Here is a number from a real personality test: 94th percentile, Assertiveness.
Here is the same number said honestly: somewhere in the top quarter, probably.
Those are the same measurement. The first one is what almost every instrument shows you, and the second is what it actually knows.
Where the width comes from
A facet on this instrument is four items. Four items is not many, and the reliability of a four-item scale sits somewhere around .63 to .88 depending on the facet. Push that through to a percentile and you get roughly plus or minus fifteen points of measurement noise on a single facet.
Fifteen points. Your 60th percentile Trust is a 45th or a 75th about as easily.
That is not a defect anybody introduced. It is what four items buys. The defect is printing "60th" and letting a reader believe the two digits.
Why the band is the finding
On a short scale, the interval is not a hedge attached to the result. It is the result.
If a facet comes back at the 94th with a band from the 78th to the 99th, the honest reading is "high, clearly". If it comes back at the 55th with a band from the 32nd to the 77th, the honest reading is "we do not really know, and neither do you from four questions". Both are useful. Only one of them is what the number alone suggests.
So every percentile on this site carries its interval, and the copy beside it says in words how wide it is — measured precisely, measured roughly, treat the range not the number, or barely distinguishable from average. Words, because a band is a picture and people read the point estimate out of pictures.
The coverage is 80%, not the conventional 95%, and that is a trade I want to be caught making. A 95% band on a four-item facet spans roughly forty percentile points and simply reads as "this test does not work". 80% is what the bar shows; the wider figure stays available in the data. Every band is labelled with its coverage so the trade is disclosed rather than hidden.
The part that costs something
Showing intervals makes this instrument look less impressive than its competitors. Two tests can have identical precision and the one that hides it will feel sharper, more insightful, more worth the fifteen minutes.
That is a real cost and I am paying it deliberately, because the alternative is a product whose confidence is manufactured by omission — and this whole project exists because a confidently presented single number was hiding something.