This page states, plainly, how much empirical evidence stands behind The ArcProof Verification Framework today.
ArcProof has published a methodology. ArcProof has not yet published an empirical validation study. There are no third party validity studies of the ArcProof rubric, no peer reviewed papers authored by ArcProof, and no longitudinal outcome data, because the credential population needed to produce them does not yet exist.
The framework's own Stage 4 treats predictive validity as an empirical claim to be tested over time rather than asserted at issuance. Applying that standard to ourselves means saying so here rather than implying evidence we do not have.
What ArcProof does claim today: constructs are defined before scoring, evidence is performance based, scoring is applied consistently against a fixed rubric, and each issued credential carries a tamper evident record that a third party can trace. Those are properties of the process and are checkable now. Predictive power is not yet among them.
The rubric's weights and internal logic remain proprietary. That is a deliberate tradeoff: it protects the assessment against gaming, and it means external validity evidence has to come from outcome studies rather than from rubric inspection. We treat that as an obligation, not an excuse.
Categories follow the Standards for Educational and Psychological Testing (AERA, APA and NCME, 2014).
AERA, APA and NCME (2014), evidence based on test content
Constructs are defined against occupational demand before any score is issued. No independent content validity panel has yet reviewed the construct map.
AERA, APA and NCME (2014), evidence based on response processes
Tasks are performance based rather than recognition based. No think aloud or process tracing study has been conducted.
AERA, APA and NCME (2014), evidence based on internal structure
Reliability and dimensionality analysis requires a sufficient volume of scored attempts. That volume does not exist yet.
AERA, APA and NCME (2014), relations to other variables
Predictive validity against hiring, advancement, earnings and retention is Stage 4 of the framework. No outcome data has accrued.
AERA, APA and NCME (2014), fairness in testing
Differential item and score functioning analysis requires demographic and performance data at scale, with consent. Not yet available.
Messick (1994), interplay of evidence and consequences
Consequential evidence depends on documented decisions made using ArcProof credentials by employers and programs.
Each study below is planned, not completed. Prerequisites are stated so a reader can judge how far away the evidence is.
Design: External subject matter panel reviews the construct map and task specifications for a single occupational family, rating role relevance and coverage.
Output: Published content validity report, including panel composition and disagreements.
Prerequisite: Three or more independent reviewers per occupational family.
Design: Blind double scoring of the same evidence by trained human raters and by the automated scorer, reporting agreement statistics and disagreement patterns.
Output: Reliability report with agreement statistics and error analysis.
Prerequisite: A minimum sample of scored performance evidence and funded human raters.
Design: Longitudinal cohort study linking verified capability profiles to hiring, advancement, earnings and retention outcomes, with a pre registered analysis plan.
Output: Predictive validity report, including null and negative findings.
Prerequisite: An employer or workforce board partner willing to share outcome data under a data use agreement.
Design: Analysis of score distributions and item functioning across demographic groups, with consented data and an independent statistician.
Output: Fairness report and any resulting rubric or task revisions.
Prerequisite: Sufficient consented sample size per subgroup.
The studies above are blocked on data that only institutions hold. If you run a workforce program, a community college, or a hiring function and you can share outcome data under a data use agreement, we will run the study jointly and publish the results, including findings that do not favor ArcProof.
Partners keep the right to review findings before publication and to be named or to stay anonymous.
ArcProof Research Institute is a research initiative of SkillUpArc LLC. It is not a separate legal entity and is not an accredited institution.
Trademark & claims notice
"ArcProof"™, "The ArcProof Verification Framework"™, "ArcProof Research Institute", and related product names are unregistered trademarks used by SkillUpArc LLC to indicate intent of common law trademark rights. The ™ symbol denotes intent of use and does not represent a registered trademark with the USPTO or any other trademark office. Nothing on this site constitutes legal, financial, or compliance advice. Product descriptions, verification methods, and outcome claims are marketing summaries and may evolve. For authoritative information about credentials, integrity protocols, or licensing, contact us directly.
We use cookies to improve your experience and analyze site usage. By continuing, you accept our use of cookies. Learn more in our Privacy Policy.
Founding member pricing: ArcProof Pro at $29/month or $199/year (≈$17/mo). Your founding rate stays the same for as long as your subscription stays active.
See pricing