Saville Assessment
HireAIScore rates Saville Assessment 48 out of 100 (grade F, Substantial gaps) under rubric v1.0, last reviewed September 25, 2026. Saville Assessment ranks 30 of 82 vendors rated in Assessment AI, against a category average of 46. That score is drawn from 40 evidence items across 7 rubric criteria for Saville Assessment — 34 with a cited source, 6 recording that nothing was located.
Score breakdown
Saville Assessment, 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
50
+10.0
Bias Audit Transparency
Weight 18% · category avg 41
42
+7.6
FRIA Support
Weight 15% · category avg 31
40
+6.0
Data Governance Disclosure
Weight 15% · category avg 53
55
+8.3
Human Oversight Design
Weight 12% · category avg 55
55
+6.6
Post-Market Monitoring
Weight 12% · category avg 38
38
+4.6
Customer Documentation
Weight 8% · category avg 58
60
+4.8
Total weighted score
47.8(rounded to 48 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
60raw / 100
Contributes 4.8 to the total. 60 against a 58 category average.
Weakest criterion
Post-Market Monitoring
38raw / 100
Contributes 4.6 to the total. 38 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
- 40
- Documentation
- 27
- Audit report
- 2
- Public statement
- 4
- Product feature
- 1
- Absence
- 6
- With a source URL
- 34 of 40
- Source hostnames
- 3
- Fewest items
- 4
- FRIA Support
These figures measure how thoroughly Saville Assessment was reviewed, not how Saville Assessment performed — an absence row, recording that nothing was located, is counted like any other item.
Article 11 Technical Documentation
6 items50
- Audit reportCaptured Sep 25, 2026
A copy of the British Psychological Society's 2014 test review of Wave Professional Styles confirms BPS test registration and describes a 744-page Handbook of norms, reliability and validity data, while noting that users cannot see raw scores and depend wholly on computer-generated reports.
https://irp.cdn-website.com/e25a87b9/files/uploaded/Wave%20BPS%20Review.pdf
- DocumentationCaptured Sep 25, 2026
Saville's 'Handbooks & Technical Summaries' page shows public visitors the message 'Sorry – this area is only available to partners', so the technical manuals cannot be downloaded publicly.
https://www.savilleassessment.com/cra-handbooks-technical-summaries/
- DocumentationCaptured Sep 25, 2026
The Saville Screener page explains the method in plain terms: twelve Wave items are scored against a norm group as 'fixed arithmetic rather than a model's judgment', and AI is used only to read the job description and weight the score.
- DocumentationCaptured Sep 25, 2026
A Saville article reports an internal study comparing AI-generated role profiles for 15 roles with profiles written by its psychologists, finding close similarity, good consistency and 'no evidence of adverse impact being introduced by the AI role profiling process'.
- Public statementCaptured Sep 25, 2026
Saville markets Wave as using 'smart AI (Artificial Intelligence) to test for consistency and mitigate overly-positive or negative profiles' but publishes no documentation of that AI component.
https://www.savilleassessment.com/wave-professional-and-focus-styles/
- AbsenceCaptured Sep 25, 2026
A review of the savilleassessment.com sitemap and site searches for 'AI Act', '42001', 'model card', 'instructions for use' and 'technical manual' found no AI policy, model or system card, EU AI Act technical documentation, instructions for use or ISO 42001 certification.
Bias Audit Transparency
5 items42
- DocumentationCaptured Sep 25, 2026
Saville's January 2025 fairness article reports, without underlying tables, that selection ratios for minority-ethnic and female applicants exceeded 1.0 among 350,000 applicants to a simplified retail SJT, and that two client programmes showed no adverse impact at the 31st- and 82nd-percentile cut-offs.
https://www.savilleassessment.com/striving-for-fairness-in-assessment/
- DocumentationCaptured Sep 25, 2026
An October 2024 Saville R&D article analyses more than 130,000 Swift Analysis Aptitude completions and reports that asking demographic questions after the test, rather than before, narrowed gender and ethnic group score gaps.
https://www.savilleassessment.com/stereotypes-in-assessment/
- Audit reportCaptured Sep 25, 2026
The BPS's independent 2014 test review reports the gender and cultural make-up of Wave's norm samples and notes that the norm tables were not split by group because mean score differences were 'relatively small'; it is a psychometric review, not a bias audit.
https://irp.cdn-website.com/e25a87b9/files/uploaded/Wave%20BPS%20Review.pdf
- DocumentationCaptured Sep 25, 2026
Saville's 'Diversity, Equity & Inclusion' resource page in its client area is restricted to partners, so any adverse-impact data held there is not public.
https://www.savilleassessment.com/cra-diversity-equity-and-inclusion/
- AbsenceCaptured Sep 25, 2026
No NYC Local Law 144 bias-audit summary, independent algorithmic audit or LL 144/AEDT statement was found: site searches for 'Local Law 144', 'AEDT' and 'bias audit' returned only general guidance pages, and the Warden AI trust pages for 'saville' and 'saville-assessment' return 404.
FRIA Support
4 items40
- DocumentationCaptured Sep 25, 2026
Saville's 'HR Compliance & Audit-Ready AI' page tells deployers that hiring is high-risk under the EU AI Act and recommends plain-language notice to candidates, named human sign-off on adverse decisions, pass-rate checks at each stage and audit trails, but it never mentions fundamental rights impact assessments.
https://www.savilleassessment.com/ai-compliance-playbook-hr/
- DocumentationCaptured Sep 25, 2026
The Data Processing Protocol in Saville's standard terms commits it to help clients with breach notifications and data protection impact assessments, the closest thing to impact-assessment support it offers.
- DocumentationCaptured Sep 25, 2026
Saville's public 2024 accreditation workbook teaches test users the four-fifths rule, the adverse-impact risk of high cut-offs, GDPR principles, how to handle accommodations and why an organisation needs a test-use policy.
https://www.savilleassessment.com/wp-content/uploads/2024/07/IAC-Ability-Workbook-2024-1.pdf
- AbsenceCaptured Sep 25, 2026
No EU AI Act Article 27 FRIA template or FRIA guidance was found on savilleassessment.com, and the playbook's '90-Day Compliance Plan' is available only through a download form.
Data Governance Disclosure
6 items55
- DocumentationCaptured Sep 25, 2026
Saville's privacy policy (updated 1 October 2025) says optional Research Data such as gender, age and cultural background 'is not used in the assessment' and serves only fairness monitoring, and that assessee data is kept for seven years before only an anonymised copy is retained.
- Public statementCaptured Sep 25, 2026
Saville's homepage states that it is 'ISO/IEC 27001:2022 & Cyber Essentials Plus certified' but publishes no certificate or scope.
- DocumentationCaptured Sep 25, 2026
Saville's infrastructure providers page names its sub-processors and their locations, including Microsoft Azure for assessment delivery, Bridewell for 24×7 SOC and penetration testing, Constructor for remote proctoring and Kombo for ATS integrations.
- DocumentationCaptured Sep 25, 2026
The Data Processing Protocol in Saville's standard terms (updated 21 July 2026) limits processing to client instructions, requires a written information-security programme, lists race/ethnicity and health/disability as possible special-category data and lets Saville use anonymised data for research after seven years.
- Public statementCaptured Sep 25, 2026
The Screener page states that 'AI touches your job description and nothing else' and that 'nothing learns from who you hired last year', which explicitly limits what the AI uses.
- AbsenceCaptured Sep 25, 2026
No disclosure was found of the AI model or provider behind the Wave Connect role profiler or the Screener's job-description analysis, of any training data, or of ISO 42001 certification.
Human Oversight Design
6 items55
- DocumentationCaptured Sep 25, 2026
The Screener FAQ says AI does not decide rejections: Saville returns 'a fit band and never a decision', the client sets the threshold and its team makes the call, and the fixed scoring means a candidate's score can be explained.
- DocumentationCaptured Sep 25, 2026
Saville presents AI-generated role profiles as a starting point that stakeholders review and adjust, stating that 'AI should assist rather than replace human experts'.
- DocumentationCaptured Sep 25, 2026
Saville's Ethical Guidelines restrict expert reports to qualified, registered users, say psychometric data should be combined with other evidence such as structured interviews, and let participants ask for clarification of their feedback.
- DocumentationCaptured Sep 25, 2026
The privacy policy tells candidates that 'our assessment results are reached by automated means (i.e without meaningful human involvement)' and that clients make the decisions, so requests about automated decision-making go to the client.
- DocumentationCaptured Sep 25, 2026
Saville advises that AI proctoring flags on its aptitude tests be confirmed by a human proctor, with a second human reviewing unclear cases, and that a flagged candidate retake the test under supervision rather than be labelled a cheat.
- Product featureCaptured Sep 25, 2026
Wave Integrate writes results onto the candidate record in the client's ATS and is marketed as 'one auditable source of truth', but it documents no override, logging or reviewer-trail features.
Post-Market Monitoring
6 items38
- DocumentationCaptured Sep 25, 2026
Saville's standard terms commit to 99.5% availability in UK business hours and 99% at other times, with fee rebates for extra outages, and to notify clients 'without undue delay' of security incidents.
- DocumentationCaptured Sep 25, 2026
Saville names Bridewell Consulting as its provider of '24×7 SOC services and penetration testing services'.
- DocumentationCaptured Sep 25, 2026
Saville reports comparing client scores from before and after generative AI became widely available, found no significant score inflation, and says it 'will continue to monitor client data'.
https://www.savilleassessment.com/the-integrity-of-assessments/
- DocumentationCaptured Sep 25, 2026
Saville's 'Product Updates' page is restricted to partners, so change notices are not published openly.
- DocumentationCaptured Sep 25, 2026
A public Knowledge Hub notice announces a refresh of the Oasys admin interface and the date the old interface will be retired, one of the few product updates visible without a login.
https://knowledgehub.savilleassessment.com/general-support-platform-look-and-feel-refresh
- AbsenceCaptured Sep 25, 2026
No public status page, incident history, security.txt or vulnerability-disclosure channel was found after probing status.savilleassessment.com, status.sc-oasys.com, the savilleassessment, saville, sc-oasys and oasys statuspage.io slugs, and /.well-known/security.txt on savilleassessment.com and scuk.sc-oasys.com.
Customer Documentation
7 items60
- DocumentationCaptured Sep 25, 2026
The Wave Connect page describes AI role profiling from job descriptions, results automatically calibrated to the role, and dashboards that combine behavioural and cognitive data.
- DocumentationCaptured Sep 25, 2026
The Saville Screener page explains what the tool measures and how it is scored, gives completion rates and ATS connectors, and answers FAQs about AI's role.
- DocumentationCaptured Sep 25, 2026
Saville publishes its full standard terms, including a service-level agreement, a GDPR Data Processing Protocol and California (CCPA) and China (PIPL) addenda.
- DocumentationCaptured Sep 25, 2026
An AI-compliance page covers the EU AI Act, NYC bias-audit rules and Colorado duties in FAQ form, though its dates predate the Digital AI Omnibus deferral and Colorado's SB 26-189.
https://www.savilleassessment.com/ai-compliance-playbook-hr/
- Public statementCaptured Sep 25, 2026
Saville's 'Confidently wrong' report states that 'Employment AI is high-risk under the EU AI Act' and urges buyers to ask any vendor for validation evidence and for who audits for adverse impact and how often.
https://www.savilleassessment.com/where-ai-brings-risk-to-your-hiring-and-how-to-fix-it/
- DocumentationCaptured Sep 25, 2026
The candidate area offers free timed practice tests, questionnaire guidelines and FAQs, and explains that results go to a trained feedback provider or a qualified project administrator.
https://www.savilleassessment.com/practice-preparation-advice-for-candidates/
- AbsenceCaptured Sep 25, 2026
No customer guidance specific to NYC Local Law 144, Illinois HB 3773 or Colorado SB 26-189 was found, and the technical handbooks, DE&I resources and product updates are restricted to partners.
Editorial notes
Company overview
Saville Assessment Limited is a UK psychometric publisher registered in New Malden, Greater London. Professor Peter Saville founded it in 2004 as Saville Consulting, and it took its current name in 2017. It was WTW's talent-assessment business until the private equity firm Tenzing agreed to buy it in April 2023. At the time, WTW said Saville would remain its preferred assessment software partner. Graham Bell has been CEO since September 2024. The portfolio centres on the Wave personality questionnaires (Professional Styles, Focus Styles, Match 6.5 and Work Strengths), Swift aptitude tests and situational judgment tests. These are delivered through the Oasys platform and, since January 2025, through Wave Connect, which adds AI role profiling and candidate-fit dashboards. The newer Saville Screener gives every applicant a single fit band inside the employer's ATS. Pricing is quote-based and pay-per-use, and some assessments may only be used by trained, accredited users.
Regulatory exposure
Saville's own AI-risk report says employment AI is high-risk under the EU AI Act, and its ranking products fit Annex III point 4. Wave Connect and the Screener use AI to turn a job description into a weighted profile that drives the fit scores or bands used to shortlist applicants, and Wave is marketed as using 'smart AI' for consistency checks. Fixed-arithmetic scoring makes results easier to explain but does not take these products out of scope. Saville therefore faces the high-risk provider duties, including Article 11 documentation and human-oversight design, from 2 December 2027 under the Digital AI Omnibus. It has not published how it classifies its own products. Its self-report and cognitive tests use no biometric data, so the Article 5(1)(f) ban on emotion recognition does not apply. In New York City, an employer that screens applicants with a single fit band or a Match 6.5 role-fit score can be using an AEDT, which needs a yearly bias audit and candidate notice. DCWP has promised tougher enforcement after the December 2025 Comptroller audit, and Saville publishes neither an audit nor LL 144 guidance. Illinois employers have owed notice under HB 3773 since 1 January 2026. Colorado's SB 26-189 disclosure regime starts on 1 January 2027. California's FEHA rules on automated decisions have required four-year record retention since October 2025. In the UK, Saville's privacy policy concedes that results are reached 'without meaningful human involvement'.
Path to a higher score
The biggest gains would come from publishing what Saville already holds. That means public technical summaries for Wave, Swift, Match 6.5 and the Screener, with adverse-impact tables by sex, ethnicity and age, instead of partners-only handbooks. A yearly independent bias audit of the Screener and Wave Connect fit scores in NYC Local Law 144 format would lift its weakest category. An AI policy and system documentation should name the model behind the role profiler and the job-description reader and describe its data and limits. Adding EU AI Act instructions for use and a deployer pack with a FRIA template would cover Articles 11, 13 and 27. Saville should also update its compliance playbook for the Omnibus timetable and Colorado SB 26-189, and publish its ISO 27001 certificate and scope. Finally, it should add a public status page, a security.txt file and a changelog for scoring and AI-model updates.
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
Saville Assessment against its nearest-scoring peers.
The closest scores in Assessment AI, side by side.
- Side-by-sideSaville Assessment vs X0PA AI48F
- Side-by-sideSaville Assessment vs Sova Assessment48F
- Side-by-sideSaville Assessment vs VidCruiter48F
- Side-by-sideSaville Assessment vs The Predictive Index48F
Ranking
Others rated in Assessment AI
Ranked 30 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.
- #27Sova Assessment48F
- #28VidCruiter48F
- #29The Predictive Index48F
- #30Saville AssessmentThis profile48F
- #31Thomas International48F
- #32Crosschq47F
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.