Methodology

How the scan works

A guided live session, a hard quality gate, and results reported as ranges with a confidence level — never a single score. The trust problem is consistency, not raw accuracy, so every stage exists to protect consistency.

The session

Five views, guided one at a time: hairline and front, crown from above at about 45°, part line, temples and sides, and a scalp close-up with the hair parted. The session picks the best frame per view and checks quality as you go — blur, light, framing, and whether the view is the one asked for.

The gate

Never score a bad photo. A frame that fails gets a retake with a specific, actionable reason — for example, too dark, face a window. After repeated failures on a shot, you can continue with lower confidence on that dimension, or finish later. The lower confidence is shown, not hidden.

The result

Five dimensions: hairline, thinning and bald-spot areas, density, dryness and damage, and scalp condition. Each one comes back as a graded range with a confidence level — high, medium, or low — and a one-line note in plain language. Confidence comes from image quality, agreement across repeated readings, and, for returning users, plausibility against your history. A result that jumps implausibly from your baseline prompts a retake to confirm, rather than an instant trend claim.

The baseline

Your first scan is your baseline. Rescans run on an 8 to 12 week loop, and the product’s real signal is change over time.

What it is not

The scan does not diagnose. It describes signs consistent with patterns, in assessment language, and routes to a human specialist when the signs call for one. If the scan shows something needing a specialist promptly — sudden patchy loss, scalp pain with shedding, signs of scarring, or rapid diffuse shedding with other symptoms — the flow routes to dermatology care first and the standard plan is set aside for that pathway.

Skeleton note: the scan on this staging site runs in sandbox mode on sample frames. No live analysis is running yet, and every result shown is labeled SAMPLE. The live scoring service is a separate workstream; the front end is built against the ScanAnalysisProvider seam with fixtures so the two never block each other.