Status: current as of October 17, 2026. This page is maintained in public, and its changelog is part of the record. If a claim here ever stops being true, this page changes in the same release.
TestMyself is a six-assessment, server-scored battery (reasoning, knowledge, cognitive flexibility, technology adaptability, values under trade-off, persistence) sold as one structured signal among several in a hiring process. Scoring is deterministic, published in its rules, and replayable: any stored attempt can be re-run through the scorer and produces the identical result, and every result carries a checksum and a verification ID your team can check independently.
Under Title VII and the Uniform Guidelines on Employee Selection Procedures (29 CFR Part 1607), a selection procedure that screens out a protected group at a substantially higher rate must be justified as job-related and consistent with business necessity. The four-fifths rule (§1607.4D) is the standard screening trigger for "substantially higher": a group's selection rate below 80% of the highest group's rate calls for justification. It is a trigger, not a verdict — and not a safe harbor either.
The legal duty sits with the employer, including when the tool is ours. Agency enforcement priorities change with administrations, and guidance documents come and go; Title VII and the Uniform Guidelines do not, and neither do candidates' private rights or state-law duties. Some jurisdictions (New York City and California among them) impose additional notice, audit, or recordkeeping requirements on automated selection tools; whether a given use is covered is fact-specific, and your counsel should assess it for your situation. Our documentation posture does not depend on who is currently enforcing — it is built to the standing framework. If you use TestMyself in selection, you should:
What we contribute is structural: we collect no demographic data from candidates at all — no race, sex, age, or ethnicity is captured, stored, or inferable from our side — so there is nothing on our servers to leak, misuse, or be subpoenaed into a narrative. Adverse-impact analysis runs on your applicant data joined to our scores, and we will supply complete attempt-level data for your candidates on request, in full. If we ever detect a scoring anomaly affecting any group, it will be disclosed on this page.
Candidates who need an accommodation (extended time, assistive technology, an alternative sitting) should request one before starting; accommodated sittings are flagged in the attempt record so the result is interpreted in context. Tests are administered identically to every candidate within a sitting — same instructions, same timing rules, same randomization discipline.
We do not currently publish population norms, and nothing on this site should be read as a percentile or a comparison to a representative population. Score bands on results ("Developing / Solid / Strong") are content-referenced — they describe the score's position on the instrument's own scale, not a standing among other people. The band boundaries were set by the instrument authors against each scale's score range; a formal standard-setting study has not yet been conducted, and when one is, its method will be documented here.
Completed administrations already contribute anonymized item-level data to a growing calibration sample, and norms will be published — with the sample described (size, source, dates) next to the numbers — when the sample supports honest ones. When the first norm table appears, it will appear here first, with its sample attached. Until then, any vendor in this space who shows you a percentile owes you the same description of theirs.
We answer them in writing, and the answers become part of this page's FAQ when they recur. Contact: thegemologist@gmail.com.
Changelog: 2026-10-17 — page created. Battery scoring rules, integrity-signal definitions, and the technical manual (summary edition) are available on request and supplied to employer customers as part of onboarding.