FinOps

WHAT IT IS. FinOps is the operating discipline that brings financial accountability to cloud spend: engineers see the cost of what they run, finance sees where the money goes, and neither side needs a spreadsheet war to get it. The FinOps Foundation's framework (linked below) organizes it as a loop — inform (allocate and show every dollar), optimize (act on what the data says), operate (govern, review, repeat as a culture, not a project). The output that matters is unit economics: cost per customer, per transaction, per deploy — a number the business can reason with. WHY IT EXISTS. Cloud converts capex into a metered utility that engineering controls with every terraform apply — but the invoice arrives monthly, after the decisions, addressed to people who did not make them. Without allocation, nobody owns waste; without anomaly alerts, a forgotten NAT gateway or a runaway retry loop burns quietly until the invoice lands. FinOps closes the loop by putting price and usage next to the people whose choices move them. WHEN TO USE IT. The moment the cloud bill is big enough that a surprise in it would be a business event — or when growth makes spend scale faster than headcount. Below that threshold, a monthly bill review is honest and sufficient. FinOps capability should arrive before the pain: allocation retrofitted onto an untagged estate is a quarter of archaeology. WHO OWNS IT. Not finance, and not a lone "cloud cost person." Engineering owners per cost center (the service teams whose choices move the spend), a FinOps practitioner running the cadence and the tooling, and finance as a partner defining what good unit economics means. The practitioner's job is the loop's rhythm — allocation reports, anomaly triage, commitment reviews — not being the company's conscience about spend. WHAT GOOD LOOKS LIKE. Cost allocation coverage is effectively total: every dollar maps to a team and service through tags and account structure enforced at creation (that enforcement is the /cloudops guardrail doing FinOps work). Anomalies alert within hours, not at invoice time. Rightsizing and idle-resource cleanup run as a recurring cadence with named owners. Commitment coverage (savings plans, reserved capacity) is managed deliberately — neither 0% nor a bet on growth nobody re-examines. Unit cost is on the product dashboard next to usage, and it trends down while the product grows. HOW TO IMPLEMENT. Sequence matters: allocate first — enforce the tag/account standard on new resources and backfill the top spenders until coverage is real. Then alert — anomaly detection on daily spend per service. Then optimize — rightsizing and idle cleanup as a scheduled review. Commitments come last, on stable, understood usage; a discount on waste is still waste. HOW TO ASSESS YOURSELF. The production readiness assessment scores cost visibility and cost governance as operational capabilities: /assess. The maturity ladder below runs from a monthly invoice surprise (L1) to unit-economics-steering-investment (L5). Related: /cloudops enforces the allocation structure, /platform-engineering bakes sensible defaults into the golden path so the cheap route is the default one.