How career matching works

Four numbers, shown separately

Every result shows four distinct measurements. They are never combined into one hidden “best career” score, and none of them is a guarantee.

• Exploration match— cosine similarity between your questionnaire answers (interests, work values, preferred activities, work context, self-reported skill affinity, and learning goals) and each career family's reference profile. This is the only score used to rank and group careers.
• Practical fit — how well a career respects any optional constraints you set (for example a non-negotiable requirement). Constraints can only lower or remove a career from view; they are never a hidden bonus inside the exploration score.
• Learning readiness— how your current self-reported skill affinity and learning goals relate to a career's typical entry paths. It informs recommended lessons and aptitude preparation; it does not exclude a career permanently.
• Market context — wages, outlook, shortages, and current openings for a specific country and occupation, each labelled with geography, period, source, and last-verified date. Market context is informational only and never changes the exploration or practical-fit numbers above it.

Weights for the exploration score (scoring version cm3-weights-v1) are versioned: interests 30%, work values 20%, preferred activities 15%, work context 10%, skill affinity 10%, learning goals 10%, and your optional emphasis 5%. Weights only change after validation, and every export records the exact version used.

What inputs are used

Scoring reads only what you enter in the questionnaire:

• Interests (RIASEC-mapped: Realistic, Investigative, Artistic, Social, Enterprising, Conventional)
• Work values (achievement, independence, income, security, work-life balance, and more)
• Preferred activities (build and repair, analyze and research, teach and help, and more)
• Work context (independent/team, remote capability, travel, shift work, and more)
• Self-reported skill affinity — kept separate from any demonstrated Learn performance
• Learning goals (quick entry, degree, apprenticeship, licensing, and more)
• Optional practical constraints you explicitly add, and an optional weight emphasis

What is never used

The matching engine actively strips a fixed list of protected and sensitive attributes before scoring runs, even if one appears in a questionnaire response by mistake or in a malformed request. None of the following ever reaches the scoring function:

• race
• ethnicity
• religion
• sexualOrientation
• politicalAffiliation
• unionMembership
• healthDiagnosis
• disabilityDiagnosis
• birthDate
• photo
• facialFeatures
• socioeconomicStatus
• gender
• sex
• iq
• nrcScore
• novusReasoningScore

This covers race, ethnicity, religion, sexual orientation, political affiliation, union membership, health or disability diagnosis, exact birth date, photographs or facial analysis, inferred socioeconomic status, sex/gender, and IQ- or reasoning-style scores. Age bands are collected only for age-appropriate content and consent — for example a youth exploration mode with broader families and no salary promises — never as a general reason to suppress or promote a career. Removing any of these fields never changes the ranking or scores you would otherwise see; the engine is tested to produce identical results with or without them present.

Confidence and limitations

Every result includes a confidence percentage built from four honest inputs: how many questionnaire dimensions you actually answered, response consistency, current catalog coverage, and data freshness. Low confidence is shown, not hidden. Missing sections reduce the weight given to that dimension rather than being guessed at or defaulted to a favorable value, and the exact dimensions used or missing are listed in each result's explanation.

Explanations trace back to your inputs: each positive reason and trade-off names the specific dimension it came from (for example “aligned on activities”), so a result can always be inspected, not just trusted.

Jobs never alter your exploration score

“Find jobs” sections show live, user-controlled outbound searches (or clearly labelled example data while no live API is enabled). Novus Learn does not fetch, store, rank, or use job counts to adjust exploration or practical-fit scores. A career with zero current local openings can still be your strongest exploration match; the two facts are shown side by side, never merged. See the country labour-market pages for sourced wage and outlook data.

Career-professional and pilot review

A user comprehension pilot and independent career-professional review of the matching engine and its explanations are planned before this feature is treated as production-validated. That review has not happened yet — see the honest status in the release readiness notes below. Nothing on this site claims that review is complete.

Until an independent review is recorded, treat every result as a deterministic, explainable starting point for exploration — not a validated psychometric instrument.