Everything here stays in your browser. Your answers are not stored at all, and what you choose to save is written to this device only.
If you have narrowed your type down to two and cannot get past it, this works through a short set of questions written for that specific pair. It is not a second quiz: the questions come from the function stacks that actually separate those two types, and there are only twelve general ones elsewhere on the site for that job.
It can also come back and tell you the questions did not separate them. That is a real answer, and the tool will give it rather than round a one-question lead up into a verdict.
This tool runs in your browser
It needs JavaScript, because every part of it, the questions, the scoring and anything you save, happens on this device and nothing is sent to a server. With JavaScript off, the comparison pages set any two types side by side as ordinary text.
How this decides
Each pair has its own set of questions, written from the two function stacks in Model A rather than from dichotomies. Dichotomies are what the quiz uses, and they are exactly the wrong instrument here: two types you cannot separate usually share three of their four letters, so the letters have already told you everything they can. What is left is the order of the functions, which is what these questions ask about.
Every question has an option for "neither really fits", and choosing it counts. A forced choice between two descriptions that both feel wrong produces a confident-looking number out of an answer you did not really give, and that number would then be recommending a type to you.
A type is named only when two things are both true: at least two questions separate the pair, and the leading type takes a majority of the whole set. One question is inside the margin of a set this short, and a majority counted against every question asked is what makes the "neither fits" answers count rather than quietly disappear. If either condition fails, the tool says the questions did not separate the two and leaves both candidates in place.
Sets are written one pair at a time and most pairs do not have one yet. When there is no set for the two types you pick, the tool says so instead of generating questions on the spot.
What it cannot do
It compares two types, not sixteen. A clear result tells you which of your two candidates your answers lean towards, and tells you nothing at all about whether the answer is a third type you have not considered. If you are not confident the real answer is one of the two, start from the type profiles or the small groups instead, which is the faster way to narrow sixteen down to a few.
It also cannot see anything you did not tell it. Six behavioural questions are a much narrower instrument than a full typing, where the reasoning runs from a long questionnaire in your own words. Where the questions come out level, that is the honest place to hand the problem to a written analysis, which weighs the two candidates against each other directly and shows its working.