Granularity — how big is a skill?

Cut skills too big and people stall on them for a year; too small and the forest becomes confetti. The method uses a single rule of thumb:

Progressing a skill by one step should take roughly a month, up to about three.

That's the whole rule. It gives you two tests:

• Too big? If moving a skill up one level reliably takes more than ~3 months, split it into smaller skills. A skill that stalls for a quarter isn't teaching anyone anything — it's just sitting there being discouraging.
• Too small? If there's nothing meaningful to grade — you either know it or you don't — don't force levels onto it. Make it a no-level (yes/no) skill and move on.

When in doubt, aim bigger. A broad skill has a natural escape hatch: it gets separated into levels, and each level behaves like its own step. An over-split skill has no such rescue — you just end up with twelve checkboxes where one honest skill should be.

The month-to-three-months window isn't arbitrary. It matches the method's rhythm elsewhere: individual progress is reviewed roughly monthly, so a well-cut skill visibly moves within one or two check-ins. Progress you can see on that cadence is what keeps people engaged; a map where nothing ever changes is a map people stop reading.