Provenance: how MKDC knows each fact in your witness bundle

Why provenance has to be explicit

A compliance report that simply says a control "passes" or "needs attention" hides the more important question: how do we know that? A fact captured from a live management-plane API, a fact your operations team typed into a form, and a target your team set for next year are three different kinds of confidence — and audit committees are asked to defend all three the same way if the report does not distinguish them.

Model risk creeps in exactly here. A committee that cannot tell a measured fact from a stated intent has approved a conclusion it cannot independently defend under follow-up questions.

The provenance ladder

Every finding in the witness bundle is tagged with the provenance ladder: observed from your infrastructure, proposed by MKDC from observed structure, attested by your team, or intended as a target rather than a measured fact. A proposal is never presented as evidence — it becomes ordinary attested content only when your team accepts it.

That distinction is what lets MKDC bootstrap documentation for an estate that has never been mapped before without weakening the evidence trail for the parts that already are.

What this replaces

Manual consulting deliverables rarely separate these threads — a narrative deck states conclusions without saying whether they came from an interview, a spreadsheet someone updated eighteen months ago, or a live capture from this cycle. Reconciling that after the fact, under examination pressure, is exactly the scramble a reproducible witness bundle is designed to prevent.

Provenance is also why a control blocked on missing information is never described as a vague "gap" — it is either not yet observed, or it needs your team's attestation. Committees can act on either; neither is a mystery.

How committees and auditors use it

Fixed fee · 4–6 weeks. Advisory, not formal attestation.

• Trace any finding back to whether it was observed, proposed, attested, or intended
• Distinguish measured recovery capability from a target your team set
• Confirm a proposal was reviewed and accepted before it appears as attested content
• Re-derive the same distinctions independently from the source captures in the bundle