Learners and working engineers keep asking the same thing: how do I know whether my team’s compliance practice is healthy, or just survivable? This checklist is built for that question. It turns the ideas behind ComplianceOps into twenty yes-or-no checks you can run against your own organisation in about an afternoon, with notes on what a “no” usually means and where to read more. It is general educational material, not legal or audit advice.
How to use it
Score each item yes, partly or no, and write down the evidence for every yes. If you cannot point to a dated record, mark it “partly” at best. The principle comes from the Ops Atlas page on ComplianceOps as a standing operation: a claim without a fresh record behind it counts as missing.
Section A: Obligations and ownership
- Is there a written register of the obligations you actually answer to? Customer contracts, card rules, sector regulations and internal policies should all appear.
- Does every requirement have a named owner? Shared ownership usually means none.
- Is scope written down? For payment-card work in particular, the careful assumption is that every system sits inside the boundary until somebody proves it does not.
- Is it clear what is not your job? Risk appetite, privacy requests, building safeguards and legal opinions belong to neighbouring teams.
Section B: The control library
- Do you keep one library rather than one list per framework?
- Does each control map to every requirement it supports?
- Does the mapping preserve intervals? If one standard says six months and another says annually, the note should say so, not just “covers two”.
- Is each control testable by a stranger from the written wording?
Section C: Evidence quality
- Does each record show source, timestamp, method, test and freshness rule?
- Is out-of-date evidence reported as missing?
- Are old records superseded rather than edited?
- Do tests check behaviour, not only that a policy exists? A document is not enforcement.
- Is machine-generated text kept out of the evidence store? A summary is not a record of an event.
Section D: Exceptions and gaps
- Does each exception name the exact failing resources?
- Does each have an end date, and does an expired one count as a failure?
- Is renewal blocked unless a compensating safeguard is recorded?
- Is there a runbook for missing evidence that forbids back-dating? It should require a retrospective label and disclosure.
Section E: Reporting and rhythm
- Do reports show untested controls next to the pass rate?
- Do you know your fresh-evidence percentage and your automated percentage?
- Do checks run on a schedule independent of the audit calendar?
Common reasons for a “no”
Most failing answers trace back to a small set of causes. Teams built the process around the assessor’s calendar instead of their own systems. Separate people look after separate frameworks and never compare notes. Exceptions were renewed for convenience. Tests were written against documents rather than live configuration. Or evidence collection began before the collectors were actually connected, leaving the start of the period with nothing behind it.
Spotting which cause applies to you is more valuable than the score itself, because each one has a different fix. Calendar-driven work needs schedules owned by the system team. Framework silos need a single library. Exception drift needs end dates that bite. Narrow tests need rewriting. Premature periods need a rule that nothing officially starts until every collector reports fresh data.
Reading your score
Count the yes answers. Fifteen or more suggests a routine that is already running; between eight and fourteen means the foundations exist but upkeep is uneven; below eight points to the scramble pattern, where most effort happens just before an assessment. These bands are rough guides rather than official thresholds, and your own context matters more than any number. Repeat the exercise every quarter, keep the previous sheets, and look at the direction of travel rather than a single result.
Turning the results into a plan
Resist the urge to repair every gap in one sweep. The Atlas offers a three-stage ninety-day shape you can borrow. In the first month, appoint an owner, finish the obligation register and take on one framework. In the second, wire up collectors and write tests covering your top ten controls, and start an exception register. In the third, start the formal evidence period only when everything it relies on is fresh, run one deliberate failure to confirm it would be detected, and map a second framework to the same library.
Skills worth practising while you do this
The checklist doubles as a training ground. Turning vague requirements into measurable tests builds the habit of structured thinking that the piece about building a sharper analytical mindset for solving DevOps problems describes. And if you are mapping where this fits in a larger programme of change, the roadmap in practical skills and strategy for IT transformation gives the wider context. Tool-wise, cloud-native policy checks such as Azure Policy and open-source scanners such as Prowler are common starting points, though the Atlas suggests proving that your control library can be exported in full before you sign up with any platform.
One last caution
A high score does not mean you are secure. The Atlas recalls a 2013 retailer that held a recent card-industry assessment and still lost payment data. Treat this list as a way to see your own habits clearly, then continue to test whether the controls themselves are the right ones.

