load?
- Is this bill line flat-growing while usage isn't?
Step 5: Set tripwires so it never happens again
Budget alerts at 50%, 80%, 100%. A weekly 10-line script that emails the top 5 resources beats any dashboard nobody opens.
What this looks like in practice
A 3-person team I worked with found $740/month of orphaned test resources this way. The fix took 90 minutes. The billing export took 5.
I sell small operational products for solo founders and tiny teams — runbooks, automation kits, cost-audit scripts. The lineup: Hive80Lab store. Launch code LAUNCH50 takes 50% off everything this week.
If you want a free 1-line read on your own cloud bill, paste your top 5 services in the comments.
Top comments (0)