The retention conversation you're not having

If you have led a People function for more than a couple of years, you have presided over a version of this conversation. An employee gives notice. Their manager is surprised. You look at the engagement scores for that team and they are, frankly, fine — middle of the pack, no flags. You go back through the 1:1 notes, if any exist. Nothing was raised. The exit interview produces a polite, vague answer about wanting a new challenge.

You quietly add a retention-risk model to the next quarterly People analytics review. The model uses email metadata, calendar density, tenure, and recent compensation movement. It identifies people at risk. It does not catch the next person who leaves either.

The reason is that you are measuring the wrong layer. The earliest reliable signal of disengagement is not in any of those data sources, and it is not in your engagement survey. It is in whether the employee is still showing up to think about their own work.

What disengagement actually looks like before it becomes attrition

The behavioral progression that ends in a resignation email tends to follow a recognizable pattern, if you have visibility into the right layer. It begins with a quiet stop. The employee stops doing the small acts of intentional thinking that mark someone who is invested in their own trajectory. They stop preparing for their 1:1. They stop reflecting on the week. They stop articulating goals for next week. The work continues; the meta-work about the work stops.

This happens months before the calendar density drops, before the OOO time spikes, before anything appears in the manager's view. It happens before the employee themselves has consciously decided to leave. It is the upstream signal of someone who has stopped owning their own engagement.

You cannot see this from a survey. The employee will continue to fill out the survey at the same rate, with answers that are slightly more neutral than they were a year ago. You cannot see this from email metadata. The job continues to require email. What you can see, if your tool produces it as a byproduct of the employee doing useful work for themselves, is the cadence at which they are still thinking about their work.

The retention dashboard you actually want

Imagine the chart at the team level. The horizontal axis is the week. The vertical axis is the fraction of seats on that team that completed a weekly reflection. For a healthy team, you see a steady line around 70-80%. The number flexes a little around holidays. People take vacation. Someone has a brutal launch week. But the line returns.

When you start to see that line trend down — when the fraction of seats reflecting weekly drops from 80% to 50% over six weeks — you are seeing the earliest reliable retention signal that any aggregate-only system can produce. You are not seeing who. You are seeing that something is happening to the team's relationship to their own work. The conversation to have is with the manager, about the team. Not with a specific employee about a specific score.

Open-loop closure rate is the second signal. An open loop is something the employee flagged in a 1:1 agenda that has not yet been closed with an outcome. A team where loops are getting closed at a healthy rate has functional 1:1 conversations and a manager who is following through. A team where loops pile up — flagged, never closed — has a relationship problem that will show up in attrition six months later. You can see it now.

The third signal is the gap. The employees who have stopped engaging with their own workspace stand out, structurally, against the team average. The system tells you that the team's collective last_active_at distribution has changed shape, without telling you which individual changed. You go to the manager. The manager goes to their team. The conversation happens at the layer where it can actually do something.

Why this works as a retention signal

The mechanism is straightforward once you name it. An employee who continues to draft weekly reflections, capture wins, build 1:1 agendas before important conversations, and close open loops is signaling that they still believe their trajectory at this organization is worth thinking about. The act of recording is the act of investment. When the investment stops, the disengagement is already complete; the resignation is the formal acknowledgement, often two or three quarters later.

Self-Determination Theory points at the same mechanism from the motivation side. Continued autonomous engagement with one's own work is the behavioral signature of intrinsic motivation. When intrinsic motivation collapses — usually because autonomy has been eroded, competence is no longer being developed, or relatedness with the team has frayed — the autonomous engagement stops first. The work continues, for external reasons. The intrinsic engagement is gone.

Amabile and Kramer's progress principle adds the daily lens: employees who can see their own progress over time stay engaged in a way that employees who cannot do not. A weekly reflection cadence is what makes progress visible to the employee themselves. When the reflection stops, the employee loses the felt sense of progress, regardless of whether the work is going well objectively.

Catch the cadence drop. You catch the disengagement at the layer where it is still addressable.

What this looks like in your function

A retention program organized around this signal looks different from the standard one. The dashboard reports four numbers per team: weekly reflection cadence, open-loop closure rate, distribution of last_active_at, and adoption rate at the seat level. Each gated below five active seats per team. No individual content. No sentiment scores.

When a team's cadence drops, the conversation is with the manager. The intervention is at the relational layer where engagement actually lives. The manager — equipped with their own 1:1 agenda — has the structured conversation with each direct report that surfaces what is actually going on. None of that flows back to HR as text. The aggregate signal moves; the underlying relationships do the work.

This is the operating model for a function that has stopped trying to catch disengagement by measuring the employee directly and started catching it by measuring whether the employee is still investing in themselves.

The honest separation

Di Stefano on reflection, Amabile on progress visibility, and SDT on autonomous engagement collectively underwrite the theory: the cadence at which employees engage with their own structured workspace is a leading indicator of retention. That is the mechanism.

What the literature does not yet prove is that this specific product, in your specific organization, produces a retention signal that beats your current model. That is what a six-month pilot measures. You see the data from your own teams. You compare against your existing retention model. You make the call on whether the signal is rich enough to act on.

The bet is well-founded. The test is yours.

How Nela Helps

Use Nela to log your wins, track your challenges, and build a private 1:1 agenda from your own evidence for your next conversation. Your data is owner-only at the database — enforced by Postgres Row-Level Security, not just hidden in the UI — and only you can read it back through the app. Request pilot access.

Further reading

• Di Stefano, G., Gino, F., Pisano, G. P., & Staats, B. R. (2014). "Learning by Thinking." Harvard Business School Working Paper 14-093.
• Amabile, T., & Kramer, S. (2011). The Progress Principle. Harvard Business Press.
• Deci, E. L., & Ryan, R. M. (2000). "Self-Determination Theory." American Psychologist.