The AssessAll Blog
Assessment science, decoded weekly.
Measurement, hiring, proctoring, and the future of skills — written for people who make talent decisions.
Recognition of Prior Learning, Explained: An RPL Certificate Is an Assessment Result, Not a Service Record
India has certified 54.28 lakh workers through Recognition of Prior Learning. Prior experience predicts job performance at r = .06, which is why RPL rules drop tenure from the calculation - and why the assessment record behind the certificate is the only thing an employer can read.
Read articleScreen the Applicant or Certify the Workseeker? Two Places to Put the Same Assessment
A South African experiment with 6,891 workseekers found a credibly shareable skill certificate raised employment 17% and weekly earnings 34%; the identical scores, unshareable, moved employment not at all. Employer-side screening vs candidate-side certification, and when each is right.
A Visa Points Score Is Not a Capability Score: What Cross-Border Hiring Still Has to Measure
Canada's 2026 consultation proposes dropping study, sibling and spousal points as weak predictors, and Australian linked data found only about half of points-tested migrants work in their nominated occupation a year after grant. What a migration points score does and does not measure.
Separate Interviewers Agree at .44, a Panel at .74: How to Build a Structured Interview Guide
Across 125 reliability coefficients and 32,428 interviews, panels averaged .74 interrater reliability and separate interviewers .44. A seven-step walkthrough for building an interview guide that measures - job analysis, question format, anchored scales, fixed probes, independent scoring and records.
The EU Moved the Hiring-AI Deadline to 2 December 2027. A Practitioner FAQ on What Already Applies
The Digital Omnibus pushed the AI Act's high-risk obligations for recruitment systems to December 2027 — but prohibited practices, transparency, AI literacy and the GDPR already apply. Eleven questions assessment teams are asking, answered.
77% of India's Web Traffic Is Mobile. Five Myths About Assessing Candidates on Phones.
Across 3.5 million applicants, mobile GMA scores ran substantially lower - and in a five-assessment study, measurement invariance held while mobile SJT means sat half a standard deviation below desktop. Five myths about device effects, tested against the evidence.
The Same Leadership Programme Scored d = .69 and d = 1.22. Only the Evaluation Design Changed.
One Situational Leadership programme measured four legitimate ways produced effects from d = .69 to d = 1.22, and a control scale that should not have moved shifted too. A comparison of the three training-measurement designs and when each is right.
Your Assessment Scores Are Personal Data: What India's DPDP Rules Require Before May 2027
India's DPDP Act recognises only consent and a closed list of legitimate uses - no legitimate-interest ground. What the Rules notified on 13 November 2025 require of candidate assessment data, and why the erasure duty collides with your validity archive.
A 15-Item Adaptive Test Runs on a 95-Item Bank. That Ratio Is the Whole Decision.
Computerized adaptive testing reaches a given precision in fewer questions - but only off a large pre-calibrated bank. What CAT-ASVAB, NCLEX and PISA 2022 reveal about the real cost, and when a fixed form is the better call.
Your Job Grades Are Now Legal Evidence. Most Job Evaluation Schemes Were Never Validated.
EU Directive 2023/970 makes skills, effort, responsibility and working conditions the legal basis for pay differences - but a 2025 review of 199 studies found job evaluation research declining since 2010 with almost no recent work on reliability or validity.
Two Marks Short: Walking Through One Team's Assessment Retake Policy
A retake policy is four numbers and one sentence about which score counts. An illustrative walkthrough of the six decisions, each settled against published retest research.
Dropping the Degree Requirement Changed Fewer Than 1 in 700 Hires: A Practitioner FAQ
Indeed put 19.3% of US postings at bachelor-or-higher in November 2025, up from 16.6% two years earlier, while Burning Glass and HBS found skills-based hiring touched under 1 in 700 hires. Eleven questions on what to assess once the credential filter comes off.
Work Sample Validity Fell From .54 to .33. The Test Didn't Change - the Correction Did.
Sackett et al. (2022) found range-restriction corrections had inflated selection validity estimates by .10-.20 points. What the correction does, where the overcorrection came from, the revised numbers, and six questions to ask before accepting a validity coefficient.
The Mean Is a Choice, Not a Summary: Turning Individual Scores into a Team-Readiness Decision
Mean, minimum and maximum aggregation of members' ability scores each predicted team performance at .21 to .29, so the meta-analysis cannot pick your rule. An illustrative walkthrough of the five decisions behind one defensible team-level capability claim.
Screening for the Bench vs Screening for the Seat: What India's GCC Shift Changes About Hiring
India's GCCs now employ 2.36 million while the top five IT services firms cut net headcount in FY26 - the two models predict different outcomes, so they need different assessment designs.
One Assessment, Five Countries: What Must Be True Before You Compare the Scores
Translating a test does not make its scores comparable. What construct, method and item equivalence actually require, and the six checkpoints a cross-border assessment has to clear before one ranked scale is defensible.
Fewer Than 3% of Applicants Admitted Using AI. Your Problem Is Variance, Not Cheating.
A 2026 study of 5,675 applicants found under 3% self-reported generative AI use, and use alone showed no systematic link to job-fit scores. The measurement risk sits elsewhere: five chatbots answered a teamwork SJT with a fifth of the human spread. Six decisions behind an AI-use policy.
Applicants Score Half a Standard Deviation Higher on Conscientiousness. Here Is How to Audit Faking on Your Own Assessment.
Meta-analysis puts applicant score inflation at d = .52 on conscientiousness and .50 on emotional stability, and in one experiment Likert validity fell from .15 to .01 once the ideal-employee factor was removed. A seven-step audit you can run on your own applicant data.
A 360 Moves Ratings by 0.15 of a Standard Deviation: Six Myths About Multisource Feedback
Meta-analysis puts improvement after 360-degree feedback at d = 0.05-0.15, and a third of feedback interventions make performance worse. Six claims about multisource feedback tested against the evidence: panel sizes, rater self-selection, cross-source disagreement, and development vs decision use.
Probation Is Not a Second Screening. Four Markets Just Shrank the Window to Six Months.
US federal agencies acted on 0.4% of newly hired supervisors during probation. With the EU capping probation at six months, the UK cutting unfair dismissal qualifying service to six months in 2027, and India and the Philippines at six, probation cannot carry a selection decision.
Career Breaks Cost Indian Women 49% of Their Callbacks. A Certificate Does Not Fix It.
An Indian correspondence experiment found women returning from career breaks got 49% fewer callbacks, and upskilling certificates did not help. Five selection-design decisions that move the evidence to where the penalty happens.
Internal Mobility vs External Hiring: You Are Comparing Two Incomparable Kinds of Evidence
External hires were paid ~18% more and performed worse for two years in Bidwell's 5,260-worker study - but tenure carries .07 validity and top salespeople make measurably worse managers. Why the two candidate pools are judged on non-comparable evidence, and a seven-step fix.
What an EQF, NSQF or AQRF Level Actually Tells You About a Candidate
A qualifications framework level describes the complexity of a qualification, not how well a person performs. What referencing to the EQF and AQRF does and does not establish, the four things a level cannot tell you, and how to use one in a hiring funnel.
Extra Time Is Not a Favour: A Practitioner FAQ on Assessment Accommodations
Extra time helps everyone, and candidates with disabilities somewhat more - it is not an advantage. Nine practitioner questions on assessment accommodations: what the evidence shows, why score flagging ended in 2000, and what the ADA, Equality Act 2010, EU law and India's RPwD rules each require.
Your Assessment Centre Measures the Exercise, Not the Competency. Here's What to Report Instead.
Meta-analysis puts exercises at 34% of assessment centre rating variance and competencies at 22%, and a 2026 development-centre study found perfect assessor agreement on scores that matched nothing external. What is actually defensible to report back to a participant.
You Verified the Skill. Did You Verify the Person? Identity Assurance in Remote Hiring
A valid assessment score is conditional on the score belonging to the applicant. What DOJ enforcement records, the ITC administration modes and NIST SP 800-63A-4 say about designing identity assurance into a cross-border remote hiring funnel.
18 of 391 Employers Published the Bias Audit. Here's How to Run One on Your Own Funnel.
A step-by-step guide to adverse impact analysis in hiring: what the four-fifths rule actually says, why passing it is not a clean bill of health, and the records US state rules now expect employers to keep.
A Percentile Is Not a Standard: Norm-Referenced vs Criterion-Referenced Scoring
Norm-referenced scores rank candidates against each other; criterion-referenced scores test them against a defined standard. A comparison of when each is correct, what the Standards and the Uniform Guidelines require, and why percentile cuts quietly expire.
Candidates Quit in the First 20 Minutes. Shortening Your Assessment Won't Help.
Across 69 selection systems and ~250,000 applicants, assessment length had no relationship to withdrawal - 21% quit, most inside the first 20 minutes. Why shortening a validated test widens your error band without fixing drop-off, and what to fix instead.
Accent Is Not Proficiency: Five Myths About Testing Workplace English in India
Two meta-analyses put accent bias in hiring at d = 0.46-0.47 - and in both, comprehensibility did not explain it. What workplace English assessment should actually measure, and why the 1-to-5 communication box is not a measurement.
Map Once, Read Anywhere: How to Make Your Skills Data Portable
ESCO has 13,939 skills, Lightcast 34,000+, O*NET 277 descriptors - and none of them record how well a person performs a skill. A seven-step walkthrough for attaching level, evidence type, date and identifiers so skills data survives leaving your system.
The Rater Is the Instrument: Why Calibrating Assessors Beats Redesigning the Rating Form
Only 21-25% of variance in performance ratings comes from the person being rated. Frame-of-reference training moves that number - trained assessors tracked actual performance in 54.6% of rating variance vs 33.5% untrained. How to run a calibration session that works.
Same Cut Score, Different Questions: A Practitioner FAQ on Test Forms, Item Banks and Score Comparability
Randomising questions is not the same as having multiple test forms. A practitioner FAQ on equating, item bank size, exposure control, retest effects and the quarterly checks that keep scores comparable.
Credential Recognition vs Skill Testing: Germany Fully Recognises 43% of Foreign Qualifications. Japan Tests Instead.
Germany fully recognised 43.2% of foreign qualifications in 2024 and referred 46.6% to a compensatory test; Japan's Specified Skilled Worker route skips equivalence and examines instead. A factual comparison of the two models and when each is right.
India Is Upgrading 1,000 ITIs. The Practical Exam at the End Is the Weak Link.
One long hands-on practical is an unreliable skill test: generalizability research shows task sampling, not assessor disagreement, drives the error. What that means for employers hiring from India's ITI and apprenticeship pipeline.
Skills Don't Have a Half-Life. They Have a Use Rate.
The five-year half-life of skills is a metaphor, not a finding. What 1,352 effect sizes on skill decay actually show: loss tracks non-use at roughly 0.06-0.08 SD per month, with variance so large that any single organisation-wide interval is guesswork.
Multiple-Hurdle vs Compensatory Scoring: How You Combine Assessment Results Decides Who You Reject
Sequential hurdles or one weighted composite? The evidence on accuracy, adverse impact, cost and legal defensibility — and how to choose between them deliberately.
Testing Is Not Just Measurement: What 272 Effect Sizes Say About Quizzing After Training
Practice testing beats restudying by g = 0.51 across 272 effect sizes, and spaced retrieval beats massed by g = 0.74 after bias correction. Why the end-of-course quiz is mistimed, what retrieval practice does not do, and a seven-step design for L&D.
31.5% of Distractors Are Dead Weight. Your Expert Reviewers Catch Two-Thirds of Them.
Most assessment questions are never checked against response data. Evidence from 2,948 distractors, an 80-year meta-analysis and 2026 AI item-generation research on the four item statistics that find broken questions expert review misses.
The 12-Month Work Sample: How to Turn an Apprenticeship into a Hiring Decision You Can Defend
India has 11.84 lakh NAPS and 5.23 lakh NATS apprentices, and most employers convert them on a single manager rating - the least reliable evidence available (.45 interrater). A seven-step design for a conversion decision you can defend.
Codified Knowledge Got Cheap. Tacit Knowledge Did Not. What That Changes About What You Assess
Stanford's payroll data shows a 19% employment gap for young workers in AI-exposed occupations, with declines concentrated in codified-knowledge roles while tacit-knowledge work grew. What the split does - and does not - mean for how you assess.
The Same Role in Singapore and Malaysia: Six Decisions Behind One Selection Process
An illustrative walkthrough of the six decisions behind building one selection process across Singapore and Malaysia - construct equivalence, national skills frameworks, language load, local cut scores, integrity bands and the records that make it defensible.
The Level 3 Decision: How to Prove Training Actually Changed Behaviour
A walkthrough of the three measurement designs L&D teams choose between when asked to prove training changed on-the-job behaviour — and what the meta-analytic evidence says about each.
Reliability vs Validity: Your Assessment Can Be Perfectly Consistent and Completely Wrong
Reliability tells you whether a score repeats; validity tells you whether it means what you claim. A practitioner comparison of the two checks, with the maths on standard error of measurement, attenuation ceilings and what validity evidence actually requires.
Four Myths About Early-Career Hiring in India, Tested Against 1.53 Lakh Internship Offers
India's PM Internship Scheme published its full hiring funnel: 1.53 lakh offers, ~50,700 acceptances, ~16,000 joiners. The data contradicts four common assumptions about early-career selection — and shows where assessment budgets are actually wasted.
Tamper-Proof Is Not Trustworthy: What a Verifiable Credential Actually Proves
W3C Verifiable Credentials 2.0 and Open Badges 3.0 have solved credential provenance — but a signature says nothing about the rigour of the assessment behind the claim. The three layers of a credential, the seven evidence fields almost nobody publishes, and when a credential is genuinely enough.
Emiratisation, Saudization and the Skill Verification Exam: A Practitioner FAQ on Gulf Hiring
Twelve practitioner questions on hiring under Gulf nationalisation rules in 2026 — what counts as a skilled role, what the Saudi Skill Verification Programme does and does not tell you, and how to assess fairly when eligibility is set by regulation.
Your Competency Framework Is a Vocabulary, Not a Measurement System
Competency frameworks name the right things and measure almost none of them. The evidence on self-ratings (r = .29), administrative manager ratings (interrater reliability .45), and what it takes to retrofit a real evidence layer onto a framework you already own.
70% Is Not a Cut Score: How to Set a Passing Mark You Can Defend
A cut score is a policy judgment, not a property of the test. A step-by-step guide to setting a defensible passing mark: modified Angoff panels, selection-ratio reality checks, measurement error at the boundary, and an adverse-impact check before launch.
India's Employability Rate Hit 56.35%. Two-Thirds of Unemployed Youth Are Still Graduates.
India's national employability index rose to 56.35% in 2026 while graduate unemployment climbed. Both numbers are right — here's what an employability benchmark measures, why it can't screen a candidate, and an 8-step role-anchored alternative.
Volume Screening vs Capability Screening: What the Philippines' Revised BPO Roadmap Changes About Who You Hire
IBPAP cut its 2028 headcount target by up to 650,000 jobs and redefined the workforce as AI-enabled. A practical comparison of volume screening and capability screening, when each is right, and a seven-step migration checklist.
The Agent-to-Agent Hiring Loop: What Survives When Both Sides Are Bots
Candidates' AI agents now apply to jobs that employers' AI agents screen — and every generated signal is losing value. Why detection is a dead end, and how live, proctored, scenario-based verification becomes hiring's trust layer.
No Single Right Answer, Real Predictive Power: The Science of Situational Judgment Tests
Five decades of research and a 2025 review of 524 studies show SJTs predict job performance with smaller group differences than cognitive tests. What they measure, why instruction wording matters, and how AI-scored open-response formats change the economics.
The Accidental Manager: Why Your Best Performer Is Your Riskiest Promotion
82% of managers were never trained for the job — and most were never assessed for it either. Why promoting on past performance fails, and how scenario-based readiness assessment before the promotion fixes management's most expensive selection mistake.
Quality of Hire: The Metric 89% Call Critical and Only 25% Can Measure
89% of recruiting leaders say quality of hire matters more than ever, yet only 25% can confidently measure it. Why the lagging metric fails, and how standardised, integrity-verified pre-hire assessment data turns it into a routine quarterly report.
2.7 Lakh Seasonal Jobs, a 15% Talent Gap: Screening India's Festive Hiring Ramp Without Guessing
India's 2026 festive season will create up to 2.7 lakh temporary jobs amid a 10-15% talent shortfall, and ~25% of seasonal hires convert to permanent roles. How mobile, scenario-based, integrity-banded assessment fixes volume screening at festive speed.
The EU AI Act's Hiring Deadline Just Moved to 2027. Don't Waste the Extra 16 Months.
The Digital Omnibus pushed the EU AI Act's high-risk hiring-AI deadline from August 2026 to December 2027 — but bans, transparency duties and enforcement powers are already live. What HR and assessment teams should do with the deferral.
Can You Trust an AI-Graded Answer? What 2026's Scoring Studies Actually Show
Three 2026 studies settle the AI grading question: hybrid LLM scoring now matches trained human raters on open-response tests and beats them on consistency. What agreement benchmarks, bias checks, and architectures to demand before trusting machine-scored assessments.
Only 8% of Companies Have Reliable Skills Data. Here's How to Build a Real Skills Inventory
88% of organizations admit they can't see their workforce's skills, and half hire externally for skills they already have. Why self-reports and manager judgment fail, and how assessment-based skills inventories fix the visibility gap.
The 20-Interview Hire: Your Funnel's Most Expensive Hour Is Being Wasted
Interviews per hire are up ~40% and applications have tripled to 300+ per role. Why interviewer-hours-per-hire is the metric to cut — and how assessment-first funnels reclaim senior time for the interviews that matter.
Want this thinking applied to your organisation?
Talk to the team behind the platform.
solutions@bodhih.com