How these maps are built

Source data

The O*NET 30.3 Database, text distribution, published by the U.S. Department of Labor, Employment and Training Administration. The analysis on this site was generated on 2026-08-23. Files used: Occupation Data, Related Occupations, Job Zones, Work Activities, Knowledge, Transferable Skills, and Essential Skills. Entry requirements come from a second official source: BLS Employment Projections Table 5.4, Education and training assignments by detailed occupation, 2024.

Source occupations

Some professions people search for are a single O*NET occupation. Others are several. “Registered nurse” is one occupation; “teacher” is four. Each map declares its source set rather than pretending a broad word is a single occupational code, and where a set is used, importance ratings are averaged across the set.

How destinations are selected

Candidates are the occupations O*NET lists as related to a source occupation, plus the occupations related to those. Occupations in the source set are removed, and so are the rest of the same occupational family where that family would otherwise dominate: a move from middle school to high school teaching is not an alternative career, and neither is a move from registered nurse to a nursing subspecialty.

Destinations more than two Job Zones away from the source are removed. Emergency medicine physician is not an alternative career for a registered nurse: it is three zones and a separate decade-long credential path, and presenting it as an adjacent career would mislead. Within two zones, the gap is shown rather than filtered.

We considered ranking every one of the 894 occupations with profile data by similarity instead. We do not, because it performs badly: unweighted similarity put travel agent among the closest occupations to journalist, on the strength of generic communication work. Anchoring to O*NET relations keeps the candidate pool defensible.

How overlap is measured

Four domains are compared: work activities, knowledge, transferable skills, and essential skills. An element counts as important when its O*NET Importance rating is at least 3 on the 1 to 5 scale. Shared means important for both the source and the destination. A new demand means important for the destination and not for the source.

The headline percentage is the shared portion divided by the union of everything important to either side, so an occupation cannot score highly merely by being a simpler version of the one you already do. Elements are weighted by how distinctive they are: an element important for a small share of occupations counts for more than one important for almost all of them. Sharing “active listening” with another occupation means very little; sharing “therapy and counseling” means a great deal.

How preparation level is used

O*NET assigns each occupation a Job Zone from 1 to 5, describing the education, training, and experience typically needed. Destinations are labelled by whether the destination Job Zone is the same as, higher than, or lower than the source. The label is arithmetic on a published value, not our judgement about how hard a move would be.

Job Zone is not used as a hidden feasibility filter. An earlier version dropped anything more than one zone away, which removed obvious nonsense but also silently removed legitimate long-range pivots. The friction is now shown rather than hidden.

RoleSprint does still limit candidate destinations to occupations within two O*NET Job Zones of the source, to keep these maps focused on reasonably adjacent exploration. That range is a RoleSprint product-scope decision, not an O*NET rule. It is not a transition probability, and it is not a claim that moves outside the range are impossible. Registered nurse to physician sits outside it because a three-zone gap and a separate decade-long credential path is a different kind of decision from the one these maps are built to support.

Transition friction

Occupational similarity is not the same as transition feasibility. Two occupations can share most of their work and still be separated by a graduate degree. Every destination therefore carries what it takes to cross, alongside what carries over.

Typical entry education, work experience commonly required, and typical on-the-job training come from the U.S. Bureau of Labor Statistics, Employment Projections program, Table 5.4, “Education and training assignments by detailed occupation, 2024”. These are published values reproduced as written. We do not model them, and we do not average them.

BLS publishes against 2024 National Employment Matrix (SOC) codes. O*NET-SOC codes extend SOC with a decimal suffix, so the crosswalk takes the part before the decimal point. Where the two classifications genuinely differ, an explicit alias is recorded in the generator: O*NET “Mental Health Counselors” (21-1014), for example, sits inside the BLS row “Substance abuse, behavioral disorder, and mental health counselors” (21-1018). Where no BLS row exists, the fields read as not published rather than being guessed.

Where the destination’s typical entry education is higher than your field’s, the destination carries an explicit note saying so. That note reports the BLS value and nothing more.

We make no licensing claims. BLS Table 5.4 publishes education, experience, and training; it has no licensing field, so inferring licensure from an education category would be us guessing. Many occupations on these maps are in fact licensed or certified, and the requirements vary by state. Check the authoritative source for your state and occupation before planning a move.

Direct and extended relationships

Each destination states whether O*NET lists it as related to your occupation directly, or whether it is an extended adjacent path: related to an occupation that O*NET lists as related to yours. You should never have to assume a two-step destination was classified by O*NET as related to your work, because it was not.

Career change and related field

Destinations are split by whether they sit in a different SOC major group from your occupation. Registered nurse and nurse practitioner share a major group, so the latter is labelled a related field rather than an alternative career, and a page answering “alternative careers for nurses” lists career changes first.

We deliberately do not call the same-family group a progression. Sharing a SOC major group establishes that two occupations sit in the same broad family. It establishes nothing about seniority, pay, or whether one leads to the other, and we hold no career-ladder data that would let us claim otherwise.

The split is a fact about the classification, not about working life. Two occupations in different major groups can still share a job site and a toolbox. Where that made a map read as neighbouring trades rather than alternative careers, we removed the map rather than relabel it.

What is source data and what is ours

Neither USDOL/ETA nor BLS has approved, endorsed, or tested anything in the right column. Where these pages recommend, characterise, or rank, that is RoleSprint speaking, not a federal agency.

What this analysis does not mean

• It is not a probability. No figure on these maps estimates the chance of being hired, or of a transition succeeding.
• It is not a fit score for a person. Occupational data describes what work requires, not what any individual has done.
• Occupational similarity is not transition feasibility. A high overlap next to a doctoral entry requirement means the work is familiar and the door is still gated.
• It is not a record of what people actually do. O*NET publishes occupational requirements, not labour-market transition flows. We do not have flow data and do not claim any.
• It does not account for licensing, geography, pay, employer preference, or whether you would enjoy the work.
• Importance ratings come from surveys of incumbents and occupational analysts, and carry the sampling limits described in the O*NET documentation.

Reproducing this

The generator is committed at scripts/generate-career-transition-data.mjs. It reads an unzipped copy of the O*NET 30.3 text distribution and writes the JSON these pages render. The raw O*NET release is not redistributed here; download it from the O*NET Resource Center and point the script at it.