Don't promote your best engineer
He was the strongest engineer on the team. You made him a manager. Six months later he's unhappy, the team is slower, and you've got a tier-three manager.
Who's the best engineer on your team? Now run the tape forward. You promote them, and inside a few months the team is slower, the new manager is unhappy, and you've traded a tier-one IC for a tier-three manager. The trade is common enough at companies under 300 people that it should have a name.
The skills don't transfer. The best IC is the person who can solve a hard problem alone in a four-hour stretch. The best manager is the person who can resist solving it alone, and hand it to someone who'll take two days and learn something. These aren't the same person. They're usually opposite people. A promotion ceremony doesn't change which one someone is.
The cost lands in two places at once. The team loses velocity, because your fastest engineer now spends the day in one-on-ones and Slack threads instead of in the code that used to move twice as fast as anyone else's. And the new manager loses the work that made them senior. The deep, solitary problem-solving is now structurally out of reach. They're doing different work for the same hours, and the new work is worse for their particular brain.
The cause is almost always the same, and it's a comp problem wearing an org-chart costume. The company has one ladder, and the ladder is management. There's no way to pay a senior engineer more, hand them more scope, or grow their title without turning them into a manager. So the promotion isn't a promotion. It's a raise with a reporting-line side effect. It works on the books and breaks the team on the floor. This is the same failure mode as handing someone responsibility without the structure to carry it: the title moves, the support doesn't.
What holds people is a senior IC track with real money and real status behind it. A staff engineer who out-earns the engineering manager. A principal who reports to the CTO and stands in the same all-hands slot the VPs get. The org has to prove, in comp and in visibility, that the IC path is a career and not a consolation prize. Most companies under 300 people never build it, and they lose their strongest technical people to the ones that did.
The other half is making the manager promotion deliberate instead of automatic. The engineer has to have asked for it, shown coaching instinct before anyone required it, and had a real alternative to turn down. If there's no alternative, there's no choice, and what looks like a promotion is the org chart quietly absorbing a comp decision.
So next time you're about to move your strongest IC into management, ask one question: are you handing them a career path, or a pay raise wearing a title? If it's the raise, fix the comp structure instead. The team keeps its best builder. The builder keeps the work that makes them good. And the manager you hire from outside will be a manager because they wanted the job, which is the only kind worth putting in front of people.