How career matches are computed

Your top 5 career matches are computed from your 29-dimension score vector against the PRISM career taxonomy — 448 degree-requiring career families, each with a 29-dimension importance profile reviewed by an industrial-organizational psychologist.

The math

For each family, we compute mean-centered cosine similarity between your vector and the family's importance profile. Mean-centering means we subtract the average of each dimension across all candidates first, so families don't all rank the same person at the top just because everyone scores higher than zero on most dimensions. The 10 highest-similarity families make your match list.

Top drivers

Each match shows three "top drivers" — the three dimensions where your score and the family's importance profile contributed most to the similarity. The driver list is the right place to look when you want to understand why a particular family ranked well.

The LLM explanation

Underneath each driver list is a short paragraph generated by a large language model the first time you open your report. It reads your three drivers and writes one or two sentences about how those drivers connect to the day-to-day of the career. The text is cached into the result document, so re-opening the report doesn't regenerate it — you get the same wording every time.

If the model returns a low-quality response, we fall back to a templated sentence so you always see something meaningful.