Engineering Management

Competing paths that diverge at L4. Staff engineers increase impact through technical depth and influence without direct reports (L4-L7 salaries $200-400k+). Managers increase impact through team leverage and organization building ($150k EM → $300k Director → $600k+ VP). Choose staff if: you prefer technical problems, architecture, mentorship at scale, autonomy. Choose management if: you enjoy 1-on-1s, hiring decisions, career development, organizational strategy. Both paths are equally valued; most orgs need both. Some people do both (tech lead manager) for 1-2 years before choosing.

IC Manager (L2-L3): Direct reports code, you unblock them daily, you own delivery, you attend their 1-on-1s. Manager of Managers (L3-L4): Reports don't code, they manage people, your 1-on-1s cover their hiring/retention/org design, you rarely know the tech details of their team's work. Skip-level 1-on-1s become critical (you need to know IC sentiment even though you don't manage them directly). Biggest challenge: letting go of code to focus on org health.

Feedback only works if it's: (1) timely (same week as the behavior), (2) specific (example, not generality), (3) about behavior not character, (4) invited (person asked or you have standing), (5) actionable (here's the fix, not just the problem). Worst: 'You're not a team player' (vague, judgmental). Better: 'In the meeting on Tuesday, you didn't ask for opinions before deciding the approach. I want to see you spend 5 minutes gathering input next time.' Modern frame: Radical Candor (care personally + challenge directly) or SBI (Situation-Behavior-Impact). Deliver in 1-on-1, never in group.

30min, same time weekly, YOUR calendar, never cancel, no status updates (they can email status). Use the first 5min for them to set agenda (what do you want to talk about?). Then: career questions (where do you want to go?), blockers (what's slowing you down?), feedback (here's what I noticed), personal stuff (how are you?). Take notes so you follow up next week. Skip 1-on-1s predict burnout 6 months later. Over-communicate in remote setups — you lose hallway convos.

PIPs (performance improvement plans) are usually the end, not a reset — most people don't recover. If you have doubts: (1) Document everything in the 1-on-1 notes (weekly written record), (2) Get legal/HR involved before the PIP conversation (don't DIY this), (3) Make the bar clear (what does success look like in 30/60/90 days?), (4) Have check-in meetings every 2 weeks, (5) Escalate to your manager if it's not working. Big red flags: Sudden poor performance (health issue? manager issue?) vs pattern (should have been addressed 6mo ago). Never surprise people in a final conversation.

The '10x' problem. High output but poisons the team or blocks others. Three moves: (1) First 1-on-1: 'Your code is great AND you're creating tension. Here's what I've observed [specific examples]. It needs to change.' (2) Give them 30 days to course-correct with clear metrics (fewer complaints, better code reviews, respectful meetings). (3) If no change, let them go — one toxic person costs more than the productivity gain. Your job is not to salvage individual genius, it's to build a team that compounds. Staying too long signals to the team that bad behavior doesn't have consequences.

Market rates (use Levels.fyi, Blind, company surveys). New hire at L3 should get $190-210k, not $160k because the last hire was underpaid. If internal people are underpaid vs market, fix it in the next review cycle (2-5k bumps, document it). Compression kills morale fast ('new guy makes the same as me after 3 years?'). Your CFO will push back on budget; push back harder with data. No transparency on salaries = people always think it's unfair.

Psychological safety = people don't fear embarrassment/punishment when taking interpersonal risks (asking dumb questions, admitting mistakes, disagreeing). How to build it: (1) Invite input in meetings, then actually use it, (2) Admit your mistakes publicly ('I made a wrong call on that roadmap, here's what I learned'), (3) Ask questions instead of declaring answers, (4) Respond to bad news with curiosity, not blame ('that bug happened — what system failure let it slip?'), (5) Protect people who speak up (if someone questions a decision and you overrule them, follow up 1-on-1). Over 6 months, teams with high safety ship faster and retain better.