Evidence before claims.

A claim that cannot be traced to observable conditions is difficult to govern and harder to assure.

Institutions routinely make material representations: a system is secure, a model is reliable, a reserve is sufficient, a process is compliant, a vendor is controlled, a decision was independently reviewed. Each statement contains an implicit evidentiary burden.

Evidence architecture makes that burden explicit. A material representation is decomposed into the conditions that would make it true, the controls that produce or preserve those conditions, the evidence that records their operation, the test that can challenge them, and the consequence when the evidence does not support the claim.

This is more than documentation. Evidence has its own integrity requirements. Source, time, version, access, custody, retention, reproducibility, and known limitations can determine whether a record supports the decision for which it is later used.

The result is a system in which institutional confidence follows evidence rather than precedes it.