What is career adjacency? The full method, the acronyms, and the numbers

Career adjacency is the measurable distance between two occupations, set by the skills they share rather than the titles they carry. Two jobs are adjacent when the skills one of them demands already cover most of what the other demands. It is not a metaphor and it is not a personality quiz. It is a number, it is read from live job postings, and this piece is the whole method: the formula, the three signals behind it, every acronym spelled out, and what the numbers actually say.

Why titles are the wrong unit

A job title is a marketing decision. The same work is a “product designer” at one company and a “UX designer” at the next, a “financial analyst” here and a “finance business partner” there. Titles also hide the opposite case: two roles that sound identical and share almost no daily work. Search by title and you inherit every one of those distortions.

Skills do not have that problem. A posting for a role lists what the role requires, and those requirements are comparable across titles, companies, and countries. So we throw the titles away as a matching key and keep them only as labels. Adjacency is computed on the skill vectors underneath.

Signal one: skills coverage, the readiness score

The core number is coverage. For every occupation we take the skills its postings most often demand and keep the top 20. Then, for a given origin and destination, we ask a single question: what share of the destination’s top 20 skills does the origin already require? That share, times 100 and rounded, is the readiness score.

A worked example. A software engineer moving to backend developer scores 73 percent: of the fifteen skills that define a backend developer in the postings, a software engineer already brings all fifteen, and the two roles overlap enough that the move is a lateral step, not a leap. The Jaccard figure, 0.54, is the tiebreaker: shared skills divided by all distinct skills across both roles, so it rewards pairs that are close in both directions, not just one.

The three-skill floor matters. Two occupations that share only a word like “communication” are not adjacent, they are unrelated, and scoring them would manufacture a route that does not exist. Below three shared skills we return nothing rather than a small, false number.

Signal two: who employers say they welcome

Coverage is forward-looking demand, but it cannot see intent. So we read the postings a second way, for the sentences where a destination explicitly welcomes an origin background: “architecture or industrial-design background a plus,” “former teachers encouraged to apply.” This is employer-attested adjacency, and it catches the non-obvious human moves that a skills model alone would rank too low.

Signal three: where people actually went

Demand is what employers want next. It is not the same as what workers actually did. The third signal is observed mobility: real occupation-to-occupation transitions, from government and resume data. We use three sources, chained by strength.

The rule for this layer is fixed: corroboration, never ranking. Postings measure where demand is going; flow measures where people already went; when the two disagree, that disagreement is the interesting part, not an error to average away. A route the skills model loves but nobody has ever walked is a different animal from one that is both skill-close and well-trodden, and we keep them distinct.

The acronyms, spelled out

The method leans on public datasets, and the field is thick with initials. Here is every one this instrument touches, in plain language.

What the numbers say: adjacency is rare

Run the coverage score across every pair in the taxonomy and the shape is stark. We scored 7,946 origin-to-destination routes in the July run. Sorted into readiness bands, as a share of all routes:

Nearly two-thirds of all occupation pairs score under 20 percent. The vast majority of careers are simply far from each other in skill space, which is the honest and slightly deflating truth the motivational literature skips. Only 93 routes out of 7,946 reach 60 percent readiness or higher. Adjacency is not everywhere. It is a thin, specific set of connections, and the entire point of measuring it is to find the few that are real for you.

The routes that clear the bar

At the top of the distribution, the strong non-licensed moves, the ones where the skills genuinely carry and no credential stands in the way:

The rules that keep the number honest

A readiness score is easy to fake and easy to misread, so the method is fenced by a few non-negotiable rules.

Skills over titles, always. The matching key is the skill vector read from postings, never the job title. Titles are labels on the output, nothing more.

Read demand, do not trust self-report.The skills come from what postings require, not from a candidate’s self-assessment or an employer’s tag. The documented weakness of tag-based boards is that the tags are wishful; ours are read from the description text.

Readiness is not eligibility. Coverage measures skills, and skills are not the only barrier. Medical assistant to dental hygienist scores 90 percent on skills and is still gated by a license that takes years. So the license is flagged separately, on its own axis, and never folded into the readiness number. A high score with a legal gate is a real finding, not a mistake, and we show both.

No manufactured adjacency. The three-shared-skill floor means a pair either clears the bar or returns nothing. We would rather show fewer routes than invent a connection out of one generic overlap.

Corroborate, keep disagreement. Demand, employer intent, and observed flow are three separate readings. Where they agree, confidence is high. Where they disagree, we surface it rather than blend it into a single reassuring average.

Only licensable data. Every external dataset in the stack, the CPS mobility network and the resume trajectories among them, is public domain or Creative Commons, which is what makes it legal to build a paid product on. Nothing is scraped against its terms.

What it is for

Career adjacency is the measurement under a single practical question: given the skills you already have, which roles are actually within reach, and how far. The instrument takes an origin occupation, reads your skill vector, and returns the reachable destinations ranked by readiness, each with the salary band, the specific skill gap, and the honest odds attached. A measured route is one row of that answer, opened up. The glossary defines every term and links every source.

The number is not a promise. It is a map of the skill distance between where you are and where you are thinking of going, built from what the market is actually asking for this week. Start from your skills, read the distance, and the few real routes separate themselves from the many that were never close.