How this is measured
Most free personality sites won't tell you how their tests work, because the answer is embarrassing. Here is exactly how Wrenlight's work — including the parts that aren't finished.
How questions are written
Every question on this site follows four rules, and roughly ninety questions that broke them have been rewritten since launch:
A question can only measure one thing. "Do you plan ahead and stick to your plans?" is two questions wearing one trench coat — planning and follow-through are different traits, and bundling them corrupts both readings.
We never ask whether you consider yourself organized — people are unreliable narrators of their own categories, but surprisingly honest about specifics. So we ask what your counter looks like, what happens to the form that needs ten minutes, where the unread messages sit.
Agreeing that "honesty is the best policy" measures your relationship to proverbs. Noticing a pricing mistake in your favor at checkout measures you.
Every dimension is measured by at least three differently-shaped questions — a scenario, a frequency, a recognition item — so a single odd answer can't swing your result. Agreement across shapes is what we trust.
How scoring works
Each answer carries weights toward one or more dimensions. Your dimension score (0–100) is your accumulated weight as a share of the maximum, and your archetype is the profile geometrically closest to yours — with one honesty mechanism attached: when you sit near the midline on a dimension, your result page says so and shows the type you'd read as from the other side. Borderline results are presented as borderline, never rounded into false confidence.
Direction matters: the valence framework
A low score is not a deficit — it depends entirely on what the scale measures. Every dimension in the catalog is classified one of four ways, and all generated language respects it:
Positive (capacities — low is a genuine growth area) · negative (costly patterns like neuroticism or attachment anxiety — low is the healthy end, and we say so) · neutral (styles like humor type — no good end exists) · bipolar (paired poles like E vs I — a "low" score is simply more of the other pole, rendered from the midline outward, never as an empty bar).
This is why Wrenlight will never tell an introvert to practice extraversion, or congratulate anxiety for being high.
Percentiles come from real takers
Where you see "higher than 72% of takers," that is computed from the anonymous score distribution of people who actually completed that test here — no identities, no IPs, just bucketed scores. We display nothing until a test has at least 100 takers, because small samples masquerading as norms are how this industry lies. Counts are shown next to every percentile.
Retakes are kept, not overwritten
When you retake a test, the prior result is archived, and your result page tells you what moved — or tells you plainly that nothing did, which across months is the signature of trait rather than mood. Over time these retakes also let us measure each test's stability against itself, and weak instruments will be revised or retired.
What the integrated portrait is — and isn't
The Portrait weaves your results into one reading: every algorithmic sentence in it is grounded in your actual scores, drawn from a hand-vetted lexicon, and skipped entirely when there's nothing accurate to say. No filler is generated to look thorough. Where the portrait compares how you read yourself with how someone who knows you reads you, both readings are shown raw, side by side.
The honest limitations
These are self-report instruments: they measure how you see and report yourself, which is exactly the thing personality sometimes distorts. They are not clinical tools, diagnose nothing, and are not substitutes for professional support. Population percentages on archetype pages are informed estimates, not census data — they'll be replaced by real distributions as taker counts grow. And Wrenlight's instruments are internally designed: face-valid and carefully built, but not yet externally validated against academic benchmarks. We'd rather tell you that than imply otherwise.
Privacy, briefly
Results live in your browser and a keyed record you control. Norming data is anonymous score buckets. Pageview counting hashes IPs and stores no profiles. Nothing is sold, and there are no ads to optimize you for.