A service-band audition has to be fair, and it has to be provably fair when someone asks. Cadenza produces the record behind that answer: a reproducible random draw, a printable Fairness Certificate that cites the recorded seed, blind screening, aggregate equal-opportunity and adverse-impact reporting, and a full data-egress audit trail. Every piece below maps to a mechanism that ships today — nothing here is aspirational.
Five controls, each independently verifiable, each backed by a real mechanism in the product rather than a policy statement.
Audition order is set by a seeded random draw, not by hand. The draw record — the seed plus the canonical candidate and slot lists — is append-only, and anyone holding it can re-run the published algorithm and obtain byte-identical assignments. Any alteration of the recorded order fails verification, so tampering is detectable rather than deniable.
Each round produces a downloadable, printable Fairness Certificate. It records the draw seed, the independent-verification result, who ran the draw and when, and how many candidates went into how many slots — the facts a reviewer needs to confirm the order was determined by the draw and not by a person.
When blind review is on, applicant identity is masked from the reviewers themselves until a decision is recorded — reviewers screen on the audition material only. Masking is enforced server-side: identity fields never leave the API for a masked application, so a reviewer cannot read them from the network tab.
Applicant self-ID is reported only in aggregate, with any bucket below the small-cell threshold suppressed to a "<5" marker so a report can never re-identify an individual. On top of that aggregate, Cadenza computes a four-fifths (80%) adverse-impact analysis to flag where a group is selected at less than four-fifths the rate of the strongest group.
Every fairness-pack, Fairness Certificate, and scored-CSV export is recorded as an audit event capturing which admin took the data, when, and from which IP address. An institution can show an inspector exactly who exported an applicant dossier.
The same five steps run every round, and each one leaves an artifact you can hand to a reviewer.
Candidate and slot lists are canonicalized, then shuffled with a seeded, dependency-free PRNG. The draw record is stored append-only — a re-run never edits history, it supersedes the prior draw with a recorded reason, and every superseded draw is retained.
With blind review on, reviewers see the audition material but not the applicant identity until a decision is recorded. The masking is server-side, so identity cannot leak to the browser during screening.
Produce the printable certificate for the round. The government/federal variant adds an attestation block built only from facts the system holds — the recorded seed, the verification result, the operator, and the timestamps — with no external-standard conformance claimed.
Run the aggregate EEO report (small-cell suppressed) and the four-fifths adverse-impact analysis over the applicant pool and the selected subset — an informational screen for spotting potential adverse impact, not a certified OFCCP/EEO filing.
Every export is logged with the admin, timestamp, and IP. If anyone asks who pulled a dossier and when, the audit trail already has the answer.
The same record serves three readers — the office running the audition, the office reviewing it for equal opportunity, and whoever asks about it afterward.
You need to stand behind the result. The certificate and the reproducible draw let you show the audition order was random and unedited — a fact anyone can re-verify from the recorded seed, not a claim they have to take on faith.
You need the applicant-flow picture without re-identifying anyone. The aggregate EEO report with small-cell suppression and the four-fifths adverse-impact analysis give you a defensible, privacy-preserving read of selection rates across groups.
When the order or the outcome is questioned, the record answers. The Fairness Certificate (government/federal variant), the tamper-evident draw, and the egress audit trail form a coherent file of how the order was set and who handled the data.
This page describes only the fairness and auditability controls that ship today. Cadenza does not currently hold a FedRAMP authorization, a SOC 2 report, a NIST 800-53 attestation, or a Section 508 conformance certification, and does not yet offer SSO/CAC sign-in or a signed DPA out of the box. The adverse-impact analysis is an informational screen, not a certified OFCCP or EEO filing, and the Fairness Certificate attests only to facts the system records — it makes no representation of conformance to any external standard. We would rather tell you that up front than overstate our posture.
Running a formal evaluation? We are glad to walk through our security questionnaire and data-processing terms with your team.
By a seeded random draw, not by hand. The draw record — the seed and the canonical candidate and slot lists — is stored append-only. Anyone holding it can re-run the published algorithm and reproduce byte-identical assignments, and any alteration of the recorded order fails verification.
A downloadable, printable document produced for an audition round. It cites the draw seed, the independent-verification result, who ran the draw and when, and the candidate and slot counts. A government/federal variant adds an equal-opportunity/inspector-general-defensible attestation block built only from facts the system records.
Yes. Applicant self-ID is aggregated with small-cell suppression (buckets under the threshold show as "<5"), and Cadenza computes a four-fifths (80%) adverse-impact analysis across the applicant pool and the selected subset. It is an informational screen for spotting potential adverse impact, not a certified OFCCP or EEO filing.
No. Cadenza does not currently hold a FedRAMP authorization, a SOC 2 report, a NIST 800-53 attestation, or a Section 508 conformance certification, and does not yet offer SSO/CAC sign-in or a signed DPA out of the box. We describe only the fairness and auditability controls that ship today. We are glad to walk a formal evaluation through our security questionnaire and data-processing terms.
Explore the full military program and its security & data posture, or ask our team for the security questionnaire your evaluation needs.
Cadenza is an independent platform and is not affiliated with, endorsed by, or sponsored by the U.S. Department of Defense or any branch of the U.S. Armed Forces. The controls described here are product features; they are not legal advice and do not by themselves establish compliance with any law, regulation, or agency standard.