Snapshot
Evidence state is pinned at publication
A later source refresh cannot silently rewrite the basis of a previously reviewed figure.
Proof
Actual Valty interfaces use illustrative data to show the product. The separate fictional worked example shows how an estimate, assumptions, evidence lineage, confidence, freshness, and claim limits can be reviewed together. Neither is a customer outcome.
Illustrative control-evidence example
Fictional record REC-EXAMPLE-01 shows the review logic for one payroll database. It is an illustrative workflow, not a screenshot, an executed Valty test or an observed customer outcome.
Source recorded
The fictional record names one payroll database and a configuration snapshot showing backups enabled. It records source, scope and capture time; it is not a restoration test.
Evidence gap open
No restoration test artifact is supplied. The reviewer keeps that gap visible instead of treating an enabled setting as proof that the database can be recovered.
Action proposed
The proposed task identifies the system owner, isolated test scope, approval, recovery criteria and due date. No change or test is represented as executed.
Verification pending
The reviewer would need the executed test record, observed result, source and time, and a check against the agreed criteria. Until then, the gap remains open.
Evidence integrity answers whether a record is attributable and unchanged. Control effectiveness requires relevant validation; third-party acceptance is a separate decision. This example establishes none of those outcomes and assigns no loss-reduction credit.
The actual Valty interface below contains illustrative data and is separate from this fictional record. The financial worked example separately shows how a funding decision can retain its assumptions and claim limits.

Inside the proof pack
Reproducible proof
The Proof Pack carries more than a source list. It binds the displayed figure to an evidence snapshot, serving model version, public prior rows, and the publication state that controlled the export.
Snapshot
A later source refresh cannot silently rewrite the basis of a previously reviewed figure.
Model card
Public prior rows cite study, year, sample, and license; dormant licensed sources cannot silently contribute.
Recompute
A mismatch blocks the pack from claiming reviewed and writes the result into provenance.
Unsupported
Blocked claims and unsupported companies remain visible with the next evidence action needed to advance them.
Replay chain
This page shows the review grammar with static example data. Replay and node-level challenge operate inside entitled workspaces and reviewed exports.
Proof matrix
See which claims are ready to share, which need access controls, and which stay blocked until the evidence is current.