Test Partnership
HireAIScore rates Test Partnership 50 out of 100 (grade F, Substantial gaps) under rubric v1.0, last reviewed September 25, 2026. Test Partnership ranks 22 of 82 vendors rated in Assessment AI, against a category average of 46. That score is drawn from 32 evidence items across 7 rubric criteria for Test Partnership — 26 with a cited source, 6 recording that nothing was located.
Score breakdown
Test Partnership, criterion by criterion
Each rubric criterion is scored out of 100 on the cited evidence, then weighted into the total.
Weighted contribution shown to the right of each bar.
Article 11 Technical Documentation
Weight 20% · category avg 50
57
+11.4
Bias Audit Transparency
Weight 18% · category avg 41
48
+8.6
FRIA Support
Weight 15% · category avg 31
36
+5.4
Data Governance Disclosure
Weight 15% · category avg 53
57
+8.5
Human Oversight Design
Weight 12% · category avg 55
58
+7.0
Post-Market Monitoring
Weight 12% · category avg 38
33
+4.0
Customer Documentation
Weight 8% · category avg 58
62
+5.0
Total weighted score
49.9(rounded to 50 for display)
Category avg is the mean raw score on that criterion across the 82 Assessment AI vendors in scope of this rubric, this one included.
Strongest criterion
Customer Documentation
62raw / 100
Contributes 5.0 to the total. 62 against a 58 category average.
Weakest criterion
Post-Market Monitoring
33raw / 100
Contributes 4.0 to the total. 33 against a 38 category average.
Cited evidence
The evidence behind the score
Every score is backed by at least one cited piece of evidence. Open a criterion to see what was found, when, and where.
Download diligence recordEvidence ledger
- Items
- 32
- Documentation
- 24
- Public statement
- 1
- Product feature
- 1
- Absence
- 6
- With a source URL
- 26 of 32
- Source hostnames
- 3
- Fewest items
- 3
- FRIA Support
These figures measure how thoroughly Test Partnership was reviewed, not how Test Partnership performed — an absence row, recording that nothing was located, is counted like any other item.
Article 11 Technical Documentation
5 items57
- DocumentationCaptured Sep 25, 2026
The public 170-page MindmetriQ Technical Manual (v2.0, 22 August 2019), linked from the product page, documents the Rasch/IRT scoring model, the adaptive 'soft item targeting' algorithm (items pitched at a 70% chance of success), automated DIF bias detection, reliability, validity, group differences and the composition of its norm groups.
https://www.testpartnership.com/fact-sheets/MindmetriQ-Technical-Manual.pdf
- DocumentationCaptured Sep 25, 2026
The 94-page Insights Series Technical Manual (first published March 2017) documents the theoretical framework, Rasch calibration, reliability, validity, group differences and norm groups of the computer-adaptive verbal, numerical and inductive tests.
https://www.testpartnership.com/samplereports/psychometric-properties.pdf
- DocumentationCaptured Sep 25, 2026
A 2026 technical report on the TPAQ-45 and TPAQ-27 personality questionnaires sets out construct validity against the IPIP-NEO-120, HEXACO-PI, CEI and CSE scales, Rasch reliability, group differences and 15 norm groups (the general-population norm has n = 10,935).
https://www.testpartnership.com/samplereports/tpaq-psychometric-properties.pdf
- Public statementCaptured Sep 25, 2026
Test Partnership describes its platform as validated psychometric assessments with 'no AI scoring black box', so it does not present its scoring as AI or machine learning.
https://www.testpartnership.com/alternatives/hirevue-alternatives.html
- AbsenceCaptured Sep 25, 2026
No AI policy, model or system card, EU AI Act technical-documentation mapping, ISO/IEC 42001 certification or AI-specific instructions for use was found in a keyword scan of all 173 sitemap pages. MindmetriQ's published criterion validity is against GCSE grades (n = 125), and none of the three reports includes a job-performance study.
Bias Audit Transparency
4 items48
- DocumentationCaptured Sep 25, 2026
The self-authored 2026 TPAQ technical report publishes Cohen's d for gender, age (under vs over 40) and ethnicity (white vs BME) on every personality scale and reports no effect of 0.5 or more.
https://www.testpartnership.com/samplereports/tpaq-psychometric-properties.pdf
- DocumentationCaptured Sep 25, 2026
Section 9 of the 2019 MindmetriQ manual (restated in a 2026 fact sheet) reports Cohen's d by gender, ethnicity, age, sexual orientation and disability for each game, including a moderate age effect (d = 0.52) on Number Racer, and tells test users to monitor pass rates against the four-fifths rule.
https://www.testpartnership.com/fact-sheets/MindmetriQ-Technical-Manual.pdf
- DocumentationCaptured Sep 25, 2026
The 2017 Insights manual reports group differences by gender, age (under vs over 50), ethnicity and first language, including d = 0.65 for first language and d = 0.49 for ethnicity on Insights Verbal, and recommends that clients run their own local adverse-impact analyses.
https://www.testpartnership.com/samplereports/psychometric-properties.pdf
- AbsenceCaptured Sep 25, 2026
No independent bias audit (by BABL AI, DCI, ORCAA, Warden AI, Holistic AI or anyone else), NYC Local Law 144 summary or cut-off impact ratios were found: no 'bias audit', 'Local Law' or 'impact ratio' reference appears across the 173 sitemap pages, and guessed Warden AI trust URLs for the vendor return 404.
FRIA Support
3 items36
- DocumentationCaptured Sep 25, 2026
The Data Protection page gives clients a controller checklist: avoid automated decisions (and fully explain and justify any they do make), give candidates notices, prepare for subject access requests, minimise data by using candidate IDs, and set retention periods.
- DocumentationCaptured Sep 25, 2026
The privacy notice says Test Partnership carries out DPIAs before introducing new profiling activities and regularly tests its algorithms and processes for accuracy.
- AbsenceCaptured Sep 25, 2026
No Fundamental Rights Impact Assessment template, EU AI Act deployer-obligation guidance or AI Act reference was found on the data-protection page, the resources hub, the terms, the technical manuals or any of the 173 sitemap pages.
Data Governance Disclosure
5 items57
- DocumentationCaptured Sep 25, 2026
The Data Protection page lists its data-handling controls: encryption in transit and at rest, MFA and SSO, role-based access with per-user activity logging, and external penetration testing. Data is stored in the UK and USA on 'UK-based servers with ISO 27001 accreditation', and results are kept for 24 months by default, which clients can change.
- DocumentationCaptured Sep 25, 2026
The privacy notice (updated 14 January 2026) lists sub-processors with their purpose and data residency (including Cronos Internet for UK main hosting, Kamatera and Kombo in the EU, and Google and HubSpot in the USA), gives ICO registration ZA228041, names an EU representative, and says special category data is not used for profiling.
- DocumentationCaptured Sep 25, 2026
The terms (updated 4 September 2026) include a data-processing schedule, covering approved sub-processors, breach notice 'without undue delay' and customer audit rights. A Schedule C of technical and organisational measures adds Telehouse London primary servers, logging of all access, MFA for admin access, automatic anonymisation per retention settings, and annual staff data-protection training.
- DocumentationCaptured Sep 25, 2026
A Cyber Essentials certificate from CyberSmart/IASME covers Test Partnership Ltd's whole organisation; it was issued on 2 July 2026 and is due for recertification on 2 July 2027.
- AbsenceCaptured Sep 25, 2026
No ISO/IEC 27001 certificate for Test Partnership's own ISMS, no SOC 2 report and no ISO/IEC 42001 certification were found; the site only says its information security policy was developed in line with ISO 27001 principles.
Human Oversight Design
5 items58
- DocumentationCaptured Sep 25, 2026
Test Partnership says it only provides scores and that it is up to the client to make selection decisions, and it recommends clients not make automated decisions but combine scores with structured interviews or work samples.
- DocumentationCaptured Sep 25, 2026
The anti-cheat page says a flag 'is a prompt for a human to look, not a verdict' and that the system never rejects a candidate on a flag alone. Administrators can review the session's response times, window focus and optional screen capture themselves.
- DocumentationCaptured Sep 25, 2026
The help centre's 'Understanding the Results' guide explains how reports express aptitude results as percentiles and personality results as sten scores relative to a chosen norm group, with worked interpretation.
https://testpartnership.zendesk.com/api/v2/help_center/en-us/articles/200842752.json
- DocumentationCaptured Sep 25, 2026
Before testing begins, administrators can give a candidate 25% extra time, or a custom time factor of up to double time, as a reasonable adjustment.
https://testpartnership.zendesk.com/api/v2/help_center/en-us/articles/360011066677.json
- Product featureCaptured Sep 25, 2026
Scores 'flow back into the ATS in real time' through integrations with Greenhouse, Workday, iCIMS, Lever and others. No safeguard against automated stage moves inside the ATS is documented.
Post-Market Monitoring
4 items33
- DocumentationCaptured Sep 25, 2026
The MindmetriQ manual describes automated bias detection that regularly runs DIF analysis on every question and flags biased items for investigation and removal, along with ongoing item-bank improvement. This monitoring is internal, and no results are published.
https://www.testpartnership.com/fact-sheets/MindmetriQ-Technical-Manual.pdf
- DocumentationCaptured Sep 25, 2026
The Diversity and Inclusion page says assessments are checked for adverse impact before publication and 'monitored afterwards', but gives no monitoring results or schedule.
https://www.testpartnership.com/diversity-and-inclusion.html
- DocumentationCaptured Sep 25, 2026
The security Q&A lists a Security Incident Management Policy and a Business Continuity Plan (redacted versions available on request) and external penetration testing. The only contact given is the general info@testpartnership.com inbox.
- AbsenceCaptured Sep 25, 2026
No public status page, incident history, security.txt, dedicated security contact or product/model changelog was found. status.testpartnership.com only returns a Cloudflare 526 error from a wildcard DNS record, and none of the vendor's guessed Statuspage, Instatus or Better Uptime subdomains exists.
Customer Documentation
6 items62
- DocumentationCaptured Sep 25, 2026
The Data Protection and GDPR page combines overviews for clients and candidates, a Q&A in the style of a security questionnaire and a checklist for clients as controllers on using the assessments lawfully.
- DocumentationCaptured Sep 25, 2026
A public Zendesk help centre with 37 articles covers project set-up, interpreting results, norm groups, Job Fit scores, extra time and FAQs on the Workable integration.
https://testpartnership.zendesk.com/api/v2/help_center/en-us/articles.json?per_page=100
- DocumentationCaptured Sep 25, 2026
A public API integration guide (version 2.02, 17 March 2026) documents creating candidates, retrieving scores (percentile, sten and z-scores) and PDF reports, and an optional webhook for results.
- DocumentationCaptured Sep 25, 2026
The resources hub offers the MindmetriQ technical fact sheet, an ability-tests guide and whitepapers, including one on improving fairness and legal defensibility with psychometrics and one on AI resilience in assessments.
- DocumentationCaptured Sep 25, 2026
A 2026 guidance article tells employers to ask providers for pass-rate data by demographic group at any cut-off they are considering, and warns that higher cut-offs tend to increase adverse impact.
https://www.testpartnership.com/blog/impact-cut-off-scores-recruitment.html
- AbsenceCaptured Sep 25, 2026
None of the 173 sitemap pages offers deployer guidance on NYC Local Law 144, the EU AI Act, Illinois HB 3773 or Colorado's automated-decision law.
Editorial notes
Company overview
Test Partnership Ltd (Companies House 08020351; incorporated April 2012 as AD Assessment Ltd and renamed in 2014; registered office in St Albans, England) is a UK psychometric test publisher. Its team traces its origins to a candidate-practice platform it ran in 2010. It sells computer-adaptive aptitude tests (the Insights verbal, numerical and inductive series and others), the TPAQ adaptive personality questionnaire, situational-judgement and soft-skills questionnaires, and MindmetriQ, a six-game cognitive battery marketed as resistant to AI-assisted cheating. Scoring uses classical psychometrics: Rasch/IRT ability estimates reported as percentiles and stens against a norm group the employer chooses. The company markets this as having 'no AI scoring black box'. Employers review, rank and export results in a dashboard, through a public API or inside ATSs such as Greenhouse, Workday and iCIMS. It claims more than 6,000 client organisations and prices by quote, on pay-as-you-go or annual-licence terms.
Regulatory exposure
Employers use Test Partnership scores to sift and rank applicants, but the company states no position on any hiring-AI law. None of the 173 pages in its sitemap mentions the EU AI Act, NYC Local Law 144, Illinois or Colorado. Because it scores with classical statistics (Rasch/IRT, norm-referenced percentiles) rather than machine learning, it can argue that its tests are not 'AI systems' under the EU AI Act. If they are, recruitment use is high-risk under Annex III point 4, with obligations now due on 2 December 2027. It has an EU GDPR representative in Bonn, which indicates it assesses candidates in the EU. NYC's AEDT definition expressly covers statistical modelling that outputs a score used to substantially assist hiring decisions. NYC employers who apply its cut-offs may therefore need an annual independent bias audit, and none exists. DCWP has also promised tougher enforcement after the State Comptroller's December 2025 audit. The company claims CCPA compliance and stores some data in the USA. Its US clients would face Illinois HB 3773 notice duties, California's FEHA automated-decision rules and, from 1 January 2027, Colorado SB 26-189 disclosures. California's rules require four years of record retention, while Test Partnership deletes results after 24 months by default. Under UK GDPR, it advises clients not to make solely automated decisions based on its scores.
Path to a higher score
Test Partnership already publishes more psychometric evidence than most UK test publishers, but all of it is self-authored and much of it is dated. The largest gains would come from three steps. First, an independent bias audit of Insights, TPAQ and MindmetriQ, reporting LL 144-style impact ratios by sex, race/ethnicity and intersectional group at realistic cut-offs, published and repeated annually. Second, group-difference data refreshed from live candidates, with finer ethnic categories. Third, a study of criterion validity against job performance. A public statement on whether its tests count as AI systems or AEDTs would lift its FRIA and documentation scores, especially if backed by deployer guidance covering EU AI Act Articles 26 and 27, NYC notices, Illinois notice and California record retention. On the operational side, the weakest categories would improve with a public status page, a security.txt file with a dedicated security contact, a changelog for item-bank and norm updates, and ISO/IEC 27001 certification of its own ISMS (its policies are currently only described as 'inline with' ISO 27001).
Regulatory frame
What applies to assessment AI.
Assessment is the most-regulated surface in AI hiring. NYC LL 144, EEOC and OFCCP frameworks, Illinois HB 3773, and EU AI Act Annex III §4 all apply, with overlap on bias-audit obligations.
Compare
Test Partnership against its nearest-scoring peers.
The closest scores in Assessment AI, side by side.
- Side-by-sideTest Partnership vs BrightHire51F
- Side-by-sideTest Partnership vs SHL51F
- Side-by-sideTest Partnership vs Xobin50F
- Side-by-sideTest Partnership vs Versant by Pearson49F
Ranking
Others rated in Assessment AI
Ranked 22 of 82 by weighted total under rubric v1.0. The ordering is arithmetic on the rubric and carries no view on which tool suits a given hiring process.
- #19BrightHire51F
- #20SHL51F
- #21Xobin50F
- #22Test PartnershipThis profile50F
- #23Versant by Pearson49F
- #24Jobma49F
Conflicts of interest
No vendor pays for placement, scoring, or removal. Casework - the consulting firm that operates this directory - provides paid services to some vendors. Any active or recent (within 24 months) commercial relationship is disclosed on the affected vendor profile and the review is reassigned to an independent reviewer. See the full policy on About.
Casework has no commercial relationship with this vendor.