The Engineering Manager's Guide to Performance Reviews
2026-06-22 · 9 min read · Career & Promotions
Performance reviews are one of the most high-stakes things an Engineering Manager does. They affect compensation, promotions, and how people feel about their career trajectory. Done badly, they destroy trust and make your best people update their CVs. Done well, they're one of the most powerful tools you have.
Most managers do them badly — not because they're careless, but because they've never been taught how.
The Problem with Most Performance Reviews
The typical performance review process looks like this: once a year, a manager scrambles to remember what someone did over 12 months, writes something vague and balanced, puts a rating in a box, and has an awkward conversation. The engineer leaves either confused ("what do I actually need to do to get promoted?") or frustrated ("I shipped three major features and got a 'meets expectations'").
This happens because most performance reviews are:
- Backward-looking without being useful: recapping what happened rather than what to do next
- Based on recency bias: the last 2 months count for 80% of the assessment
- Disconnected from day-to-day feedback: nothing in the review should be a surprise
The goal of a good performance review is to give someone a clear picture of where they stand, why, and what specifically would change that assessment.
Preparation: The Real Work
The review conversation itself is maybe 30 minutes. The quality of it is determined entirely by the preparation you do beforehand.
Gather evidence, not impressions. Go back through your 1:1 notes, incident reports, project retros, and Slack messages. What did this person actually do? What were the outcomes? Where did they have impact beyond their immediate work?
Collect peer input early. Don't rely only on your own observations. Ask 3–4 people the engineer works closely with: "What's one thing [name] does particularly well? What's one area where you think they could have more impact?" Weight this input but don't delegate the assessment to it.
Map to the engineering ladder. Be explicit about which level the person is performing at, which behaviours they're consistently demonstrating, and which ones are missing or inconsistent. This is the difference between "you're doing well" and "you're operating at L5 in technical scope but at L4 in cross-team influence, which is what's holding back the promotion case."
Check for bias. Before you finalise your assessment, ask yourself: am I rating this person based on their output, or based on how visible they are? Quiet, heads-down engineers often get underrated. Confident self-promoters often get overrated. Adjust accordingly.
Structuring the Conversation
A 30-minute performance review conversation should be structured roughly as:
5 min — Set the tone. This isn't a verdict, it's a conversation. "I want to share my perspective on how things have been going, and I want to hear yours. Nothing I say is final — I want us to leave with a shared understanding."
10 min — Their self-assessment first. Ask them how they think the period went before you share your view. This does two things: it tells you if there's a significant gap in perception (a red flag), and it gives them ownership of the conversation.
10 min — Your assessment. Be direct. Share what you see as their key contributions and strengths, then the areas where you think there's a gap. Use specific examples. Don't soften the hard parts to the point where they don't register.
5 min — What next. What does success look like over the next 6 months? What support do you need to provide? What do they need to do differently? Make this concrete.
The Rating Conversation
If your company uses ratings, the rating should never come as a surprise. If it does, you haven't been giving enough feedback throughout the year.
When sharing a rating lower than someone expected, be direct about why: "I've given you a 'developing' rating because while your technical work is strong, you're not yet operating at the scope or independence we'd expect at your level. Here's what that would look like..."
Don't over-explain or apologise for the rating. Explain it, give examples, and move to what changes it.
Calibration: Protecting Your Engineers
Most companies run calibration sessions where managers compare ratings. This is where bias often enters the system — the loudest voices in the room disproportionately influence outcomes.
Prepare for calibration. Bring written evidence for each person you're advocating for. Know your engineers' accomplishments well enough to defend them against other managers who don't have the context you do. If someone's getting downgraded in calibration, ask what evidence is being used.
Your job in calibration is to be your team's advocate — not to agree with the room to avoid conflict.
Continuous Performance Management
The best performance review is one with no surprises, because everything in it has already been discussed.
This means:
- Giving feedback in the moment, not saving it for the review
- Checking in on growth areas in 1:1s throughout the year
- Being explicit when someone is at risk ("I want to flag that if the communication pattern we discussed doesn't change, it's going to affect your rating")
Performance reviews should feel like a summary of an ongoing conversation, not a new one.
When the Conversation Gets Hard
Sometimes someone disagrees strongly with your assessment. Stay grounded:
- Listen to their perspective fully before responding
- Acknowledge what's valid in their view
- Be clear about where your assessment stands and why
- Don't revise your assessment under emotional pressure — but do reconsider it if they bring evidence you hadn't considered
It's also okay for someone to be disappointed. Disappointment is a natural response to an assessment that doesn't match someone's self-perception. Your job isn't to make them feel good — it's to give them an honest picture of reality that helps them grow.
The managers who are most respected are usually the ones who were honest when it was hard, not the ones who told people what they wanted to hear.