The exit interview is HR's best attempt to learn from a failure that has already happened. Someone has decided to leave. They've probably already told their manager. They may have already signed an offer somewhere else. And now you're asking them to tell you, honestly, why they're going.
Most people don't tell you honestly. They cite career development or a new opportunity or wanting a change. These things may be true. They're rarely the whole story. The full story - the manager who made them feel small, the promotion that went to someone else without explanation, the restructure that was communicated badly and never properly recovered - usually stays unsaid. The relationship is ending. There's nothing to gain from burning it.
So the exit interview data you collect is filtered, diplomatic, and months out of date. The decision to leave was made weeks or months before the notice was handed in. The underlying problem is older still. You are learning about something that started in the past and ended in the recent past, from a person with every incentive to keep it vague.
This isn't an argument against exit interviews. Done well, they can surface systemic patterns over time, and some people do use them to say difficult things they couldn't say earlier. But they are not a retention tool. By definition, they can't be. The person is already gone.
Retention happens before the decision is made. Research on voluntary turnover consistently shows a gap of weeks to months between the moment someone starts seriously considering leaving and the moment they hand in notice. That gap is the window. It's the only one available, and most organisations don't know it exists because they have no mechanism to see into it.
What the window looks like in data: a sentiment decline, often gradual, sometimes sharp after a specific event. A drop in response rates as someone starts to disengage from the process. A shift in the themes of free-text responses - from specific and engaged to vague and brief. These aren't definitive signals, but they're visible, they're early, and they're actionable in a way that an exit interview never is.
The shift this requires isn't complicated. It's listening continuously rather than at the end. Not because exit interviews are bad, but because waiting until someone is leaving to ask how they're doing is too late, and everyone involved usually knows it.
Relay gives you a signal before the decision is made. Not to surveil people or predict individual resignations - but to know when something has shifted, so you can respond while there's still something to respond to.