Engineer retention: why people leave and the signals you missed

Every engineer who resigns feels, to their manager, like it came out of nowhere. It almost never did. By the time someone hands in their notice, they’ve usually been leaving for months — the decision was made weeks before the conversation, and the signals were there the whole time, just unread.

Retention isn’t about counter-offers and ping-pong tables. It’s about understanding why good engineers actually leave, and learning to read the slope before it becomes a goodbye.

Why engineers really leave

Exit interviews are polite; the real reasons cluster into a short list:

• Their manager. The oldest truth in management: people leave managers, not companies. A manager who doesn’t listen, doesn’t support, or doesn’t have their back is the most common root cause.
• No growth. Engineers who stop learning start looking. If someone can’t see how they’re getting better or where they’re headed, they’ll find somewhere they can.
• Burnout. Sustained overload with no recovery doesn’t resolve itself — it resolves with a resignation.
• Feeling unseen. Effort that’s repeatedly unrecognised curdles into resentment. People want their work to matter and to be noticed.
• Compensation — but usually second. Pay matters, but it’s often the reason people give rather than the reason they go. A happy, growing engineer rarely leaves over a moderate pay gap; an unhappy one uses it as the exit.

The signals that precede a resignation

Almost every departure is preceded by a recognisable arc. The signs:

• Withdrawal. Less presence, less chat, camera off, shorter answers. The disengagement starts before the decision.
• Dropping discretionary effort. Initiative fades, scope shrinks to the minimum — the quiet-quitting pattern. We cover it in how to spot quiet quitting.
• 1:1s going flat. They stop bringing topics, stop raising frustrations, let the meeting slip. They’ve stopped investing in the relationship.
• A change in the calendar. Suddenly taking that accumulated PTO, the occasional unexplained afternoon — sometimes the interviews are already happening.
• Engagement trending down. Not one bad week — a sustained slide over a month or two.

What you can actually do

• Have stay conversations, not just exit ones. Ask, while they’re still happy, what would make them consider leaving — and fix it before it’s a live question.
• Make growth visible. A real sense of the next step, and regular movement toward it, removes the most common reason engineers drift away.
• Recognise effort specifically and often. Cheap to do, and it addresses one of the biggest silent drivers of attrition.
• Act on burnout early. Reduce load before someone breaks, not after. By the time it shows, you’re already late.

How to run a stay conversation

The most underused retention tool is absurdly simple: ask your best people what would make them leave, while they’re still happy, and then act on the answer. A stay conversation isn’t a formal review — it’s a deliberate twenty minutes, once or twice a year, with the people you most want to keep. A few questions that work:

• What would make you seriously consider leaving, if anything?
• What keeps you here right now? What would you miss?
• Is your work giving you what you want from your career at the moment?
• If you could change one thing about your role, what would it be?
• Do you feel fairly recognised and rewarded for what you do?

The value isn’t the questions, it’s what you do next. If someone tells you they’re bored and you change nothing, you’ve confirmed their reason to go. If you hear it and act — a new project, a growth path, a comp conversation — you’ve solved the problem months before it became a resignation.

You can’t run these with everyone, and you don’t need to. Prioritise the people whose departure would hurt most — the quiet load-bearers, the ones with the deepest context, the ones the team is built around. Those are exactly the people who get taken for granted, precisely because they never complain.

Comp matters — but get the order right

None of this means pay is irrelevant. If someone is genuinely underpaid for their market and level, no amount of growth talk papers over it, and you should fight for them before it becomes a resignation rather than after. But comp is a hygiene factor more than a motivator: getting it wrong drives people away, while getting it right doesn’t, on its own, keep an unhappy, unchallenged engineer. Pay people fairly and proactively so it’s off the table — then compete on the things that actually make someone want to stay.

When the resignation comes anyway

Sometimes you’re too late, and someone resigns. A few things worth knowing in that moment. A counteroffer rarely works for long — the reasons someone reached the point of interviewing are usually deeper than the number, and most people who accept one leave within a year anyway. The resignation is also information: handle the exit well, ask honestly what you could have done differently, and take the answer seriously, because the same cause is probably acting on people who haven’t left yet. And how someone leaves is watched closely by everyone who stays — grace on the way out is one of the quietest retention tools you have.

Finally, be honest about which departures actually matter. Not every exit is a failure — sometimes a role has genuinely been outgrown, or the fit was never right, and the healthiest thing for everyone is a good goodbye. The ones to learn hard from are the regrettable losses: the people you wanted to keep, who left for reasons you could have influenced. Track those honestly, and let them sharpen how you manage the people still here.

Reading the slope in time

Here’s the uncomfortable part: in hindsight, the signals before a resignation are obvious. In the moment, spread across weeks and buried under the fact that the person is still doing their job, they’re easy to miss — especially across a whole team, relying on memory.

That gap is the entire reason CoManager’s signals exist: retention risk, disengagement and burnout surface weeks early, ranked by urgency, each with a concrete next step attached — built on per-person engagement and morale trends rather than a gut feeling. You can’t keep everyone. But most of the people you lose were telling you for months, in signals exactly like these. Retention is mostly the discipline of reading them in time.