How to read confidence and limitations

What does a confidence value mean?

It describes how strongly your photographs support one specific observation — not how confident anybody is in you, and not how good or bad the observation is. A finding such as "hair density at the crown is lower than at the front" might be high confidence when several views show the crown clearly in even light, and low confidence when the crown appears in only one view, half in shadow. Confidence belongs to a finding, so two findings from the same set of photographs can carry different values.

What lowers confidence?

Anything that hides or distorts what is being observed. Head tilt and turned angles change apparent proportions; a camera held too close distorts the centre of the face; hats, glasses on the head, hands and hair can cover the area being assessed; filters and beauty modes smooth skin and change colour; and uneven or coloured light changes how hair and skin read. A good analysis tells you which of these applied, so that you can retake the photograph rather than accept a guess.

Why are some findings withheld?

Because a polished explanation of a weak observation is worse than no explanation. When the photographs do not support a finding well enough, LEVEL withholds it and says so, rather than showing it with a caveat attached that most people skip. Less output that holds up is more useful than more output that might not, and a withheld finding can usually be recovered by retaking one view in better light.

What travels with every result?

The limitations. A released finding should say which views it used, what could not be determined, and where an assumption remains uncertain — for example, that colour judgements depend on the light being consistent across the photographs. Read these before acting on the finding; they tell you how much weight it can bear.

What can a photograph never tell you?

Anything medical, anything about health, and anything about worth. A photograph can show the outline of a haircut, where a beard grows, how frames sit, and how colour reads near the face under the light it was taken in. It cannot diagnose skin or hair conditions, and it cannot rank a person. A product that claims otherwise is claiming more than the evidence allows.

What is the difference between low confidence and withheld?

Low confidence is shown with a warning; withheld is not shown at all. A finding the photographs support only partly can still be useful if you know how much to trust it, so it appears with its confidence and the reason it is not higher, such as "crown visible in one view only". A finding the photographs do not support well enough to be worth acting on is held back entirely, and the result says which area was held back and why, so you can retake that view. The line between the two is set in advance per finding, not chosen after the fact to make a result look fuller. In practice, the commonest reasons for withholding are glare on glasses, hair covering the forehead, a profile that is not fully side-on and uneven light; retaking that one view usually brings the finding back.

How can I get higher-confidence results?

Take the photographs the way the checks ask, because each check exists to remove a known source of doubt.

Even, frontal light

A window in daylight, facing you. Overhead and coloured light are the most common reason colour and texture findings are withheld.

Hair off the forehead and ears

Unless the question is about how you wear it. Covered outlines cannot be measured.

Glasses off for the face views

Then one photograph with them on, so frames can be judged separately.

Normal distance, camera at eye level

Close cameras distort proportions; the progress photo guide explains why.

No filters

Smoothing and beauty modes change exactly what is being assessed.

Should I trust a face analysis that never says it is unsure?

No. Every photograph has limits, so an analysis that is equally certain about everything is either not measuring what it claims or not telling you what it does not know. Look for three signs of an honest result: findings carry their own confidence rather than one score for the whole face, the result names what could not be determined, and some findings are withheld when the evidence is weak. If none of those appear, treat the output as opinion rather than observation. LEVEL shows each finding with its confidence and lists what it withheld and why, so you can see what the result rests on. A useful habit is to read the withheld list first: it tells you exactly which photographs to retake, and in what light, before you act on anything else.