Methodology
How our test works
The most complete type test for $29, built to settle close calls. This page says what the questions do and what a simulation showed. It is not a claim that the test is the most accurate, or that a real-world study has already validated it.
Two tests
The free personality test is 25 statements and takes about 5 minutes. It estimates five scales: energy, information, decisions, lifestyle, and identity.
Pro is version 3 of the In-Depth Assessment: 100 statements across 20 trait facets, plus 16 forced choices between cognitive functions. That is 116 questions and about 17 minutes, before any follow-ups. Up to 6 extra questions are asked only when a letter is close.
How the items are written
Each letter scale has four facets. Every facet has items keyed in both directions: agreeing with some items moves you toward that pole, and agreeing with others moves you away. The four facets on a scale are balanced, so a habit of agreeing with everything does not, by itself, decide the letter.
Facet results are leanings. Each one comes from a handful of answers, so the bar shows a direction, not a precise percentage.
The 16 function pairs compare the eight cognitive functions. Each function appears equally often, and equally often as the first and second option.
Items are interleaved so two neighboring questions do not measure the same letter pair. The wording is shown to you. The scoring key stays on the server.
Close calls, confidence, and answer style
If a letter is near the middle, Pro asks follow-up questions on that letter before it commits. The result then shows a confidence label (very clear, clear, or close call) and, when there is one, the runner-up type.
The scorer also looks for patterns that make a result less trustworthy: answers that are almost all the same, a large share of neutral answers, very fast answers, or a letter score that does not fit the function scores. Those flags do not change your type by themselves. They add a short note suggesting a quieter retake.
Simulated recovery
These numbers come from a computer simulation, not from real people. The script is scripts/pro-sim.ts, seed 12345, 200 simulated people per type (3,200 per strength level). Each simulated person has a true type, and their answers are noisy versions of that type. The table is how often version 3 recovered it.
| True leaning | Exact type | Letters right | Asked a follow-up | Called a close call |
|---|---|---|---|---|
| Very clear | 100% | 100% | 0% | 0% |
| Clear | 100% | 100% | 4% | 0% |
| Moderate | 97% | 99% | 42% | 8% |
| Slight | 62% | 88% | 87% | 63% |
| Near 50/50 | 22% | 69% | 95% | 88% |
At a slight leaning, follow-ups were asked of 2,800 letter decisions. Letters correct went from 2,948 before those questions to 3,644 after. Those counts are letters, not people.
Among simulated people with a clear leaning, the highest function matched the type’s usual lead function 77% of the time (100 people per type in that check).
Style checks, still on simulated answers: inconsistent random answers flagged 100%; random close calls flagged 94%; straight agreement flagged 100%; straight neutral flagged 100%. Clear types who agreed with almost everything still landed on the exact type 100% of the time after the balanced keying was applied.
A study with real answers
A validation study is underway. If you opt in on the free test or the Pro assessment, we store item responses without your name, email, or account id, and we may ask a short public-domain Big Five questionnaire and whether the result felt accurate. Aggregate results will be published here when the sample is large enough. Until then, treat the table above as a simulation.