Licence the career and skills graph behind JobCannon

Three things, and the third is the one that is hard to buy anywhere else. (1) Career records — 2,521 of them, each carrying a salary band, a demand signal, a typical education level, day-to-day responsibilities and a written profile. (2) Skill records — 1,863 of them, each with a learning curve, the tooling associated with it, its prerequisites and where it sits in a progression. (3) The graph — 57,490 weighted relationships joining traits to skills, skills to careers and careers to adjacent careers. Anyone can assemble a list of occupations. The weighted mapping between a personality result and a realistic next role is the part we built, and it is what a licence buys.

Includes information from O*NET 29.0 (2024) by the U.S. Department of Labor, Employment and Training Administration, used under CC BY 4.0. JobCannon has modified the source data. The trait, skill and career relationships, the written profiles and the translated layer are JobCannon works. In practice that means three layers with different rights: the occupational taxonomy tracks O*NET 29.0 (2024) and is public domain under CC BY 4.0 — you could obtain it without us; the relationships, weightings and written profiles are JobCannon works and are what the licence covers; the compilation, refresh and support are the service around them. Any surface that shows the data carries the attribution notice through. We will put all of this in the contract in writing, because a licensee who overclaims provenance creates a problem for both of us.

The API is a live read: you call us, we answer with current records, and you never hold a copy. It suits a product that shows career or skill information inside its own interface — 100,000 records a month are included and it is an add-on to any paid plan, from $500 a month. The snapshot is the opposite: a complete structured export, delivered as files with a data dictionary and stable record identifiers, refreshed quarterly for the contract term, running entirely on your infrastructure. That suits offline analysis, research, or volume where per-call reads make no sense — from $25,000 a year. Both carry the same two limits: No resale. No model training.

Mechanically you would need about 4,429 calls to walk every one of the 4,384 records, and a standard contract meters 1,000 calls a day. So it is not impossible — it is just slow, visible to us, and against the terms you signed. More to the point, it would get you the records and not the graph: there is no bulk relationship endpoint, deliberately, because the 57,490 edges are the part that took the work. If you want the whole corpus, the snapshot licence is cheaper than the engineering to scrape it, and it comes with refreshes and a support contact.

No — for education the two come together. The Education Licence is priced per enrolled student per academic year: $3.00 at the entry band, falling to $2.00 at statewide scale, with the full corpus included at no extra charge. Assessments are uncapped inside it — retakes included, nothing metered, nothing to top up in March — so the figure does not move if a year group tests twice. It starts at 1,301 enrolled students; below that one of our standard plans already covers your whole roll for less than a licence would, and UK schools have a flat whole-school licence at any size. Above the floor there are three enrolment bands, so you can multiply your own roll by your band and arrive at the price before you ever speak to us. Per enrolled student that price is above our standard plans, and it is a different licence rather than a larger size of one: the plans meter individual candidates for hiring and coaching, while this covers an enrolled roll for the academic year with nothing metered inside it and the corpus licensed with it. It is invoiced annually against a purchase order.

The corpus is versioned and refreshed quarterly for the term of any licence. Every refresh ships with a changelog saying what was added, what was revised and what was retired, and record identifiers are stable across versions so your own joins do not break when we update. API licensees get the new data automatically on the refresh date; snapshot licensees get the new files. If a refresh would change a field in a way that breaks an integration, we tell you before it ships rather than after.

Yes. The written career and skill profiles exist in the languages JobCannon publishes in, and the translated layer is a JobCannon work covered by the same licence as the English one. Which languages are included is set in the contract — if you need a language we do not currently publish, tell us during the quote rather than after signature, because adding one is editorial work with a real lead time and not a switch we flip.

No resale. No model training. Those two are absolute in every licence we write. Resale means redistributing the corpus, or any substantial part of it, as a dataset — displaying it inside your own product to your own users is exactly what the licence is for, and is fine. Model training means using the records or the graph as training data for a machine-learning model, including fine-tuning and embedding an entire copy for retrieval. Beyond those two, the terms are ordinary: use is licensed to the named entity, attribution carries through to any surface that shows the data, and the licence ends when the contract does.