0Independent access
Your own read access to the real systems — client consoles, source repos, live queries. Never evidence mediated entirely through the client's own export or script output.
1Rebuild context from zero
Independently trace assets, data flows, and scope — don't start from the client's own system-description document as ground truth.
2Verify every control claim
For each "Implemented" claim: documentary evidence AND a live/runtime test AND, where possible, an interview — triangulate, don't accept one source.
3Independent risk scoring
Score likelihood/impact from what you found in Phase 2, before reading the client's own risk register. Reconcile after, not before.
4External artifact collection
Signed contracts, insurance policies, console configurations — things no repository or export can produce. Get the originals.
5Resolve legal exposure empirically
Don't ask whether regulated data touched an unauthorized system — query it and check.
6Delta report, not adoption
Your own findings first; diff against any client self-assessment second. The self-assessment is a cross-check on your work, not the source of it.
7Re-verify remediation
"Done" gets tested in production, by you, before it's closed in the final report — not accepted on the client's word.