By the time someone hands in their notice, the decision is months old.
This is one of the most consistent findings in employee research, and one of the least acted upon. The resignation letter is the final visible signal in a sequence that typically started with a shift in sentiment six to twelve weeks earlier. A change in engagement, a growing disconnection from the team or the role, a sense that raising concerns won't change anything. The moment it shows up in a manager's inbox is almost never the moment it began.
This matters because most organisations respond to turnover reactively. When someone leaves, you find out why. You conduct an exit interview, if you have a process for one. You ask questions, get partial answers, file the information, and perhaps make a note for the next performance cycle or hiring brief. By which point the context is stale and the person is gone.
Exit interviews are a particular problem. Not because they're useless (they can surface information you might not get at all otherwise). But because people leaving a job are weighing the same calculation they've been weighing for months: is it safe to say this? They don't want to burn a bridge. They want a reference. They may be moving into a space where they'll encounter the same people again. So they give the polished version. " Great opportunity came up ." " Right time for a change ." The real reason is often something they decided wasn't worth the risk to say out loud.
What you need isn't better exit interviews. It's the data from the weeks before the decision was made.
Organisations with weekly feedback data can see the moment a team member's sentiment dropped. They can see whether it was isolated or whether others were feeling the same thing at the same time. They can see whether it was connected to a specific event - a restructure, a performance review, a change in management - or whether it was a slower drift that went unaddressed. They often can't see who dropped, because the data is anonymous. But they can see that something shifted, and when, and whether the response that followed moved things back.
This is useful in a way that annual data simply isn't. A once-yearly snapshot taken in March tells you roughly how people felt in March. It does not tell you what happened in August. It does not tell you that sentiment in one part of the business dropped in October and never recovered. It does not tell you that the three people who resigned in January had been showing declining engagement since September.
The lag between signal and consequence is where real-time data creates real operational value. Not in the sense of predicting who will resign, that kind of framing misses the point. But in the sense of knowing when something shifted, and having enough time to do something about it before the cost arrives.
Most organisations learn what their teams were thinking six months after it mattered. Relay tells you now.
If you're relying on exit interviews to understand why people leave, you're reading the end of the story. Set up Relay in five minutes and start reading from the beginning.