Laurence Peter published the observation in 1969 as satire, and it survived because it is accurate: in a hierarchy, people tend to rise to their level of incompetence.

The mechanism is short. A promotion is given for doing the current job well. The next job is a different job. Performance in the current one is evidence about it only to the extent that the two overlap, and often the overlap is small. So people move up while they keep being good at the new thing, and stop when they are not — which means the system tends to settle everyone into a role they are bad at.

The commonest instance

The best engineer becomes an engineering manager. The best salesperson becomes sales director. The best technician opens their own shop.

Each of those is a different job. Writing good code and getting good code written by other people share some knowledge and very little else. Selling and running sellers share a vocabulary. And being excellent at the craft is nearly unrelated to running the business built on it, which is the point competency confusion makes from the other direction.

What makes this expensive is that it costs twice. The company loses its best practitioner and gains a poor manager, and the person themselves usually ends up doing work they enjoy less and are worse at, for more money. Nobody in the chain made an unreasonable decision.

Why companies keep doing it

Three reasons, and they are all rational.

Promotion is the only reward mechanism most companies have. If the only way to pay someone more or recognize them is to change their title, then excellent practitioners must be moved out of practice to be kept.

Performance is the only evidence available. Managing ability is hard to assess in someone who has never managed, and the current job's results are right there and easy to defend.

Refusing to promote reads as a punishment. Telling someone they are excellent and will stay where they are is a conversation almost nobody is prepared to have well.

What to do instead

Separate pay from level. A practitioner track where the best people can earn as much as managers removes the need to promote anyone out of what they are good at. This is the single highest-return fix and it is mostly a compensation design problem rather than a management one.

Promote on evidence about the new job. Give the work before the title — let someone run a project, mentor two people, own a line item — and watch. A trial is cheap; an appointment is not.

Make a return possible. If stepping back down is available and is not treated as a failure, a bad fit corrects in months. If the only way back is leaving, it corrects in years or never.

Ask what the person actually wants. A surprising number of people accept a promotion because declining feels like saying they lack ambition, and would rather keep doing the work. Asking directly is free and rarely done.

The version that catches owners

An owner promotes themselves, continuously and without noticing.

You start doing the work. You become good at it. The reward is a job running people who do the work, then a job running the company that employs them, then a job selling the company. Nobody checks at any step whether the skill transfers, because there is nobody to check, and the evidence keeps being results from the job before.

This is the honest version of why a business stalls at the owner, and it is the argument for the skill sorter: decide which of these jobs you are actually doing, and staff the rest.