A FinOps Audit Checklist Borrowed From Mission Budgets

Space people already live inside budgets. Every satellite carries a fixed pool of propellant, a power curve that sags in eclipse and a downlink window that closes whether or not the data is ready. Technology teams on the ground face a looser version of the same problem, and the practice that grew up to handle it goes by the name FinOps. This checklist borrows the mission-planner’s habits and turns them into an audit you can run on any cloud bill, starting this week.

The short definition first. The Ops Atlas describes FinOps as an operating model for variable technology spend, one where engineering, finance, product, procurement and leadership look at the same timely cost data and share the right to decide what to do about it. The full write-up sits in the Ops Atlas page on FinOps; what follows is a field checklist, not a summary of it.

Before the audit: agree on the three equations

A flight controller never argues about whether fuel exists. They argue about burn rate. The Atlas gives FinOps three short equations that keep a review equally concrete.

  • Cost equals usage times rate. Every bill change is either you using more or paying a different price. Separate the two before anyone speculates.
  • Value equals outcome divided by cost. A rising bill is not automatically bad if the outcome rose faster.
  • Accountability equals data times decision rights times cadence. If any of the three is zero, nothing changes, however good the dashboard looks.

Copy them onto page one of the audit sheet. Each item below maps back to one of them.

Checklist part one: can you see the spend?

Think of this as the telemetry check. A craft you cannot hear from is a craft you cannot steer.

  1. Is every charge attached to an owner? In one worked example in the Atlas, 57.6% of the effective cost was directly tagged, and spreading a shared platform pool across teams lifted the allocated share to 89.0%, leaving 5.1% that nobody owned. Yours will come out differently; what matters is measuring the three tiers separately: tagged, allocated, orphaned.
  2. Do you know what tags cannot cover? Support fees, taxes and shared platform charges cannot be labelled at the resource level. Tagging everything is necessary and still not enough, so write down a rule for splitting the rest.
  3. Is the data fresh enough? Invoices land several days after month end, but daily data is adequate for decisions. Do not wait for the invoice to learn that something odd happened.
  4. Do billing exports exist? Some providers only backfill the current and previous month when you switch an export on, so turn it on early, the way you would start logging telemetry before launch, not after the anomaly.

Checklist part two: what is the cadence?

Mission control runs on shift handovers. The Atlas proposes a similar rhythm for cost: a daily pass for anomalies, a weekly review per team, a monthly showback and forecast, a quarterly look at commitments and a yearly budget and renewal cycle. Audit yours against that ladder.

  • Who looks at anomalies each morning, and what is the first question they ask?
  • Does every team hold a short weekly cost conversation, or only hear about money when it runs out?
  • Is there a calendar entry for commitments, or does someone remember them when an email arrives?

If the answer to any is a shrug, you have found your first finding. Close it with an owner and a first date, not a policy document.

Checklist part three: are you judging by size or by unit?

Mission planners rate a launch by cost per kilogram to orbit, not by the headline price of the rocket. FinOps asks for the same kind of ratio. The Atlas gives a plain example: a bill that grows 30% is fine if orders grew 60%. A bill that shrinks while volume collapses is not good news.

Audit questions:

  1. Is there one business unit measure per product, such as cost per transaction or per tenant?
  2. Is it tracked as a trend inside one product rather than compared across unrelated ones?
  3. Do resource-level measures, such as cost per gigabyte or per vCPU, sit next to it so engineers see their lever?

In the Atlas lab, the checkout unit cost came out at 5.39 per 1,000 orders, a number that only means something because it can be watched over time.

Checklist part four: commitments and the hourly floor

Satellite operators size solar arrays for the eclipse, not for noon sun. Commitment purchases follow the same logic. The Atlas playbook sizes them to the hourly floor of eligible usage, because unused committed hours are lost, subtracts changes already planned and buys in tranches rather than all at once.

Check for these signs of trouble:

  • Commitments sized from daily averages, which hide the overnight trough.
  • A target based on coverage alone. In the Atlas lab, raising coverage to 80.9% dragged the effective savings rate down to 6.0%, and a bigger purchase still went negative, at -9.4%, worse than paying on demand.
  • Savings quoted by adding up two different recommendation lists, which counts the same dollar twice.
  • Rightsizing done after buying resource-based commitments, leaving paid-for capacity stranded.

Checklist part five: guardrails on trimming

You can shave mass from a spacecraft only until you cut into the margin that keeps it safe. The Atlas playbook for rightsizing carries the same instinct: choose from utilisation data that includes peaks, agree on service-level guardrails first, change one dimension at a time, watch a full business cycle and roll back if any objective slips. Audit whether your last three trims followed that order.

Where to look next on this site

If the sky is your usual territory, a gentler route in is the orbital basics. Our note on why a geostationary orbit holds still above one spot and the beginner’s walk-through of how low Earth orbit works show the same budget-first thinking applied to altitude and energy.

Run the audit in one sitting

Print the lists above and give yourself ninety minutes with whoever owns the bill. Score each item as present, partial or missing, and then pick three. Do not try to fix everything; a crew that tries to burn on every axis at once ends up spinning. Pick the three, name one person for each and set a review date no more than a week away. When you come back to the lists in a quarter, compare the scores rather than the totals; the movement is the signal.

That is the whole method: see the spend, set a rhythm, judge by unit, size commitments to the floor and trim with guardrails. It reads like checklist discipline because that is what it is.