People Management: How Leaders Decide When to Coach or Exit a High Performer

We believe disagreement is welcome but public humiliation is not. Strong teams grow when we question ideas with respect and curiosity. Healthy debate improves decisions because different views are openly shared. That belief helped us make a difficult talent decision together with greater confidence and clarity.

A capable colleague often found real problems in proposals during team reviews. We asked them to pair every criticism with a practical solution and share sensitive feedback privately. We noticed teammates stopped offering ideas which confirmed the behavior was hurting collaboration across the group. We ended the relationship and meetings became more thoughtful because everyone felt safe to contribute with honest opinions.

Coaching has an expiration date. The fair process is not infinite patience, it is a defined opportunity to demonstrate a pattern. Anyone can appear collaborative in a meeting. Evidence arrives during a deadline, a disagreement, or a moment when credit is scarce.

I once set a boundary around information flow. A leader could not withhold context, selectively brief allies, or ambush colleagues for leverage. Feedback came from people who worked beside them, rather than dashboards or the leader's account. Improvement had to be visible downstream. When that evidence failed to appear, the separation was not punitive. It protected the company from normalizing political behavior as ambition.

I gave a strong performer two direct, specific conversations about how her comments in team calls were shutting down other people, each one with a concrete example from that exact week, not a vague pattern I was gesturing at. That was the boundary: two named instances, addressed within days of happening, not saved up for a quarterly review. If the behavior repeats after that, the decision isn't really about talent anymore, it's about whether I'm willing to let one person's output justify what it's costing everyone else's willingness to speak up.

In her case it didn't change after the second conversation, and I let her go within two weeks of that talk. Three people on that team told me afterward, unprompted, that meetings felt different, more people were actually offering ideas instead of staying quiet until she'd spoken first. The output hit didn't show up for about a month, then recovered past where it was because the team was contributing more collectively than her individual numbers had ever covered for.

A brilliant but abrasive staff member doesn't work in a residential setting. When they see a staff member being talked over during a handoff, and they're being taught that contempt is normal between adults who say they care about each other, then I'm done coaching. That was the line I set before I knew whose name would land on it.

The conversation about performance and the conversation about conduct each have their own meeting, their own week, their own notes. When we fold conduct into performance, the strong numbers always buy a discount on the conduct, and the team can smell that discount across the building. By splitting the conversations into two, it felt easier to call that hard conversation. When split up, the praise is genuine and the correction is genuine.

I apply the standard to the top of the org chart. If I am not going to hold the director to that, I should not hold the tech to that.

After more than a decade in healthcare operations, much of it at Penn Medicine before starting Alpas, I've learned that people don't leave because your standards aren't high enough.

I decide by watching what the team starts doing around the person. If the nurses are starting to time their questions around a clinician, then the counselor is not sharing what he or she should have shared at shift change. If it's not interpersonal but it's structural, then coaching can change that. If they quietly changed the way they were working to do what needed to be done, then coaching was not the right answer. If something is not being shared during handoff in a detox facility, then that's a patient safety issue. That's my line and that's how I handle that.

Until that moment I coach hard, and I coach specific. From that moment I move.

The boundary that made something that lands on a hard call fair is that productivity does not buy unpredictability. This is what I say in the first conversation, in those words, and I name the two or three behaviors I need to see change and the date that I will look again. What I no longer do is let the plan live only between the two of us. One time, I handled a situation privately, the person in question eventually left, and my team saw it as arbitrary, because they never knew a standard had been set.

I do not tell the group the details, I only tell them the standard. They can accept a loss with a strong performer but they can not accept that they do not know the rules.

We're working on all the new people. Not the star, not the people complaining about the star. But say somebody has been working here for four months, clips her tone in the same way as everybody else, or she's missing a handoff because she saw somebody else missing a handoff and nothing happened. We have gone from "It's a problem with this one person" to "It's a house standard." At that point it's my call. I can take care of one person. It takes a year to get a norm back after twenty people have seen it bend, and we don't have a year in a residential environment with clients who are in early recovery.

The other signal word is "but." In 12 years working in behavioral health, I have heard a lot of accountability conversations, and the conversations that get anywhere start with the person saying "I did X," flat-out, no "but" after it. "But we were short-staffed" means I'm coaching a defense attorney, not a clinician.

What makes a hard line fair was the fact that the damage was repaired in front of the people who saw the damage done. A private apology for a public dressing-down does not reset anything and just protects the star's image. When we started, two people fixed themselves and one person quit, and the team saw both as fair.

I quit coaching if I don't hear the apology, I get the apology. Then I get the person who got hurt. Okay, these are the people that are effective in this business development of ours, and the ones that are good at business development are great at persuasion, and great at fixing the relationship with my boss and leaving the intake coordinator exactly where she left off. I quit when I hear 'I know, I'll fix it, we're good' and the clinician doesn't hear a word, and that is not coaching behavior, that is me being managed.

I told the top performer plainly what had to be different. "I told the staff that I said this, and I asked them a month later, 'has anything changed?'" we had a conversation, not a survey. It was about the boundary where there was a hard call that was fair - letting the person on the receiving end grade the change, not me.

It's serving two purposes, it stops me from thinking a smooth flow is a fixed issue and it tells the team their read on the room counts for something.

I'm in long term recovery so I have a bias on this. Amends are made to the person you harmed, in their language, on their timeline. Everything else is "image management." The high performing person who won't do this is costing me three good people to keep one.

I had a contractor on my team who was pulling in more revenue per project than anyone else, but every week someone different came to me with a complaint about how he communicated. He was dismissive in group calls, ignoring shared processes, and going around people to get what he wanted faster. I gave him two direct conversations with specific examples and a 30-day window to adjust. I told him exactly what needed to change and how I'd measure it.

Two weeks in, I started tracking something simple. After every team call I asked myself whether anyone had gone quiet who normally contributed. That was my signal. By week three, two of my best people told me privately they were thinking about leaving.

I let the high performer go before the 30 days were up. His output was real, but the drag on everyone else was compounding faster than his contributions. Within a month my remaining team was producing more collectively than we had with him on board.

I ask whether the behaviour is a habit or a position.

Somebody who is short with colleagues under pressure, and who changes when it is pointed out, has a habit. That is coachable and worth coaching, because the underlying capability is usually real and the person is not defending the behaviour.

Somebody whose behaviour is a position, who believes their output entitles them to it, will not change. Coaching becomes a long private negotiation while the rest of the team watches it fail, and the failure is what they learn from.

The signal I trust is what happens after the first honest conversation. Not whether they agree in the room, where almost everyone agrees, but whether anything is different a month later.

The boundary that made it feel fair was applying the same standard upward. That conversation lands completely differently in a team that has seen a senior person corrected for the same thing.

What people underestimate is not the disruption. It is what the team concludes about the rules when somebody is visibly exempt from them.

Decide based on whether the behavior keeps eroding norms faster than coaching can fix it. Continue coaching if steady change appears and the group stays intact, but move on if the morale cost doesn't drop. Unmet standards after a clear support window is a clear signal that there needs to be a change.

Performance does not cancel out behavior that makes everyone else perform worse. I give people a clear opportunity to change, but the boundary is repeated behavior after expectations and consequences are explicit. In manufacturing execution, teamwork matters because supplier, quality, compliance, and shipment work depend on handoffs. A brilliant person who consistently damages those handoffs can become an operating risk. The fair test is whether the person responds to clear feedback. Talent matters, but accountability determines whether it is sustainable.

The importance of safeguarding the collective psychological safety of the team surpasses the significance of keeping the achievements of any particular high performer, irrespective of their expertise. Over the years of running complicated delivery teams, I have come to the conclusion that a toxic high performer sets a limit to the performance of the entire team, which cannot be compensated by the productivity of individuals in any manner. Once the behavior of the high performer starts affecting the morale of the team and the cultural standards of the company, they should be fired not just because it is the obligation of the human resources department, but also because it will be beneficial for the future success and scalability of the project. The high speed of work or technical efficiency of one person cannot outweigh the damage done to the atmosphere of professional trust in the working environment.

To differentiate coaching from letting an employee go, I use the behavior assessment tolerance threshold. This is how I come up with the corrective approach that starts with informing the employee of the feedback regarding their behavior and how it impacts relationships with their colleagues. If the employee says something like "I have been delivering consistently high metrics and have a great technical background, and that is why I should not follow behavioral norms," they prove that they put their ego above the interests of the entire team. At this moment, they demonstrate that they already view cultural fit as an unnecessary burden rather than a prerequisite for their success.

It is essential to act right in these situations to maintain the respectful environment. Having observed how the leaders refuse to deal with toxic performers, I can confidently say that the talented people remaining in the team tend to look for new job opportunities whenever they feel that their organization is not serious about its principles.

What made the call feel fair was deciding the rule before I had a name attached to it. On a remote team of about 60, you rarely see morale drop directly. You see people stop tagging someone in the channel and start routing work around them.

The boundary I hold to is that output stops counting as a defence once 2 teammates raise the same behaviour without comparing notes. That happened with one of our strongest salespeople. He got 6 weeks with one specific thing named, which was how he spoke to juniors on calls.

He did improve for about 4 of those weeks. You can coach a behaviour for a long time if the person actually wants it gone. His laptop came back by courier to Jaipur on a Thursday.

One boundary we have always relied on is simple and clear for everyone daily. We welcome honest disagreement because better ideas grow through open discussion without fear hesitation. We do not accept words or actions that make any colleague feel smaller personally. That clear boundary gives everyone the same standard and removes personal bias during collaboration.

When someone crosses it we start with direct feedback and a request to improve. We watch whether positive change appears consistently during stressful moments across meetings. If the behavior continues we decide based on team responsibility rather than talent alone. Keeping this boundary builds lasting trust because everyone knows respect guides how we work together.

When a high performer's behavior starts hurting team morale or standards, I coach them only as long as I can see real behavioral change, not just better promises. In coaching athletes, I've had talented people who could move impressive weight but made everyone around them tense, dismissive, or less willing to train together. My boundary became simple: after a direct conversation and clear expectations, I watched whether the same damaging behavior continued. If someone understood the impact, had a fair opportunity to correct it, and still repeatedly crossed that line, their performance numbers no longer justified the cost to the group. Moving on was difficult, but it also showed everyone else that being talented doesn't give you permission to undermine the environment. In my experience, teams become stronger when people know the standards apply equally to the strongest performer in the room.

I'll coach a fast, strong cleaner longer only while the harm is shrinking. Once it stops shrinking, the clock is up. Great reviews buy someone patience, but only a limited amount of it.

The signal I watch is what the rest of the crew starts doing around that person. When teammates quietly swap shifts to avoid working with someone, the group is telling me the cost is already being paid. The other one is redone work. If a strong cleaner skips a step and a teammate has to go back and fix it, two people are doing one job. The team notices that within days, and they watch what I do about it.

What makes the call feel fair is that the standard gets said out loud before anything goes wrong. I name the specific behavior and give a set number of jobs to show it's changed, all in writing. After that, the decision is only about whether the behavior changed. The person knew the terms from the start, and so did everyone watching.

The checklist is the same for every cleaner, so a top performer is held to the same list as the newest hire. Skipped steps can't be excused by a good track record. When crews see that, they stop testing where the line is.

The hardest cases involve people whose contribution is undeniable and whose conduct is steadily narrowing the team around them. The decision becomes less emotional when leaders ask a practical question, would this behavior be acceptable if the person's output were average? I use that test to prevent exceptional performance from distorting basic judgment.

One boundary was that feedback could not be treated as a loyalty test. Colleagues had the right to raise concerns without being frozen out of decisions, denied information, or labeled difficult afterward. This created a clear measure of whether the person could work within a healthy culture. When feedback triggered retaliation rather than reflection, the choice to exit became an act of protection for every person still willing to speak honestly.

Talent calls clarify when viewed through those who depend on someone after meetings. A standout operator can preserve appearances upward while creating uncertainty for colleagues handling deadlines and stakeholder expectations. Coaching is warranted if trust can be rebuilt through observable behavior, not apology.

The boundary was simple, no surprises for the person responsible for the client relationship. A high performer withheld risk until a solution was available, believing this protected confidence. It instead prevented informed choices. Early escalation triggers were agreed and reviewed weekly. When risks remained concealed, I concluded the person valued control over shared accountability. The exit reinforced that transparency is part of performance, particularly where teams carry promises for one another.

My internal referral flow tells me before HR ever gets a complaint. These employees who were previously referring their friends, co-workers and classmates are no longer referring people. When I ask why, nobody says a name. They said, "The department is a lot right now." That is the group protecting people they care about from a person I am still calling a high performer.

When I see it I'm going to look at one little statement in the coaching conversation: what they say - is it about their intent, or their impact? It's intent talk, that's a story about them. It's impact talk; they've already watched the faces in the room. I'm going to coach the intent talker, but only after they can say out loud what the night tech felt. My psychology degree from UNC taught me that people can learn perspective-taking, though what they cannot learn on my timeline is caring whether they need it.

Written dates are set for coaching, and the person sets the end date with me, since the boundary made a hard call land as fair. 90 days or behaviors on calendar, never extending silently.

Tell the team the standard, never the personnel details and finally I tell the team about the rules, not the people. They're going to accept a hard outcome if they've seen the rules before they've seen the consequence.

We had a developer who shipped faster than anyone on the team, caught bugs others missed, but would openly critique how other people wrote code in standups. Not mean, just brutally honest. Two people asked to switch projects within a month.

I let it go too long because the output was real. Then I realized the cost: we weren't hiring the next person because candidates felt the culture during interviews. One prospect literally said, "your team seems smart but kind of sharp with each other."

The boundary I set was specific: feedback only in 1-1s or async code review, never in group settings. Not "be nicer" or "be more collaborative," just a structural rule. If he wanted to comment on someone's approach in standup, he had to ask them first in a separate Slack thread and get permission to bring it up publicly.

He actually kept it. The surprising part: once the public critique stopped, his 1-1 feedback became more useful because people weren't defensive. He ended up mentoring two people instead of just shipping solo.

The signal that made it feel fair was that I didn't change what he could *say*, just *where* and *when*. I didn't ask him to fake positivity. He could still be direct. I just moved the directness to a setting where it could be received instead of performed.

If he'd pushed back or the behavior returned, I would've moved him. But I gave the boundary a real chance first, and it worked. That's different from tolerating it indefinitely because he was fast.

Look at what your recognition program has been rewarding before you decide anything. If the only criterion for the top performer award is the number, the company has been telling the whole team that how you achieved it doesn't matter. Coaching someone out of a behavior is an uphill battle when the award program keeps rewarding them.

The boundary I'd set is written criteria for every performance award: the result, how it was reached, and a peer-nomination step to limit eligibility. That gives the manager a signal that isn't a gut call. If the people who work next to the high performer won't nominate them, that's the conversation to have, and it's fair because the standard was written down before anyone was measured against it.

The line I draw is "I make the decision as it happens when behavior shifts from the staff room into the client room. Being sharp with a peer at a shift change is a coaching issue. Being sharp with a peer in front of a guy on day three of a benzo taper isn't a coaching issue, the client just discovered that this place isn't safe, and I can't teach that back out of him. That's my line and I say it out loud in the first conversation, so nobody finds out where the line was by stepping over it."

The other indicator I look for is whether they can tell me what the harm is without me coming up with that for them. "Tell me how Tuesday landed on the tech standing there." If they can do it accurately, even defensively, then we can get into coaching. If they go straight to numbers, to census, to outcome, it's about their performance.

I'd given them a written 30-day period with three specific behaviors that was dated, and a promise that I'd tell them exactly where they are halfway through and there will be no ambush.

I've been through all the jobs in a treatment facility, and I was a client before that. They know before the director does.

First I had to ask myself, am I coaching a skill or am I bargaining? I can close the skill gap in a few weeks now once the behavior has been labeled plainly, but willingness does not move within a few weeks. I can quickly recognize the difference, if I need to explain something for a third time, I am no longer coaching, I am bargaining, and the team is watching me bargain.

The signal I trust most is the workaround. If people are changing work schedules so that they don't have to work with someone, or if a nurse waits for a different admissions coordinator before asking a question, I already decided on it, I just have late paperwork. I have a policy that if you change your work schedules not to work with someone, we end the coaching as a private project and document the expectation and deadline as a dated written expectation.

I think "It seemed necessary for the apology to go to the person who absorbed the behavior, and saying sorry to a CEO is career management, but saying sorry to the tech I talked over in front of a client is repair" sums up what I was getting at.

In an addiction treatment program, staff is the treatment environment and I hold that line harder than I would in a business where the product sits on a shelf. The team got quieter about the individual and louder in clinical meetings, which is exactly what I wanted.

I decide based on whether the high performer respects the team's agreed norms for experimentation and knowledge sharing. If they use autonomy to test ideas but reliably share results and align with priorities, I continue coaching; if they persistently ignore those norms and undermine the group, I move on to protect the team. One boundary I used was explicit autonomy paired with a requirement: choose your tools and run experiments, but report what you discover to the team. That rule felt fair because it set a clear expectation and it strengthened the group by surfacing learning and reducing friction.

We avoid treating a difficult employee as a sunk cost because that can mislead leadership decisions. Past results still matter but we focus on what the person adds from this point forward. We ask whether they strengthen the team's ability to make sound decisions during pressure together. Strong performance alone is not enough if it weakens trust or shared judgment.

We expect active participation in the repair process after conflict. We ask the employee to explain the impact listen to colleagues and suggest a way to prevent problems. When someone refuses that step we see a values concern instead of a coachable skill gap. This shared standard keeps our decisions fair and easier for everyone to understand.