Skills taxonomy and the skills-based organisation

Q: Where does our existing job architecture fit?

Skills augment, not replace, job architecture in the medium term. The transition typically goes: keep job titles for org clarity, add skills as the underlying language for matching, gradually let skills take over the operational decisions while titles remain for external/internal recognition.

Q: How do we deal with employees inflating their self-attested skills?

Triangulate: self-attestation + peer endorsement + manager verification + skill-demonstration (project work, code review evidence, certifications). Multiple data sources mute the inflation.

Q: Is this just job analysis with new branding?

Partly yes. The conceptual difference is that skills are portable and combinable in ways that jobs aren't. The operational difference is that modern skills platforms make this matching computationally tractable in a way 1970s job analysis couldn't.

60-Second Summary

Deloitte's Skills-Based Organization research: companies operating on skills (not jobs) report 107% higher employee placement effectiveness and 98% more efficient deployment of talent.
A useful skills taxonomy has 200–500 skills, not 2,000+. The most common mistake is over-granularity — distinguishing 'React 16' from 'React 17' creates noise without signal.
Self-attested skills are usually inflated; manager-verified skills are usually stale. Best practice: self-attest, peer-endorse, manager-verify, refresh annually.
Adoption is the harder problem than taxonomy. The companies that succeeded (Schneider Electric, Unilever, IBM) made skills the spine of talent decisions, not a parallel data set.

The job — a fixed bundle of responsibilities attached to a title — is dissolving. In its place: a portfolio of skills the employee brings, matched dynamically to projects and roles the organisation needs filled. The companies furthest along this shift are not the tech giants you'd expect; they're industrials and consumer goods firms with strong workforce-planning cultures.

From jobs to skills

A job description says: 'Senior Marketing Manager — leads ABM strategy for enterprise accounts.' A skills decomposition says: 'requires 7/10 ABM strategy, 8/10 enterprise sales partnership, 6/10 SQL for cohort analysis, 5/10 content development, 7/10 cross-functional project leadership.' The skills version unlocks: internal mobility matching, gigwork on adjacent projects, skill-based pay differentials, targeted L&D investment.

Taxonomy design choices

Choice	Trade-off
Buy (Lightcast, Eightfold, Workday Skills Cloud)	Fast start, pre-populated, vendor-defined granularity
Build	Bespoke to your business, expensive to maintain, requires librarian role
Hybrid (buy + customize)	Modal choice for 1000+ employee companies

Design principles

1
Granularity
200–500 skills total. Each skill should be coachable to in <100 hours and verifiable through observation. If it's both vaguer and more granular than that, it's not a useful skill — it's a topic.
2
Proficiency scale
1–5 with behavioural anchors. 1 = aware, 3 = independent, 5 = teaches others. Five-point scales bias to middle without anchors; four-point forces a side without center; the anchors are what matters.
3
Decay rate
Technical skills (programming languages, tools): refresh every 12 months. Durable skills (negotiation, systems thinking): refresh every 24 months.
4
Skill clusters
Group related skills (e.g., 'data analysis' cluster contains SQL, Python, statistics, visualization). Cluster-level views are usable for executives; skill-level views are usable for managers and individuals.

The data hygiene problem

The garbage-in trap

Skills-based mobility matching is only as good as the underlying skills data. Self-attested skills suffer from Dunning-Kruger inflation (people who don't know what they don't know overestimate by 1–2 levels). Manager-verified skills decay because managers don't update. Without a refresh discipline, your skills database becomes archaeology within 18 months.

Operating habits of skills-based orgs

Skills appear on every hiring requisition, not just job titles
Performance reviews include skills movement, not just rating
Internal mobility marketplace uses skills as the primary match criterion
L&D budget is allocated by skill gap, not by 'training catalog'
Compensation bands have skill-based premiums (e.g., +10% for ML platform skill in eng roles)
Promotion decisions reference skill thresholds, not just performance ratings

Frequently asked questions

Where does our existing job architecture fit?

Skills augment, not replace, job architecture in the medium term. The transition typically goes: keep job titles for org clarity, add skills as the underlying language for matching, gradually let skills take over the operational decisions while titles remain for external/internal recognition.

How do we deal with employees inflating their self-attested skills?