Why Good Employees Leave Without Warning

The resignation lands and it feels like it came out of nowhere. One of your better people — reliable, tenured enough to matter, the kind you'd have named if someone asked who holds a team together. No complaints on record. The last engagement survey looked fine. And now there's two weeks' notice on your desk and a hole you didn't see coming.

If you've asked yourself why are good employees leaving without warning, or stared at a survey that didn't predict a single departure, the problem usually isn't that the warning wasn't there. It's that the warning wasn't in a form anything you run was built to catch.

Burnout is a measurable condition, not a mood

Christina Maslach spent four decades turning burnout from a vague complaint into a measured construct. Her work established that burnout isn't a bad week or a soft personality — it's a progressive occupational condition with a recognizable arc: emotional exhaustion first, then cynicism and detachment from the work, then a quiet collapse in sense of effectiveness. It builds over months. In 2019 the World Health Organization formally recognized it as an occupational phenomenon. The research is well past the contested stage.

The part operators should sit with is the sequence. By the time someone is detached and going through the motions — the stage most visible from the outside — they are already two stages into the arc. The exhaustion that started it was visible much earlier, to anyone measuring the right thing. And the people furthest along are very often your best ones, because the best ones absorb the most friction before they say anything. They cover the broken handoff. They re-do the work that came back. They hold the team together right up until they decide to stop.

Why the survey didn't catch it

An annual engagement survey is a sentiment snapshot taken on one day, answered by people deciding how candid to be. It measures how someone feels about the company at a moment in time. It does not measure how the work is actually going week to week — and it asks the question least reliably of the people already detaching, because detachment shows up first as going quiet. The person furthest into the arc is the least likely to write the comment that would have warned you.

That's the core limitation: the survey is a lagging report. It tells you how things felt after they already changed. What you needed was a leading indicator — something that moves while the person is still on the team, still doing the work, months before the notice.

That is the gap Organizational Intelligence is built for. Instead of asking people how they feel once a year, Sigma G reads the recurring friction in the actual work through four lenses — Challenges, Ownership, Drive, and Learning Strength. The signature of someone moving down Maslach's arc is specific and it shows up in operations long before it shows up in sentiment: high Drive going into the same Challenges over and over, Learning Strength flattening because nothing the person does changes the situation, Ownership quietly contracting as they stop raising things that never get fixed. Effort up, traction down, voice fading. That pattern is readable now, not in the exit interview.

What it costs to keep finding out too late

Turnover is usually the single largest line in an operation's hidden cost, and it's the most back-loaded — by common estimates replacing an employee runs between one-half and two times their annual salary once you count recruiting, onboarding, lost productivity, and the load that lands on everyone who stays. For a frontline-heavy organization that number compounds fast, because the departures cluster among the people who were absorbing the most. Our friction calculator models this on federal labor data for an organization your size; turnover is typically where the recoverable cost is largest.

The expensive part isn't any single resignation. It's that each one removes the person who understood a piece of the operation, hands their load to people already carrying their own, and pushes the next person a little further down the same arc. Unread, it's a cycle. Read early, it's an intervention.

There's a version of this concern that's more specific than "people are leaving." Sometimes you know exactly who is about to walk out the door — a tenured associate, a planned retirement — and the worry isn't morale, it's the knowledge leaving with them that nobody else has. If that's the sharper version of your problem, here's what a structured pilot looks like when knowledge is about to walk out.