There's usually a period between when someone decides to leave and when they tell you. It can be a few weeks. It can be several months. During that time they're still in your meetings, still replying to messages, still technically on the team. But something has shifted. The question is whether you have a way of seeing it.
Most managers notice something is off, eventually. It shows up in smaller ways first: slightly less initiative, a little less energy in conversations, responses that feel more guarded. The problem is that these are easy to explain away. By the time they're visible enough to prompt a direct conversation, the person has often already made their decision.
Anonymous weekly check-ins don't give you certainty about who is disengaging. But they give you signal well before the conversation needs to happen... if you know what to look for.
Sentiment scores flatten before they fall
The first sign is often not a dramatic drop. It's a shift to uniformly average responses. 75% positive drops to 65%. Negatives jump a little. Things are fine. Not bad, not good. Fine.
This matters because "fine" is a specific kind of response. It means the person has stopped investing in the question. They're not engaged enough to say things are going well, and not unhappy enough to say things are difficult. They've entered a neutral zone: still showing up, not yet leaving, but no longer present in any meaningful sense.
A sudden return to neutral after a period of more candid responses is a signal worth paying attention to.
Written responses get shorter and safer
Someone who is genuinely engaged, even when frustrated, tends to write specifically. " The handover meeting is running too long and it's cutting into my prep time ." " I'd like more clarity on the Q3 priorities before we go further. " That kind of thing.
Someone who is disengaging tends to write in generalities, or not at all. "All good this week." "Nothing to add." Responses that technically satisfy the field requirement but offer nothing a manager can act on.
Watch for a trend in the length and specificity of written responses. It often precedes other changes in the data by several weeks.
Specific sub-scores move before the overall score
The overall sentiment score is a blunt instrument. It aggregates across everything: workload, relationships, growth, purpose, recognition. Disengagement tends to show up in specific areas first, and the overall score can mask that shift for several weeks.
The most common early movement is in the verbatims. Someone who has mentally checked out stops caring about their development. Questions about whether they feel they're learning, whether their work matters, whether they can see a future here - these often appear before overall scores change materially.
If you're only tracking the headline number, you'll miss the early signal. Paying attention to verbatim comments week to week gives you a more targeted picture of what's changing and where.
Response frequency becomes irregular
Before someone stops responding altogether, they usually become inconsistent. The person who answered every week starts skipping. Then answering again. Then missing another. It's not a clean decline. It's a pattern of gradually decreasing investment.
The response rate is a trust and engagement signal in its own right. When someone stops seeing the point of answering, they start opting out. First occasionally, then permanently. By the time a team member has stopped responding entirely, they've often already made their decision. The irregular period is the window where thoughtful action might still change something.
The feedback stops because nothing happened last time
This last signal is as much about the feedback programme as the person. Someone who used to raise real concerns stops raising them. Not because things improved. Because they've concluded that raising concerns doesn't change anything.
They put something specific in the free-text field three months ago. Leadership acknowledged the survey results in a general way. The specific thing was never addressed. So they stopped raising specific things. And now they're looking elsewhere.
When you see the free-text go from specific to generic over time, the question to ask isn't what happened to this person. It's what happened to the last thing they told you.
None of these signals is definitive on its own. But patterns across multiple weeks tell a clearer story, and they're visible well before someone starts updating their CV. What disengagement rarely is, is sudden. It builds. And it's far easier to respond to in the early stages than after months of going unaddressed.
If you want to see these patterns in your own team, Relay gives you the weekly data to do it. Takes five minutes to set up.