Advancing the science of talent decisions.
Original research, workforce intelligence, and evidence-backed insights drawn from over 5 million assessments, 20,000+ researched skills, and 30,000+ job descriptions.
Hiring software can be built in days. Hiring intelligence takes years.
That intelligence is what teams put to work when they make talent decisions on Xobin.
The Xobin Capability Graph.
Xobin is built on the Xobin Capability Graph (XCG): a proprietary, validated, continuously evolving intelligence system that maps how work is performed and measured. It models three independent entity types (Functions, Skills, and Outcomes) and the validated relationships between them.
Functions define what an area of work is accountable for. Skills are the capabilities required to deliver against that accountability. Outcomes are the measurable results those skills produce. Because the entities are linked rather than nested, the graph can be queried in either direction: which skills a function depends on, or which outcomes a given skill actually predicts.
Skills and outcomes are defined independently of what performs the work, so the same unit of capability can be assessed whether it is delivered by a person, an AI agent, or a combination of the two.
Functions
what an area of work is accountable for
Skills
capabilities required to deliver
Outcomes
measurable results produced
Functions
what an area of work is accountable for
Skills
capabilities required to deliver
Outcomes
measurable results produced
Independent entities, not a hierarchy. The graph can be queried either way, from a Function to the Skills it depends on, or from a Skill to the Outcomes it actually predicts.
- 40+
- industries
- 2,900+
- mapped work profiles
- 20,000+
- skills
- 25,000+
- work activities resolving to measurable outcomes
200,000–300,000
function–skill relationships
Across 40+ industries the graph resolves to roughly 200–300 functions once overlap is accounted for, with each function carrying an estimated 600–1,200 distinct skills. That implies 200,000–300,000 function–skill relationships, meaning each skill attaches to ten to fifteen functions on average, since skills are many-to-many with functions rather than divided among them.
The distribution is heavily skewed: a small set of universal skills appears in most functions, while a long tail of specialist skills appears in only one or two.
- 01Public job descriptions
- 02Standardized assessment data
- 03Open occupational taxonomies (e.g., ESCO)
- 04Learning and certification platforms
- 05Software and talent-data providers
- 06Subject-matter experts
- 07Additional authoritative open sources
Subject-matter experts have reviewed the graph across 2,000+ work profiles and 2,000+ skills, assessing both entity definitions and the relationships between them.
- Precision
- 96.2%
- Recall
- 88.5%
- Relationship (edge) validation accuracy
- 92.3%
Measured outcomes feed back into the graph, continuously re-weighting which skills predict which results. The output is a living map of the labour market, one that evolves as functions change, new skills emerge, and the relationship between capability and measurable outcomes shifts.
The Human vs AI Skills Report.
Xobin Research tested 683 skill groups against AI setups in the first half of 2026: 72% met the criteria for full or partial delegation under at least one tested setup. The 2026 Mid-Year Edition reports this and four further findings on leadership scorecards, technical hiring asks, coding-assessment requests and emerging skill groups.
READ THE REPORT →Four pillars behind every Xobin score.
Validation
Reliability, construct validity, and criterion validity, measured, reported, and re-tested for every Xobin assessment.
Read more →02 ·Proctoring
Secure, auditable testing conditions so a Xobin score reflects the candidate, not the environment.
Read more →03 ·Bias
Adverse-impact ratios and differential item functioning monitored across every scored item.
Read more →04 ·AI Bias
AI interview scoring benchmarked against blind human raters and audited for demographic drift.
Read more →The team that stands behind every Xobin score.
Xobin Research is the in-house team of psychometricians, I-O psychologists, and machine-learning researchers who design, validate, and audit every assessment and AI interview Xobin ships.
We treat a Xobin score the way a lab treats a measurement: with a stated method, a known error, and evidence anyone can request.
About the lab →Recent from the lab.
- Sep 9, 2026Flagship report
Human vs AI Skills Report
Five findings from Xobin's first-half 2026 records: 72% of 683 tested skill groups met the criteria for full or partial AI delegation, EQ-related criteria averaged 52% of leadership scorecard weight across 113 templates, business-to-technical translation ranked among the top five requested skills across 35 technical roles, traditional coding tests lost request share as AI-assisted coding tasks gained it between the first halves of 2024 and 2026, and collaboration and non-linear thinking appeared in more of 24 tracked emerging skill groups.
Read → - Dec 15, 2025Technical documentation
Xobin Psychometric Test: Reliability and Validity Evidence
Technical documentation for the Xobin situational judgment test: the behavioural model, how it was built, and its reliability and fairness evidence.
Read →
Frequently asked questions
What is the Xobin Capability Graph?
- The Xobin Capability Graph (XCG) is the validated model behind every Xobin score. It maps three entity types, Functions, Skills, and Outcomes, and the researched relationships between them, covering more than 20,000 skills and 200 to 300 functions across 40+ industries.
How does Xobin Research validate an assessment?
- Every assessment is characterised on reliability, construct validity, and criterion validity before it ships, then re-checked on a schedule. Assessments that fall out of tolerance are paused rather than left in production, and the evidence dossier is available to enterprise customers on request.
Can I get access to Xobin research data or reports?
- Yes. Press enquiries, early access to reports, methodology questions, and data-access requests against the Capability Graph all go through the contact page, and we usually reply within two business days.
Get in touch with the research team.
The methodology described here is what runs inside every Xobin assessment and AI interview, so hiring teams can make talent decisions on evidence, not instinct. See Xobin.