The Skill Rarity Index: Why Not All Profiles Are Equal in a Tight Talent Market
We mapped 1,769 skills into 54 semantic clusters and built a rarity score on top. Here's what it measures, why it matters for pay transparency, and what your tier actually tells you.
The EU Pay Transparency Directive -- in force since 2023 and applicable from 2026 -- requires employers to justify compensation using objective criteria. Most HR software is not ready for this. One reason is that "skill" has never been formally quantified. A line on a CV and an AI-assessed proficiency score exist in entirely different categories of evidence. We think there is a better starting point: not just which skills a candidate has, but how rare that combination is in the actual labor market.
The Problem with Skill Counts
Most platforms count skills. More skills = better match. But a candidate with ten common skills is not the same as a candidate with five rare ones. A full-stack developer who also knows embedded systems and formal verification does not occupy a position "25% more skilled" than someone who knows three of those five -- they occupy a fundamentally different position in the talent market. Counting does not capture this. Rarity does.
What We Built
We converted every skill in our catalog -- name and description -- into a high-dimensional numerical representation using language model embeddings. This catalog grows every day, shaped in part by you: every time a candidate loads their profile or proposes a new skill, the map becomes more complete. We then grouped these vectors into 54 clusters of semantically related skills. The map today covers 1,769 skills.
From this map, we compute a rarity score for any combination of skills. The score answers a specific question: how rare is this particular combination across the profiles currently on our platform? It is not a measure of how "good" a candidate is. It is a market signal -- the equivalent of a supply reading.
The Four Tiers
- Elite (top 5%) -- fewer than 1 in 20 profiles hold this combination. Finding this candidate in a targeted search is genuinely difficult.
- Rare (top 15%) -- a distinctive profile. Employers running this search will not find many matches.
- Uncommon (top 35%) -- above-average market rarity. A meaningful differentiator in competitive roles.
- Common (>35%) -- widely represented across the market. High demand, healthy competition for your attention.
An important nuance: the score is not a simple product of individual skill rarities. Two skills that belong to the same semantic cluster -- similar domain, similar co-occurrence patterns -- contribute less combined signal than two skills from entirely different clusters. A backend engineer who also masters a niche front-end framework adds more cross-domain depth than one who knows two closely related backend tools. The scoring reflects this.
What It Means for Candidates
If your profile lands in the Rare or Elite tier, you will see a badge on your dashboard. This is not a vanity metric. It is a signal to employers browsing the platform that your combination of skills is genuinely hard to find in a single candidate. That matters in salary conversations today, and it will matter more as pay transparency obligations come into effect.
If your profile is Common or Uncommon, it is not a criticism. It means your skills are widely in demand -- which is also valuable. It means more job posts match you, and more companies are looking for exactly what you have.
What It Means for Employers
When a job post implicitly requires an Elite-tier skill combination, we will surface this -- not to discourage, but to help calibrate. You may be describing a profile that does not exist at the salary range you had in mind. That affects time-to-fill estimates, realistic compensation ranges, and whether it makes more sense to hire two specialists rather than search for a unicorn.
When the employer portal launches, every job post will show a tier badge based on the must-have skills you define. If your post is flagged Elite, you will know before you start searching -- not six months into a fruitless pipeline.
Language Skills Are Scored Separately
Technical skills and language skills follow different market dynamics. A candidate who speaks four languages is not four times as technical as one who speaks one. We score language skills on a separate track -- taking the rarest language component and multiplying it against the technical score. This means a monolingual Elite engineer remains Elite; adding a rare language combination can push a Rare profile toward Elite, but it does not inflate technical scores artificially.
A Note on Research
We believe this kind of skill-space model has value beyond our platform. There are open questions worth studying: do semantic clusters improve match relevance compared to flat skill taxonomies? Does a rarity score predict labor-market outcomes better than raw skill counts? How stable are these clusters as the market evolves?
We are building a pathway for researchers to access the full skill position dataset -- cluster assignments, centroid distances, and rarity metrics. If you are working on labor-market research and want to collaborate, visit our Skills Map page and use the research access request at the bottom.
… And a Note on Statistical Relevance
The scoring model is already in production -- but a rarity signal is only meaningful when it is based on enough data. We require at least 50 profiles with a given skill before that skill's rarity figure becomes reliable. If your tier badge is not showing yet, that is why: the code is there, the math is running, but the pool for your skill combination is still too small to draw a confident conclusion.
The more people join the platform, the more precise the signal becomes -- for everyone. The system is built to handle any profession, any skill set. If you know anyone who is open to new opportunities -- developer, analyst, nurse, carpenter, financial advisor -- sending them here directly improves the accuracy of the map for both of you.
A World First
Giving Power Back
For a long time, employers held most of the information in a hiring process. They knew what the market looked like, what skills were scarce, and what a profile like yours was worth. You had a CV and a gut feeling. We think that asymmetry was bad -- not just for candidates, but for employers too, who spent months chasing profiles that were never a real fit.
The Skill Rarity Index is one step toward rebalancing that. Knowing where you stand in the market is not a negotiating trick -- it is basic information that should have always been yours.
The Map Is Live
The Skills Map and rarity scoring are live on Obvelum today. If you have a profile, your tier is already calculated. When the employer portal launches, job posts will surface the rarity signal automatically. And if you are a researcher -- we would like to hear from you.