The boring-but-honest methodology
What we ask about you (before any psychometrics)
Most online career quizzes are pure psychometrics divorced from context. That's why the recommendations land in space — “you might enjoy being a marine biologist” is useless if you're 24 with two kids in Mt Gambier. Before we run any of the validated instruments, we capture:
- Age — reframes the realistic time horizon for retraining, careers with long pipelines, etc.
- Location (state + metro/regional) — filters occupation matches against where employment actually exists in Australia. Adelaide is not Sydney.
- Current situation — in school, working, between jobs, returning after a break, caring full-time, etc. This is what shapes the synthesis the most: a Year 11 student deciding uni vs TAFE needs a fundamentally different report than a 25-year-old transitioning out of hospo.
- Education level — what's accessible now vs what would need additional credentials.
- Past roles — pick from common categories plus rough years. We map this against O*NET's skill profiles to surface the careers your existing skills already unlock. Most people radically underestimate how transferable their work is.
- Working-style preferences — indoor vs outdoor, solo vs team, hands-on vs screens, hours, routine. These predict retention better than psychometrics alone.
- Constraints (optional) — caring responsibilities, mobility, finances, disability, visa. We don't recommend things you literally can't do.
- Salary floor (optional) — the minimum to make a path viable. Without it, we'd recommend things that pay $45k for a senior role.
- Ruled-out (optional) — what you're certain you don't want. Often more diagnostic than what you do want.
- Free text — whatever didn't fit. Goes straight to Claude when we synthesise the report.
None of this is required. Skip anything that doesn't apply. The more honest you are, the more useful the report.
The frameworks
Holland Codes (RIASEC)
John Holland's interest typology, developed in the 1950s and validated continuously ever since. Six interest types: Realistic (hands-on, practical), Investigative (analytical, scientific), Artistic (creative, expressive), Social (helping, teaching), Enterprising (persuading, leading), Conventional (organising, structured). Your top three letters form your “Holland code.”
The 60-item assessment we use is the public-domain O*NET Interest Profiler, developed by the US Department of Labour.
Big Five (OCEAN)
The most scientifically supported personality model in psychology. Five dimensions: Openness, Conscientiousness, Extraversion, Agreeableness, Neuroticism. Predicts work satisfaction, job performance and career changes far better than alternatives like MBTI (which is, to be polite, not a serious instrument).
We use the 50-item public-domain IPIP-NEO short form developed by Lewis Goldberg.
Schwartz Work Values
Shalom Schwartz's value-circumplex theory, adapted for the work context. Distinguishes what you're good at from what you actually care about: autonomy, achievement, security, helping others, status, variety. This is often the missing piece in career advice.
O*NET occupational profiles
Each of 1,000+ occupations in the O*NET database has a Holland code, work-styles profile (mappable to Big Five), work values profile, and skill requirements. Maintained by the US Department of Labour. Free.
Australian job-market overlay
Growth %, median wage, and employment counts come from jobsearch.gov.au (Australia's National Skills Commission). So matches are filtered for actual viability in the AU labour market.
The matching
Your answers produce a profile vector across all five engines. Each occupation has its own profile vector. We compute weighted cosine similarity, filter for AU viability (must have nonzero employment), and return the top 8–10.
The synthesis
Claude (Anthropic's AI assistant, model Sonnet 4.6) is given your profile, your top matches, and your free-text background. It produces a personalised narrative, transition pathway, day-in-the-life snapshot, reality check (pay/hours/downsides), skills gap analysis, and 30/90/365-day action plan.
Claude is good at this. It's not magic. We tell it to err on the side of usefulness and honesty.
What we explicitly don't use
- MBTI — commercially proprietary, not scientifically supported, gives you the same result on different days. Pass.
- Personality colours — not a thing.
- Astrological signs — come on.
- StrengthsFinder/CliftonStrengths — proprietary, costs money, weak validation, similar limitations to MBTI.
Limitations we're upfront about
- The frameworks are robust but not magical. They suggest fit; they don't guarantee happiness.
- O*NET is American. We map AU equivalents where possible but the title-by-title match isn't always 1:1.
- Self-report instruments have measurement error. If you answer how you wish you were rather than how you are, results suffer.
- The synthesis is AI-generated. We've tuned the prompt extensively but it can still be wrong. Treat it as a thoughtful starting point, not gospel.