Verityx evaluates software delivery claims by correlating work items, code changes, pull requests, test evidence, and deployment signals across connected systems. The goal is not to replace delivery governance or make payment decisions automatically; it is to surface reviewable findings with clear evidence chains and defined limits.
Verityx is designed to help technology leaders review whether reported delivery activity is supported by observable engineering evidence. It is most useful in vendor assurance, invoice review, delivery dispute resolution, and internal governance where reported status and verifiable delivery may not always align.
Verityx currently evaluates evidence from sources such as:
Verityx does not rely on a single signal. Findings are based on correlation across multiple evidence layers:
Story state, acceptance criteria, linked artefacts.
Commits, files changed, branch and merge activity.
Pull requests, approvals, review timestamps.
Test artefacts, linked checks, acceptance proof.
Release or deployment indicators where available.
A delivery claim is marked substantiated when the connected evidence chain materially supports the reported work. Typical signals include traceable work-item linkage, corresponding code change, merged review activity, and supporting validation or deployment evidence where expected.
A delivery claim is marked partially substantiated when some supporting evidence exists but the chain is incomplete, weak, or ambiguous. Examples include code activity without clear acceptance evidence, or work-item completion without strong deployment linkage.
A delivery claim is marked unsubstantiated when the reported completion status is not supported by sufficient observable evidence in the connected systems. This does not mean fraud or misconduct; it means the claim could not be adequately supported from the available evidence set.
A finding is marked needs review when evidence is contradictory, incomplete, or likely to require human judgment. This category exists to avoid false certainty where delivery signals do not support a reliable automated classification.
Human review is more likely to be recommended when:
Verityx uses weighted evidence correlation rather than a single keyword or ticket-reference rule. A finding is only surfaced as a stronger classification when the observable signals meet the relevant confidence threshold for that evidence category; where confidence is lower or ambiguous, Verityx prefers partial substantiation or human review over overclaiming certainty.
Verityx does not:
Like any evidence review system, Verityx works within the quality and completeness of the connected data. Missing integrations, weak change management practices, manual releases, undocumented testing, or poorly maintained work items can reduce traceability and increase the number of findings that require human review.
Verityx is designed to support human decision-making, not replace it. Findings should be reviewed in operational context, especially where there are commercial consequences, disputed delivery interpretations, or incomplete source access.
This methodology may be refined as new integrations, evidence sources, and validation approaches are introduced. Material changes will be versioned and dated on this page so customers can understand which assessment model applied to a given review.
Current version: 1.0 · 4 May 2026
For questions about our methodology: methodology@verityx.io