Methodology

How your career matches are built, and what they can’t tell you

A career tool should be able to show its working. This page lists every instrument, data source and formula behind the numbers you see, including the parts that are not yet proven. If something here looks wrong to you, tell us.

The short version

• What you enjoyis measured with the U.S. Department of Labor’s O*NET Interest Profiler (Holland’s six interest types).
• How you tend to work is measured with the public-domain IPIP Big Five markers.
• What each career is like comes from O*NET, the U.S. occupational database: what people in it typically enjoy and which working styles it rewards.
• How you respond under pressureis measured by this site’s own career simulations. That part is a pilot and is not yet validated, so it carries less weight.

The instruments

Reliability tells you a questionnaire gives consistent answers; it does not by itself prove the results predict how someone will fare in a career. For interests, the wider research literature does support a link between interests fitting an occupation and later performance and satisfaction, though the effect is modest (Nye et al., 2012). That is why matches are a place to start exploring, not an answer.

The career data

Each career is linked to one or more occupations in the O*NET 31.0 database (U.S. Department of Labor), which rates 891occupations on how much they draw on each interest type and on 21 work styles (for example stress tolerance, cautiousness, innovation, empathy, leadership). A career’s profile is the average of its linked occupations. The links are our own expert judgment, not an official crosswalk, and are listed in the project data.

The database describes the United States. India’s work culture, entry routes and pay differ, so treat these profiles as a description of what the work is like day to day, not of the Indian job market.

Approximate matches. Civil Servant (IAS/UPSC track), Armed Forces Officer, Content Creator / Digital Creator, Diplomat / Foreign Service Officer, AI Safety & Governance Specialist, Esports Athlete / Professional Gamer have no exact equivalent in the U.S. database, so a close analogue is used and the match is labelled approximate.

Careers built on request

The careers above were linked to O*NET occupations by us. Careers built on request (when someone asks the site for a career that is not on the list) were not. When one is built, an AI model chooses the one to three U.S. occupations whose day-to-day work is closest, from the full list of 891that O*NET rates on both interests and work styles. The server then checks every choice against that list, so the model cannot make up an occupation, and the career’s profile is the average of the occupations that pass, calculated exactly as for any other career.

Why these matches are labelled “approximate.”Which occupations to average is the model’s judgment, and no expert has reviewed it. A poor pick, or a career that blends unlike kinds of work, moves the profile and the match with it. So these matches are always marked approximate, always count as low confidence (however many questionnaires you completed), and the “Why this score?” explanation names the occupations it was averaged from. We do not nudge the score up or down for this; it is calculated the same way and labelled. If the model sees no reasonable match, or the choice could not be checked, the career shows no profile match at all rather than a forced one.

How a match is calculated

1. Interest fit.We compare the shape of your six interest scores with the shape of the career’s six interest ratings (a profile correlation, from −1 to +1, shown on a 0–100 scale). Only the pattern counts: someone who likes everything a lot is not mistaken for someone with wide interests. If your six scores are almost equal, your interests are not used for ranking at all.
2. Personality demands. Each Big Five factor is linked to the O*NET work styles it should help with. We measure how far the career demands those styles beyond the typical occupation, and compare that with your personality profile in the same way. This is the weakest link in the chain, so it counts for 25% of the profile match, and interests for 75%.
3. Simulation, weighted by the job. After you try a career, your score on each behaviour counts for more if that career demands the matching work styles more than a typical occupation does. A firefighter simulation therefore weighs emotional regulation more heavily than a graphic-design one, and a psychology simulation weighs empathy more.
4. Overall fit for a career you have tried is 60% profile match plus 40% simulation. These weights are design choices, not fitted numbers: the validated questionnaires count for more than the unvalidated simulation. They will be revisited when there is outcome data.

Once a career is linked to O*NET occupations, your profile scores, the career ranking and the fit numbers involve no AI: the same answers always give the same numbers, and the “why” text is generated from those numbers. An AI model is used in two places, both upstream of that arithmetic. In Hard-mode simulations, the two written scenarios are scored against a rubric by an AI model, which is another reason the simulation counts for less than the questionnaires. And for careers built on request, an AI model chooses which O*NET occupations describe the career (see “Careers built on request” above).

What is not proven

• The simulations are not yet validated. We do not know how well their scores predict real-world behaviour or career satisfaction. Their answer keys were written by us, not agreed by a panel of professionals in each field, and in Hard mode an AI model scores the written scenarios, which has not been checked against human raters.
• Links for careers built on request are an AI’s judgment. They are checked against the real O*NET list but not reviewed by an expert, and we have not measured how often the model picks the occupations an expert would. That is why those matches are labelled approximate and carry low confidence.
• Sample and culture. The questionnaires were developed and tested mostly on adults and older students in the United States. We are not aware of large norm samples of Indian school students for either, so we report scores against the scale, not against a peer group, and do not use percentiles.
• The crosswalks are judgment calls. The links between our eight behaviours, the Big Five factors and the O*NET work styles are ours. They are reasonable and documented, but they are not published research findings.
• Work values are not measured. O*NET retired its occupational work-values ratings in release 31.0, and we would rather leave them out than use an outdated or altered version.
• A snapshot. Interests and self-descriptions change, especially between school and early work. Retake them in a year.
• Not a clinical, hiring or selection tool. Do not use these results to diagnose anyone or to decide who to hire, admit or reject.

How we will keep checking it

If you choose to share your answers (an optional, unticked-by-default box when you take a questionnaire), they are used without your name or email to check reliability on our own users and to test whether the simulation scores relate to the Big Five factors they conceptually should. The results of those checks will be added to this page, including any that turn out badly.

Sources

• U.S. Department of Labor, Employment and Training Administration. O*NET Career Exploration Tools, including the O*NET Interest Profiler, and the O*NET 31.0 database. O*NET® is a trademark of USDOL/ETA.
• Rounds, J., Su, R., Lewis, P., & Rivkin, D. (2010). O*NET Interest Profiler Short Form Psychometric Characteristics: Summary. National Center for O*NET Development.
• Goldberg, L. R. (1992). The development of markers for the Big-Five factor structure. Psychological Assessment, 4, 26–42.
• Goldberg, L. R., et al. (2006). The International Personality Item Pool and the future of public-domain personality measures. Journal of Research in Personality, 40, 84–96.
• Holland, J. L. (1997). Making Vocational Choices: A Theory of Vocational Personalities and Work Environments (3rd ed.). Psychological Assessment Resources.
• Nye, C. D., Su, R., Rounds, J., & Drasgow, F. (2012). Vocational interests and performance: A quantitative summary of over 60 years of research. Perspectives on Psychological Science, 7, 384–403.

This tool includes information from the O*NET Career Exploration Tools by the U.S. Department of Labor, Employment and Training Administration (USDOL/ETA). Used under the CC BY-ND 4.0 license. O*NET® is a trademark of USDOL/ETA. International Personality Item Pool (ipip.ori.org). Public domain — used without modification.