Ordinal Encoding
Mapping ordered categories to integers that respect their order — small, medium, large becoming 0, 1, 2.
The point is that the integers carry meaning. size >= 1 selects medium and large in one comparison, which a tree finds in a single split and one-hot encoding cannot express without combining columns.
So it's the right choice whenever the order is real — sizes, severities, star ratings, education levels — and one-hot encoding there throws away the signal you had.
It's also the most common way to break a nominal column. Map forty countries to 0 through 39 and a linear model reads country 39 as thirty-nine times country 1, while a distance metric puts alphabetical neighbours close together. LabelEncoder is built for targets and lands on features constantly, because it's one line and the error message goes away.