Quiz Changelog

A record of meaningful changes to the quiz. Version numbers bump when questions, scoring, or results change in a way that could affect your type. Pure copy or style fixes are not listed.

For what the quiz has actually been producing rather than what has changed about it, the aggregate distribution of every completed run is published at what everyone else got: all sixteen types, the quadra split, and what the numbers can and cannot tell you.

v3.1 — September 2026

New

  • Close results now point to the Type Disambiguator. When one axis comes out 2 to 1 your answers fit two types, and when two do they fit four. Under the email box, the result names those types and gives a button that opens the Disambiguator with them already selected. With three or four close axes, eight or sixteen types are still open, which is too many to compare in pairs, so those results point to the written typing report instead. Results with no close axis are unchanged.
  • You can share your result. When you finish, "Copy link to this result" copies a link to it. Whoever opens the link sees which type it is, a one-line summary, links to the profile and the Dual, and an invitation to take the test themselves. The link carries the type and which axes were close, and nothing else: none of your answers and nothing about you.

Not changed

  • The questions, the scoring and the answer shuffling are exactly as in v3.0, so a v3.1 result means the same as a v3.0 one.

v3.0 — September 2026

Changed — this can affect your result

  • The three Extraversion/Introversion questions have been rewritten. The first six months of results split that axis 83/17 towards introversion while the other three came out near balanced, and all eight introvert types ranked above all eight extravert types. One axis at 5:1 beside three that are not is a property of the questions rather than of the people answering them. Each leaned the same way for a different reason: the extraverted option in the first was a caricature nobody would claim ("work the room") against a double-barrelled introverted one; the second asked how you feel after "a full day of back-to-back meetings", which is draining for most people whichever way they lean; the third asked about a "difficult" problem, which pushes everyone towards going quiet, and hedged only the introverted option. All three now offer two equally attractive answers and ask about the direction your energy goes rather than how much you can take.
  • If you have taken the quiz before, your Extraversion/Introversion answers may come out differently on these. That is the point of the change, and it is why this is a major version.

Changed — how confidence is reported

  • The result page no longer describes a result as "tied or borderline". Each axis is decided by three questions, so a score is either 3–0 or 2–1 and a tie is arithmetically impossible — the old wording offered "came out tied (1–1)" as a state the quiz could never produce, and the score bars carried a "tied" label that could never appear.
  • More importantly, "borderline" fired whenever any one of the four axes came out 2–1, which was true of 87% of results. It was written as the exceptional case and shown as one, so the "close call" prompt went to 87% of finishers and the confident one to 13% — the reverse of what was intended.
  • The result page now counts how many axes came out 2–1 and says what that costs. Each close axis doubles the shortlist, because flipping it gives a different valid type: one close axis leaves two types consistent with your answers, two leaves four, and four leaves all sixteen. The note names the close axes and the shortlist size instead of calling the whole result borderline, and the prompt below it changes at two close axes rather than at one.
  • Alternative types are now recorded against your stored profile only on a genuinely close result. Previously that happened on 87% of runs.

Code

  • quiz-complete and quiz-feedback now carry soft, the number of axes that came out 2–1, in place of the borderline boolean. A new name rather than a redefinition, so the old series ends where it was and nothing compares the two by accident.
  • The version is now read from _data/quizVersion.js, which also records the date this version went live. _data/quizStats.js refuses to publish a result distribution gathered under a different version, and scripts/umami-quiz-counts.mjs drops events from before it.

Note

  • The result distribution added in v2.4 is off the page for now. It was measured on the old Extraversion/Introversion questions, so it describes a quiz that no longer exists. It returns once there are enough results from this version.

v2.4 — September 2026

New

  • The result screen now asks one question: does this type feel right? Answering Yes, Not sure or No changes nothing about your result — it points you at whichever next step your answer implies. "No" goes to the disambiguation tool for your result and its nearest neighbours, "Not sure" to the three small groups that test it, "Yes" to the full profile. The answer is counted in aggregate and is not stored, submitted or attached to you in any way. It is only asked after a run you completed yourself: opening somebody else's shared result link does not ask you to judge it.
  • The result screen now says how often a result comes up — "10% of people who take this quiz get LII — the third most common result here" — with the sample size and the period it covers. The figures are the quiz's own completions, 2,072 of them between March and September 2026. Where two types come out on identical counts the line says they are tied rather than ranking one above the other, which would assert a difference the data does not contain. The line stays off the page entirely whenever the sample behind it is too small to quote honestly.

Code

  • Result counts live in _data/quizCounts.json and are turned into the share, the rank and the wording by _data/quizStats.js at build time. That module fails closed: a partial export, or a sample below its floor, renders nothing rather than a figure the sample cannot support. Covered by scripts/test-quiz-stats.mjs.
  • The feedback answer is reported to analytics as quiz-feedback, carrying the answer, the result type and whether that result came out borderline — so perceived accuracy can be read per type and against result confidence.

v2.3 — May 2026

New

  • Result screen now links through to the expert typing service. On results where one or more dichotomies came out tied or borderline, an additional prompt explains that a 12-question quiz cannot reliably separate close types and points to a written analysis.

Code

  • The quiz-complete analytics event now records whether the result was borderline, so quiz-to-typing conversion can be measured against result confidence.

v2.2 — March 2026

New

  • Result screen now shows two compare page shortcuts: your Identity pairing (two of your type together) and your Dual. Both link directly to the relevant compare page.

Bug fix

  • Temperament anchor links on the result screen were pointing to incorrect fragment IDs and landing at the top of the page. Fixed to match the actual anchors on the temperaments page.

v2.0 — March 2026

Bug fix

  • Score display on the results screen showed a percentage where a raw count was intended. "E 67/3" now correctly reads "E 2/3". Results themselves were unaffected; only the display was wrong.

Scoring

  • Tied dichotomies (1–1) are now labelled tied on the result screen, distinct from borderline (2–1). Both still trigger the "read both sides" advisory.

UX

  • Progress bar now moves immediately on question 1 rather than starting empty.
  • Answer options are randomly shuffled (top/bottom) each session, so repeated takes are not influenced by position memory.
  • Result pages are now deep-linkable: completing the quiz writes ?type=ILE (or your type) to the URL, so results can be bookmarked or shared directly. Visiting /quiz/?type=ILE renders that result without needing to retake.

Code

  • Share-to-X URL now uses the live page URL rather than a hardcoded string.
  • Book CTA visibility now uses the same quiz-hidden class toggle as the rest of the UI, instead of an inline style.display.
  • Added role="progressbar" and aria-valuenow to the progress bar.
  • Added aria-live="polite" region around the question and options so screen readers announce changes.

v1.0 — initial release