Feedback to a senior engineer who thinks they're irreplaceable: a scenario playbook
The 10x engineer who's also tanking team morale is every engineering leader's hardest call. A step-by-step script for the conversation, the 60-day decision…
On this page▾
- The 'brilliant jerk' problem is solved by treating the behavior as a performance issue, not a personality one. Engineers respect specifics.
- Open with impact data, not feelings. 'Three engineers asked to leave the team in the last quarter — two cited your code reviews' beats 'people find you difficult'.
- Give a 60-day window with three measurable behavioral outcomes and a written decision date.
- Be ready for the bluff — 'I'll leave' is fine. Plan the bus factor before the conversation, not after.
Senior engineers know they're hard to replace. That confidence is what makes them senior — and it's what makes the feedback conversation feel impossible. The trick isn't to soften the feedback. It's to remove every escape hatch where the conversation becomes about 'feelings' or 'fit' instead of 'this is what's measurably true about the impact on the team.'
Pre-work: bus factor + impact data
- Identify the 3-5 systems where they're the only deep expert. List what would break if they left tomorrow.
- Assign a shadow / pair for each — start before the conversation, not after.
- Pull data: peer-review feedback (anonymized), exit interviews from people who left their team, 360 if available.
- Talk to skip-levels of 3 engineers on their team. Ask: 'If you could change one thing about working with [X], what would it be?'
- Pre-align with your skip: 'I'm having this conversation Thursday. Here's the data. Here's the 60-day plan. If they escalate, here's what I need you to say.'
If you walk into this conversation needing them more than they need you, you've already lost. Spend the two weeks before to make 'we'd be fine without you' quietly true.
The opening conversation
'I want to talk about something specific. In the last quarter, three engineers asked to move off your team. In two of the three exit conversations, the reason was code review interactions with you — specifically [example 1, example 2]. I'm not telling you this to relitigate any one moment. I'm telling you because the pattern is now expensive for the team and for you. You are one of the strongest engineers in the company. That is exactly why this conversation matters — your output is high, and your taxes on others are also high. We need the first to keep being true and the second to change. I want to spend the next 30 minutes on what specifically needs to change, and the 30 after on what I can do to help.'
Three things to do in this conversation: name the specific behaviors with examples and dates; tie them to measurable impact (people left, PRs slowed, recruits declined); set the 60-day window in writing before they leave the room.
Three measurable behavioral outcomes
- 1Outcome 1 — code review tonePeer rater (rotating) reviews 5 random PRs/week. Scores: 'Would I want this reviewer if I were a new hire?' Target: 4/5 yes by week 4, 5/5 by week 8.
- 2Outcome 2 — meeting behaviorSpecific: stop interrupting in design reviews. Measure: notes from a designated observer (peer or manager) at 3 meetings/week. Zero interruptions by week 6.
- 3Outcome 3 — onboarding loadPair with two newer engineers, 2 hours/week each. Their week-8 1:1 feedback to skip-level is the measurement. Target: 'I learned, I felt respected.'
Handling the 'I'll leave' bluff
'I hear you. I'd rather you stay and we work through this — that's why we're having this conversation now instead of in six months when it would be too late. If you decide to leave, I'll support a clean exit. I won't be held hostage by the threat, and I think you'd respect me less if I were. Take the weekend. If you want to come back Monday and try the 60-day plan, I'm in. If you want to plan an exit, we'll do that with dignity.'
Two out of three times in my experience, they stay. One out of three, they leave. Both are acceptable outcomes — what is not acceptable is keeping a brilliant tax on the team because you're scared of the resignation.
Day 60 — the decision
- Pull the three outcome scores 48 hours before. Look at trend, not just endpoint.
- If clear improvement: written acknowledgement, reset the relationship, never reference it again unless it slips.
- If partial improvement: one 30-day extension, with one of the three outcomes tightened. Not two extensions.
- If no improvement: separation conversation. The bus factor work you did 60 days ago is what makes this possible.
- Performance Reviews People Don’t Dread
- The PIP that doesn't feel like a trap: scripts, milestones, paper trail
- The high performer behaving badly: a step-by-step HR playbook
- Feedback Frameworks That Land
- Difficult Conversations: A Manager’s Field Guide
- The Engineering Manager's First Year: From Lead to Multi-Team Manager
- 1:1 Meetings That Actually Help
- Goal-Setting Frameworks: OKRs vs MBOs vs SMART — Which Fits Your Team
Read next
All playbooksThe system around the review matters more than the review itself. A modern approach to goals, feedback, calibration, and the conversation.
Most PIPs are exit documents in disguise. Here's how to design and deliver one that's actually defensible — 30/60/90 milestones, documented check-ins, and the…
The brilliant jerk problem — how to keep standards when the offender is your top revenue or top engineer. A 5-step playbook with scripts for the manager, the…