I Know Who's Going to Resign Next. Their Calendar Already Told Me.
By the time someone resigns, the decision is months old. The surprising part is how visible that decision is in the data well before the conversation happens. Calendars, focus patterns, and engagement trends quietly forecast departures — if you know what to look for, and you use it to help.
Resignations Are More Predictable Than You Think
People rarely quit on impulse. The decision forms over weeks or months of accumulating disengagement, and that accumulation leaves a trail. Workforce research consistently finds that attrition is preceded by measurable behavioral change — which means it can be anticipated and, often, prevented.
The point of seeing it early is not to manage someone out. It is to intervene while you still can.
The Calendar Tells the Story First
A calendar is a map of where someone is investing their attention. As people disengage, the pattern shifts: fewer optional and cross-team meetings, declining participation in long-term planning, a retreat from initiatives that pay off months out, and a quiet narrowing toward only what is strictly required.
Combined with productivity trends — falling focus time, shrinking scope, less proactive collaboration — these shifts form a recognizable pre-departure signature.
Engagement & Wellbeing — This Week
Weekly trend
Breakdown
▼ After-hours work down 30% after rebalancing workloads.
Illustrative eMonitor dashboard.
The Cluster of Pre-Resignation Signals
No single metric predicts a resignation. The forecast comes from a cluster: declining engagement scores, reduced discretionary effort, withdrawal from future-facing work, and often a late spike in after-hours activity as the person interviews or wraps up. Our guide on retention prediction with monitoring data goes deeper on modeling this.
Read together over time, these patterns give you a window of weeks to act.
Reach People Before They're Gone
eMonitor surfaces the engagement and activity trends that precede resignations, giving you weeks to retain your best people instead of replacing them.
Using This Power Responsibly
Predicting departures is sensitive. Used badly — to pre-emptively sideline or punish someone — it is both unethical and self-fulfilling. Used well, it is one of the most humane applications of workforce data: it lets you reach out before a great employee is gone.
Keep the analysis at the level of trends, act through conversation and support, and never treat a prediction as a verdict. The signal is an invitation to care, not to control.
Turning the Forecast Into Retention
When the data flags a flight risk, the response is simple and human: a genuine conversation, a look at workload and growth, real recognition, and a path forward. Most at-risk employees are not yet committed to leaving; they are waiting to see if anything changes.
The calendar told you early. What you do with that knowledge is what actually keeps people.
Building a Flight-Risk Picture
No single metric predicts a resignation; a cluster does. Combine calendar signals (withdrawal from optional meetings and long-term planning), engagement signals (declining focus, narrowing scope), and workload signals (a late spike in after-hours activity as someone interviews or wraps up). Read together over time, these form a recognizable pre-departure signature.
The aim isn't a precise prediction; it's an early-warning flag that buys time to act. Even a rough signal weeks ahead of a resignation is far more useful than the certainty of an exit interview.
Keep the analysis at the level of trends and aggregates - the goal is to reach people in time, not to surveil them.
The Ethics of Prediction
Predicting departures is powerful and therefore easy to misuse. Used to pre-emptively sideline, withhold opportunities from, or manage out a flagged employee, it's both unethical and self-fulfilling - the prediction causes the outcome. That is the line that must not be crossed.
Used to start a supportive conversation and retain a valued person, it's one of the most humane applications of workforce data. The same signal serves opposite ends depending entirely on intent and governance.
Safeguards: keep predictions out of performance reviews, restrict who sees them, and require that any action taken is supportive. A prediction is an invitation to care, never a verdict.
From Signal to Retention
When the data flags a flight risk, the response is simple and human: a genuine one-to-one about workload, growth, recognition, and what would make staying the better choice. Most at-risk employees haven't fully committed to leaving; they're waiting to see if the situation improves.
Move before the resignation, not after. A counteroffer made under duress rarely holds; a real change made while there's still goodwill often does. The earlier the signal, the cheaper and more durable the fix.
Track whether interventions work, and feed that back into how you read the signals - retention is a practice you improve, not a one-time save.
Which Signals Matter Most
Not all signals carry equal weight. The strongest combination is declining engagement (focus time, scope) plus withdrawal from future-facing work (skipping long-term planning, optional initiatives) plus a late change in working pattern (a spike or drop in after-hours activity). Any one alone is weak; the cluster is meaningful.
Tenure and context sharpen the read. Disengagement in a long-tenured top performer is a louder alarm than the normal ups and downs of a new hire still finding their feet.
Treat the cluster as a probability, not a certainty - enough to prompt a caring conversation, never enough to act against someone.
A Retention Conversation Framework
When a flight-risk signal appears, run a simple, sincere conversation. Acknowledge you value them, ask open questions about workload, growth, and what would make staying the better choice, and listen for the real driver. Then commit to one or two concrete changes and follow through quickly - speed signals that the relationship matters.
Avoid the duress counteroffer. A pay bump offered only once someone's halfway out the door rarely holds; a genuine change made while goodwill remains usually does.
Document what worked. Retention is a skill that improves with feedback, and each saved employee teaches you which interventions land.
A Governance Checklist
Before using prediction at all, set guardrails. Keep flight-risk signals out of performance reviews and compensation decisions. Restrict access to managers who can act supportively. Require that any action taken is supportive - a conversation, a workload change - never pre-emptive sidelining. Audit usage so the capability can't quietly become a tool for managing people out.
Be transparent that engagement trends inform manager support, and keep the analysis at the aggregate-trend level rather than invasive content surveillance.
Governed this way, prediction is a retention tool. Ungoverned, it's a liability - the checklist is what keeps it on the right side of that line.
What the Data Cannot Tell You
Prediction has hard limits, and respecting them keeps it honest. The data can show that someone's engagement is sliding and that the pattern resembles past pre-departure signatures. It cannot tell you why, whether they've truly decided to leave, or what would change their mind. Those answers live in a conversation, not a dashboard.
It also can't account for life outside work - a relocation, a family change, a better-fit opportunity - that no internal signal will ever reveal. Treating a probabilistic flag as a certainty leads to bad, sometimes unfair, decisions.
The right posture is humility: use the signal to reach out and listen, and let the person, not the model, define their situation.
Turning Retention Into a System
One saved employee is luck; a retention system is repeatable. Build the loop: detect flight-risk signals early, prompt a structured supportive conversation, commit to concrete changes, and track whether they worked. Feed the outcomes back so you learn which interventions actually retain people and which don't.
Combine the predictive signal with the fundamentals that prevent risk in the first place - fair pay, sane workloads, growth paths, and capable managers - so you're not relying on last-minute saves.
Governed and measured this way, prediction becomes part of a humane retention practice rather than a surveillance gimmick: it buys time, and the system turns that time into people who stay.
Key Takeaways
- Resignations are preceded by a measurable cluster of signals, not one metric.
- Calendar withdrawal plus engagement decline plus changed hours is the signature.
- Read it as a probability and an early warning, never a certainty.
- Keep flight-risk signals out of reviews and compensation decisions.
- Restrict access and require any action taken to be supportive.
- Pair prediction with fundamentals: pay, workload, growth, good managers.
- Move before the resignation - early change beats a duress counteroffer.
The Bottom Line
By the time someone resigns, the decision is usually months old - and surprisingly visible in the data along the way. Used responsibly, that early signal is one of the most humane applications of workforce analytics: it lets you reach a valued person while there's still time to change their mind.
The guardrails are everything. Prediction used to pre-emptively sideline people is unethical and self-fulfilling; used to prompt a supportive conversation and a real change, it retains them. Keep the analysis at the trend level, govern access, and act only with care.
eMonitor surfaces the engagement and workload trends that precede turnover, giving managers the lead time to retain their best people instead of replacing them - the whole point of seeing it coming.