The Norwood scale, and where you sit on it

In 1951 James Hamilton, an anatomist, published the first systematic description of how men lose hair. He had spent years observing that the loss was not random: it followed a small number of shapes, it ran in families, and it depended on androgens — men castrated before puberty did not develop it. He sorted what he saw into types.

Hamilton's scale was a research instrument. It was O'Tar Norwood, a dermatologist and hair-transplant surgeon, who turned it into the one clinics use. In 1975 he published a survey of 1,000 men that redrew Hamilton's boundaries and attached incidence figures to them. He kept the seven-stage spine, and added the Type A variants for men whose hairline retreats straight back without a separate bald spot at the crown.

That is why it is properly the Hamilton–Norwood scale, and why the diagrams on every clinic's website are, with small variations, the ones Norwood drew fifty years ago.

Nothing. It classifies.

The scale sorts the shape of loss into recognisable types. Two things move independently: the hairline, which retreats from the temples first and holds a peak in the middle, and the vertex, which thins as a separate patch. The stages describe where those two can be, and what happens when they meet and merge.

There is also the Type A pattern, where the hairline moves back as a single front with no island and no separate crown spot. It is worth naming because a man with a Type A pattern, reading the standard diagrams, may not find himself on them at all.

Someone looks at your head and picks the drawing it most resembles. That is the entire method. No instrument, no count, no number — a label, arrived at by comparison.

This is not a criticism. It is what the scale was designed for. Hamilton needed a vocabulary for a population study; Norwood needed one for a surgical consultation. Both problems are served well by a small set of recognisable shapes, and neither requires precision.

It is a shared language. “IV” carries more between a patient and a surgeon than “quite a lot at the front and some at the back”, and it carries the same thing in every clinic in the world.

It bounds surgical planning. A stage bounds the area to be covered, which bounds the graft count, which bounds the cost. More importantly it forces the question of future loss: a 24-year-old at III is not planned for as a III, because he will not stay one.

And it is a threshold. Norwood put III at the point where loss becomes cosmetically significant. Below it, most men are looking at a mature hairline rather than a problem.

It is ordinal, not interval. The stages are ranks, not quantities — the distance from II to III is not the distance from V to VI. You cannot average them, subtract them, or get a rate out of them. “He went from III to IV in two years” is not a speed.

Two observers disagree. Assignment is a judgement about which picture a head most resembles, and different people do not always land on the same one. Your stage can change because you changed clinic.

It is far too coarse to follow. Seven bins for a process that moves continuously means a year of real loss usually happens inside a stage, invisible to it. If you are trying to tell whether a treatment is working, an instrument that cannot move for two years is the wrong instrument.

It says nothing about density. Two men at V can have very different amounts of hair. Miniaturisation — terminal hairs growing shorter and finer toward vellus — runs for years before an area looks bald, and the scale is blind to all of it.

It describes one pattern only. Diffuse unpatterned loss does not fit it, and neither does female-pattern loss, for which the Ludwig scale is used instead.

The sample was narrow. Norwood's 1,000 men were surveyed in one country in the early 1970s, and the diagrams were drawn from them.

And self-placement is worse than any of this. A mirror cannot see a crown. People routinely place themselves a stage out in either direction — which is why the tool above labels its answer T4, what you told us, and says plainly that it is the least reliable reading we recognise.

hairtrack exists because of the gap between “which drawing does this resemble” and “how much scalp is visible, where, and is that changing”.

You photograph your scalp — one area in about fifteen seconds, or the whole head in five guided photographs. Computer vision then does what an eye cannot do repeatably: it segments hair from scalp pixel by pixel; it isolates eight anatomical zones, so a number belongs to a place rather than to “the top”; it counts, expressing each zone as visible scalp with a range around it and a confidence level attached; and it repeats, the same zones by the same method next month, which is what turns a label into a line.

None of that replaces the Norwood scale. hairtrack still reports a stage, because that is the word a surgeon will use. What it adds is what a stage cannot carry: which zone moved, by how much, with how much uncertainty, and in which direction.

Every figure comes with how it was produced. We grade our own confidence in tiers, and the lowest of them is what you tell us about yourself. A self-placed Norwood stage is exactly that — a reasonable place to start, and a poor place to plan from.