Create skills graphs fast with ESCO/O*NET
Andras Rusznyak
11/11/202510 min read
Ha magyarul szeretnéd olvasni a cikket, kattints ide
Why a skills graph is the missing link between headcount and capability.
A clear working model (roles → skills → proficiency → action) that you can implement today without buying a new system.
How to use ESCO and O*NET to bootstrap skill definitions quickly instead of inventing them from scratch.
A 30-day path to stand up a first version for 20 critical roles.
Where analytics value shows up first (hiring, mobility, L&D, and workforce planning).
A practical example you can directly reuse.
IMPORTANT NOTE:
We utilized generative AI in the making of this article.
Most HR data still revolves around job titles and org charts. But titles are noisy. “Analyst” can mean SQL analyst, reporting analyst, marketing analyst, or junior controller. And two “Customer Success Managers” can do completely different work depending on the product and segment.
This creates four real business problems:
Hiring stalls because job ads are vague (“strong communication skills”) instead of anchored in observable capabilities.
Pay decisions get challenged because you can’t defend why one “Manager” earns more than another.
Internal mobility slows because you can’t see who is 70% ready for role X based on what they can already do.
L&D runs blind because you’re pushing generic “leadership training” instead of closing named capability gaps.
A skills graph solves this by creating a shared, structured language for what the organization can actually do, where it's strong, and where it's exposed. This is not “HR transformation slides.” It’s a data model you can build in 30 days and start using in hiring and development in month one.
Why this matters now
What is a skills graph (in practice)
A skills graph is a structured map that connects four things:
Roles – Your defined jobs, at a given level.
Example: “Customer Success Manager L2” or “Maintenance Technician L1.”
Skills – The capabilities required to perform those roles.
Example: “Active listening,” “Root cause isolation,” “Negotiation,” “SQL basics,” “Escalation handling.”
Target proficiency – The level of depth you expect for each skill at each level.
Example: “Escalation handling: level 4/5 required at CSM L3.”
People/Learning linkage – Who has which skills, and which training closes which gaps.
In data terms, the core objects are:
Role – your canonical, structured job entity.
Skill – a reusable capability, ideally with consistent naming.
RoleSkill – the mapping of which skills belong to which role, plus how critical each skill is and what level you expect.
(Later) PersonSkill – which skills each employee has, with evidence.
(Later) LearningAsset – which training module raises which skill.
This is enough to improve hiring profiles, interview structure, and learning targeting — before you do anything fancy with AI.
When you try to define “what skills does this role actually need?” from scratch, you immediately run into three problems:
Managers give you vague language (“strong communicator,” “proactive mindset”).
Hiring teams invent different terms for the same skill.
Every country or business unit uses different wording.
ESCO and O*NET solve that starting problem for you.
ESCO
ESCO (European Skills, Competences, Qualifications and Occupations) is the European Union’s multilingual reference classification of occupations and skills.
It describes which skills are typically associated with which occupations, using a standard vocabulary.
Important for EU-based organizations: it helps align job/skill language across countries, job boards, and compliance expectations.
O*NET
O*NET (Occupational Information Network) is a US-based occupational database that breaks down each job into tasks, skills, knowledge areas, and work activities.
It is widely used for workforce planning, reskilling programmes, and job design.
It tends to go deeper on task/competency descriptions, especially for knowledge work, technical work, and customer-facing roles.
Why they’re useful for you:
You can look up “Customer Success Manager,” “Maintenance Technician,” “Data Analyst,” etc. and pull a first draft list of 15–30 relevant skills in under an hour.
You don’t have to invent terminology. You inherit a neutral, externally validated baseline.
You can align multiple regions or business units on the same language without debating wording for six weeks.
Important: you are not copying ESCO or O*NET blindly. You are using them to get to 80% fast — then having internal subject-matter experts tune the last 20% so it reflects how the job is done in your business.
The 30-day fast path (MVP for ~20 roles)
Goal: create a minimum viable skills graph for your 20 most critical roles in one month.
Who’s involved:
HR Analytics (owner of structure and data)
Talent Acquisition lead (hiring reality)
L&D lead (learning catalogue reality)
3–5 line managers / SMEs (what “good” actually looks like)
HRIS/data contact (to keep naming consistent)
Tools:
HRIS exports
Excel / Google Sheets / SQL for mapping
Power BI / Tableau / Looker for visualization
ESCO and O*NET as source vocabularies
Scope:
20 roles that are either high volume, business-critical, or hard to fill.
Week 1 — Decide the scope and gather inputs
Select your 20 target roles (example: “Customer Success Manager L2,” “Maintenance Tech L1,” “Data Analyst L2,” “Shift Supervisor,” etc.).
Collect all job descriptions, recent requisitions, performance profiles of high performers, and relevant internal training content.
Pull initial skill lists from ESCO/O*NET for each role family.
Expected output: a first-pass longlist of skills per role (15–30 each), plus evidence of what “good” looks like in your context.
Week 2 — Draft the RoleSkill mapping
For each role, tag each skill as:
Core (must have to succeed in the role)
Adjacent (accelerates performance/mobility)
Tool (platform or system skill, e.g. CRM)
Assign a target proficiency per level (L1=2, L2=3, L3=4 on a 1–5 scale).
Remove duplicated wording and merge synonyms. Keep one canonical label, store the rest as alt_labels.
This is where ESCO/O*NET save you days. Without them, you spend week 2 arguing about language instead of rating importance.
Week 3 — Validate with SMEs (fast, controlled)
Run structured 30–45 minute review sessions with the managers/SMEs for each role.
Ask: “Which of these skills do you see in your top performers every day?”
Ask: “What’s missing that really separates strong from average?”
Ask: “Which of these is not actually required anymore?”
Adjust importance and target_proficiency based on what they say.
Mark final status as version v0.9 with owner + date.
Important: do not let this turn into a philosophical debate. This is not a competency framework for posters on the wall. This is hiring, onboarding, and mobility support.
Week 4 — Publish and activate
Create Role Cards for each role:
Top 8–12 core skills (with target proficiency)
4–6 adjacent skills that enable progression or lateral moves
Interview focus areas (which skills are assessed by which interviewer)
Stand up two BI views:
Role → Skills matrix (what matters for each role)
Skill coverage by function (where you’re strong/weak across the org)
Pilot in two areas:
Talent Acquisition uses Role Cards to write job ads and structure interviews.
L&D uses Role Cards to match learning content to actual gaps.
By the end of week 4, this is live in hiring and learning. You are past “theory.”
Practical example: Building the skills map for Customer Success Managers (CSM)
Context
Customer Success Manager (CSM) roles are high-value and difficult to scale. We want:
faster hiring,
more consistent interviews,
targeted onboarding, and
a way to identify internal candidates.
Step 1: Seed the role using ESCO/O*NET (Day 1–3)
Pull skill lists for “Customer Success / Account Management / Client Relationship Management” from ESCO and O*NET.
Collect internal top performer profiles and recent CSM job descriptions.
Build an initial list of ~20–25 skills, including:
“Renewal negotiation”
“Active listening / customer empathy”
“Churn risk analysis”
“Stakeholder mapping”
“Escalation handling”
“QBR storytelling / value narrative”
“CRM hygiene / pipeline discipline”
Tag each: core, adjacent, or tool.
Assign target proficiency by level:
L1 = basic (2/5),
L2 = working independence (3/5),
L3 = expert / leads others (4/5).
Tool skills rarely exceed 3/5.
Step 2: Normalize and align language (Day 4–5)
Merge synonyms (“Customer empathy” and “Active listening” become one normalized skill with alternate label recorded).
Standardize naming so TA, L&D and business use the same terms.
Capture the source of each skill (ESCO / O*NET / SME) and rate confidence (High / Medium / Low).
Step 3: Validate with the business (Week 2)
45-minute review with 2 senior CSMs + the CSM manager.
Ask only three questions:
“What here is absolutely required to not fail in the role?”
“What separates a top performer from a solid performer?”
“What’s missing that we value commercially (renewal impact, expansion)?”
Remove irrelevant items, add missing commercial/relationship-building skills.
Finalize target proficiency per level.
Step 4: Publish the CSM Role Card (end of Week 2)
Deliver a one-page Role Card with:
Top 10 core skills + target proficiency
5 adjacent skills for growth/succession
Interview assignment grid (which interviewer tests which skills)
Onboarding focus: which skills new hires should close in first 30/60/90 days
This Role Card is now the single reference point for TA, onboarding, and L&D.
Step 5: Put it into operation (Week 3–4)
Talent Acquisition:
Rewrite the job ad to match the Role Card (no vague fluff, only named skills).
Assign each interviewer 2–3 skills to assess with behavioral questions.
Start tracking candidate “skill match score” instead of generic “fit.”
L&D / Onboarding:
Build onboarding paths that directly target gaps in “Stakeholder mapping,” “Escalation handling,” and “QBR storytelling.”
Track whether new CSMs reach target proficiency by day 60 or 90.
Internal Mobility:
Look at Support Agents or Account Managers and check whether they already meet ~70% of the CSM’s core + key adjacent skills.
Those people become priority internal candidates.
Step 6: Measure impact in the first 90 days
Time-to-hire: Should drop because interview and offer decisions are faster.
Quality of hire: After 6–9 months, compare performance to initial skill match score.
Onboarding ramp: Track how fast new CSMs reach the required proficiency in renewal negotiation and escalation handling.
Internal fill rate: How many open CSM roles are filled from inside vs outside.
If even two of those curves move in the right direction, you’ve proven value. At that point, you’re not “doing a framework.” You’re running an operating model.
What you’ll get from this article
What are ESCO and O*NET — and why should you care?
The data model you actually need
To run a usable skills graph, you don’t need a complex knowledge graph platform. You need three tables you can maintain in Excel, SQL, or your BI layer:
1. Role table
Fields:
role_id
role_name
family (e.g. Sales, Operations, Product)
level (L1, L2, L3…)
geo (if localized)
status (active, legacy)
version (v1.0, v1.1…)
2. Skill table
Fields:
skill_id
skill_name (e.g. “Negotiation,” “Root cause analysis,” “Active listening”)
alt_labels (synonyms to support search, e.g. “Customer empathy”)
type (core capability / adjacent capability / tool-specific)
3. RoleSkill table
Fields:
role_id
skill_id
importance (core / adjacent / tool)
target_proficiency (1–5 scale)
source (ESCO / O*NET / SME)
version
confidence (High / Medium / Low)
That’s enough to:
Standardize job ads.
Align interview questions.
Link learning to actual capability gaps.
Start serious internal mobility conversations.
You do not need to map every employee to every skill before you get value. Most orgs fail because they try to start at “PersonSkill” level. Don’t. Start with RoleSkill.
Governance without bureaucracy
If you don’t set light governance, the model will rot in 6 months. If you over-engineer it, it will die in 6 days.
Keep it simple:
Definitions: Publish a short glossary: what counts as a “skill,” what “core vs adjacent” means, and how to read the 1–5 proficiency scale.
Data ownership: HR Analytics owns the data model; TA/L&D own usage; business leaders own validation.
Versioning: Every quarter, document changes: new skill added, proficiency expectation raised, tool skill deprecated.
Change control: No one edits RoleSkill ad hoc in their own copy. One owner, one source of truth.
Fairness & privacy:
At MVP, you are mapping roles, not people. That avoids accusations of surveillance.
Only when governance is mature do you start attaching skills to individuals — and even then, with employee visibility and consent.
This keeps you credible with leadership and worker representation.
Where analytics value shows up immediately
You don’t need AI or full maturity to get value. You get impact in four areas on day one:
1. Talent Acquisition
Problem today: Every recruiter and hiring manager writes the job differently. Interviews drift. Decisions become subjective.
With the skills graph:
You push job ads that clearly list 8–12 core skills and required depth instead of generic fluff.
Each interviewer is assigned 2–3 defined skills to assess (with behavioral signals).
You can quantify “match quality” per candidate instead of “gut feel.”
Immediate metrics:
Time-to-shortlist
Interview-to-offer ratio
Offer acceptance for hard roles
2. Learning & Development
Problem today: Training is catalog-driven, not need-driven.
With the skills graph:
You can calculate the gap: “Role expects Negotiation at 4/5, person is at 2/5 → assign targeted module.”
You can tell whether training actually moves that skill.
You can prioritize L&D budget around high-impact capability gaps, not generic leadership courses.
Immediate metrics:
% of critical roles where all core skills meet target
Ramp time to target proficiency for new hires / internal movers
3. Internal Mobility
Problem today: Managers say “we have no pipeline.” Employees say “I can’t move internally.” Both are partially wrong.
With the skills graph:
You can identify employees who already meet ~70% of core + key adjacent skills for another role.
You can define lateral or promo-ready talent pools with evidence.
You make succession less political and more data-backed.
Immediate metrics:
Internal fill rate for target roles
Ramp time for internal movers vs external hires
4. Workforce Planning
Problem today: Workforce planning is headcount math, not capability math.
With the skills graph:
You can see capability concentration and exposure:
“Only 2 people have ‘Escalation handling 4/5’ in Enterprise Support.”
“We’re over-reliant on 1 individual for ‘Churn risk analysis’ in DACH.”
You can link budget to capability risks, not just vacancy count.
Immediate metrics:
Coverage of critical skills (how many people at/above target in each function)
Roles at high operational risk due to skill single points of failure
Pitfalls and how to avoid them
Over-engineering on day 1
Don’t attempt 200 roles. Start with ~20 roles that matter most (volume, cost, or risk).Over-granularity
A 60-skill list per role is unusable. Cap total skills per role to ~15–25 and keep ~8–12 as “core”.Confusing tools with skills
“Knows System X” is not the same as “Can run root cause analysis.” Keep them separate and cap “tool” expectations.Static library syndrome
If you don’t revisit the model quarterly, it becomes irrelevant and political. Bake in review.Jumping too fast to people-level scoring
This creates privacy and trust concerns if you’re not ready. Earn that step with transparency and governance first.
Bottom line for leaders
A skills graph is not a multi-year transformation. It’s a structured, shared way of saying:
“This role requires these skills at this level.”
“This person has these skills at this level.”
“Here’s the shortest path to close the gap.”
You can build version 1 in a month with HR, TA, L&D, and two or three line managers — using ESCO and O*NET to get started fast.
From there, you can hire faster, develop smarter, and redeploy talent with evidence instead of gut feel.
That’s the point: capability becomes measurable, and measurable becomes actionable.
Up next: Thank you for reading this article. We will be posting short snippets on HR Analytics while we are working on season 2 of the larger series. Stay tuned.
Have you read our other articles? Go to Motioo Insights
Do you have any questions or comments? Contact us
