Health score bias ?

If you group your health score between three buckets "healthy, neutral, and at-risk", how likely are you to swing in one direction?

For example, leaning more towards putting customers in the at-risk bucket but then this creates a "the boy who cried wolf scenario" where you have too many customers who are "at-risk" but not really. Or staying away from "neutral" and either having customers fall in "healthy" or "at-risk".

What's your take on this?