Verification & validation
What is established, and what is not
There are two different questions about a screening method: is the software and process sound, and does the method predict real outcomes. BiP Omega has evidence for the first and is honest that the second is not yet established.
Established — software and process verification
Continuous integration
The screening engine and its release gates run on every change, on a clean checkout, with the full test suite. A change that breaks a gate does not ship.
Deterministic behaviour
The same evidence produces the same report, byte-for-byte, on one toolchain. This is checked across operating systems on every change.
Adversarial testing of the methodology
An independent-verification-style review challenged the requirements model and the verdict logic; issues found were resolved before the model was adopted.
Security red team
The data-room ingestion boundary has been attacked with hostile files, archive bombs, prompt-injection payloads, and cross-mandate traversal attempts. Every control held; zero critical bypasses.
Workflow red team
The mandate workflow has been attacked with attempts to deliver without a sign-off, tamper with a delivered package, inject post-cut-off evidence, change a verdict after sign-off, and resolve a contradiction without a human. Every attempt fails closed.
Methodology traceability
Every material conclusion traces from the verdict to the requirement to the evidence item to the source document, and a conclusion the evidence cannot support is downgraded automatically.
Fail-closed controls
Missing decision-critical evidence produces a conditional or withheld position, never a default value. A malformed input is rejected, not silently absorbed.
Not yet established — prospective predictive performance
A predictive-performance claim would require a record of screening positions, each sealed before the outcome is known, resolved against real project events across a meaningful sample. That record is being built from real engagements. Until it has resolved a meaningful number of predictions per screening question, no predictive-accuracy, calibration, or historical-accuracy figure is published anywhere on this site.
Earlier figures once shown on this site (a retrospective benchmark set) were withdrawn: the evidence and the outcomes had been assembled together, which makes such a set susceptible to hindsight bias. That benchmark set is now used only for software regression and failure-mode testing, never for a performance claim.
A frozen holdout is also planned, to be co-signed by a genuine external second party. Until that second party is appointed and the holdout is sealed, no holdout-based figure is claimed either.
No accreditation theatre
BiP Omega displays no accreditation logos, no third-party certification badges, and no "penetration-tested by" seals, because none apply. The security controls described above are BiP Omega's own, tested by BiP Omega's own red teams, and documented for a client to review under an engagement.