The top performer who was breaking the team
His numbers were twice anyone else's. So was the rate at which his teammates were quietly interviewing elsewhere.
His commit graph was twice anyone else's on the team. His tickets closed in half the time. His PR throughput was, on paper, the reason the team was hitting its quarter. The CTO described him to me as the engine.
Three of the four people who had to work with him daily had updated their LinkedIn profiles in the previous six months. Two of them were in late-stage interviews elsewhere. I learned this in fifteen-minute one-on-ones the CTO had never thought to run.
The metrics could not see what the team was already seeing. He was a top performer the way a tornado is a fast-moving weather system: technically accurate, and missing the point. His output was real. The cost of it was a debt the team was paying in attrition, in quiet meetings that ran longer than they had to, in junior engineers who stopped asking questions in standup because the answers came with a tax.
The math, when you finally do it, is unforgiving. Replacing two engineers — fully loaded recruiting cost, six months of ramp time each, the morale tax on the people who watch their teammates leave — comfortably exceeds what a single high performer produces in a year. The dashboard never shows this number, because the dashboard is built to count what one person ships, not what two people stop shipping because of them.
Most teams know within sixty days. Most managers act after eighteen months. The gap is not made of evidence. It is made of the manager's reluctance to lose the output they can see and replace it with a problem they would have to solve. Output is legible. Culture is not. The legible thing wins by default until the moment two resignation letters land in the same week.
The cleanest fix is a question almost no performance review asks. Of the people who work with this person, how many would choose to work with them again on their next project? The honest answer is the metric. If the answer is fewer than half, the top performer is a net negative the existing system cannot see.
The CTO eventually let the engineer go. The team's collective velocity dropped for one quarter. By the end of the second, with no other changes, the team was shipping more than it had been with him on it. The two engineers who had been interviewing withdrew their candidacies. The juniors started asking questions again.
The most expensive person on a team is rarely the one whose numbers are bad. It is the one whose numbers are great and whose presence is taxing everyone else's. Read the room. The dashboard does not.