Keeping Your Best People From Leaving

The resignation email arrives and it's a surprise to everyone except, in hindsight, it shouldn't have been. The signs were there for months: quieter in meetings, less likely to raise a concern, growth conversations that kept getting postponed. By the time someone says they're leaving, the actual decision was usually made weeks earlier, and whatever is offered now is competing against a decision that's already settled, not against an open question.

Why does generic retention advice fail here?

"Offer a raise and show appreciation" treats departure as primarily a compensation and recognition problem, solvable with a counteroffer or a thank-you. It misses that most quiet departures are preceded by a longer period of disengagement, driven by a lack of growth, an unaddressed concern, or a manager relationship that's quietly deteriorated, none of which a raise alone fixes.

Generic AI prompting tends to default to the same surface-level fixes. Ask an assistant "how do I retain my best employees" and it will often return a checklist of perks and recognition tactics without addressing the deeper, harder-to-measure signal: whether people on the team feel safe raising concerns, see a real path forward, and trust that honest feedback flows in both directions.

The deeper issue is that by the time a resignation is handed in, the real decision point has usually already passed. Retention that actually works happens continuously, in how performance conversations are handled and how safe people feel raising a problem, not reactively once someone has already decided to leave.

By the time a counteroffer conversation happens, the employee has usually already mentally rehearsed leaving, interviewed elsewhere, and received an offer. Reversing that momentum with a single conversation is a much harder ask than the ongoing, lower-effort work of never letting the disengagement reach that point.

What does the retention research actually say?

Amy Edmondson: disengagement starts with low psychological safety, long before resignation

Edmondson's The Fearless Organization (2019) documents psychological safety, the shared belief that it's safe to take interpersonal risks in a group, as a leading indicator of engagement. A team member who doesn't feel safe raising a concern or admitting a mistake tends to quietly disengage well before actively searching for another role.

Applied: Rather than waiting for a formal engagement survey, watch for early signals: someone who's stopped raising concerns they used to raise, stopped disagreeing in meetings where they once would have, or started hedging around topics they used to speak plainly about. These shifts often precede a resignation by months and are addressable if caught early.

The pattern matters because by the time someone explicitly says they're unhappy, the quiet withdrawal has usually already been underway for a while. Catching the early signal requires attention to behavior change, not just waiting for someone to say something is wrong. Edmondson's research suggests this kind of withdrawal is often a rational response to a specific moment where raising a concern went badly, not a general personality trait, which means it's frequently reversible if addressed directly.

Patty McCord: be honest about performance and business reality, especially with top performers

McCord's Powerful and the Netflix Culture Deck she helped build document a specific bet: top performers generally respond better to honest, specific information, about their own performance and about the business, than to being managed around difficult truths. Treating high performers as capable of handling reality, rather than shielding them from it, tends to build more trust, not less.

Applied: In conversations with strong performers, be specific and honest about both what's working and what the business is actually facing, rather than defaulting to generic praise or withholding difficult context. Most high performers can tell when they're being managed with a soft touch, and it reads as a lack of respect for their judgment, not kindness. McCord's documented experience is that this group, specifically, tends to leave organizations that seem to be managing around them rather than with them, regardless of how generous the compensation is.

Laszlo Bock: use actual data to catch what instinct alone misses

Bock's Work Rules! documents Google's evidence-based approach to people operations: using real engagement and performance data to identify patterns, who's at risk, what's actually driving attrition, rather than relying purely on managers' individual instincts, which vary widely in accuracy and attention.

Applied: Look at the data you already have, engagement survey trends, time since last promotion or role change, patterns in one-on-one notes, for a specific person you're concerned about, rather than relying solely on a general sense of how they seem. Data doesn't replace judgment, but it catches blind spots instinct alone tends to miss, especially with team members who are quieter by nature.

Bock's framework at Google treated manager instinct as a valuable but demonstrably inconsistent signal across a large organization, worth checking against data rather than trusting uniformly. The data isn't meant to override a manager who knows their team well; it's meant to catch the cases where that knowledge is thinner than it feels.

Where does each approach fit best?

Who's actually at risk, and what can you run today?

Why it works: naming specific behavioral changes, rather than a vague sense that something's off, makes it easier to decide whether a conversation is warranted and how to start it without sounding alarmed.

Why it works: a drafted, specific opening is easier to deliver well than an improvised one, especially when the instinct under pressure is to soften a message that needs to stay clear.

Why it works: checking the actual data against gut feel catches cases where a quiet, undemonstrative person is at real risk despite seeming fine on the surface.

What's the honest limit here?

These approaches improve how early disengagement gets noticed and how honestly performance gets discussed. None of them can prevent someone from leaving for a reason entirely outside your control, a life change, an opportunity genuinely better than anything the role could offer, or simply a readiness to do something different. Retention done well improves the odds considerably. It doesn't guarantee anyone stays, and treating every departure as a preventable failure misreads the ones that genuinely weren't.

Amy Edmondson, Patty McCord, and Laszlo Bock all sit in the People & Culture category alongside Kim Scott. For writing a specific, honest, defensible performance review without the usual HR language, Performance Review Writer is a $79 tool built for exactly that.

All the copy-paste prompts from this guide, and the rest of the Situations series, live at /prompts too, free to copy without reading the guide first.