Your best senior engineer gave notice at 4pm on a Thursday.
Everyone was surprised.
The signal was in her calendar 60 days earlier and nobody was looking.
Six weeks before the resignation, her GitHub commit cadence had already shifted. The week before, she had moved three internal 1:1s and accepted two external coffees. The afternoon she gave notice, her calendar was clear. Each tool reported half a fact. Together, the half-facts named her.
This is what every senior departure looks like in the data. The board pack you defend after the loss describes the departure as a personnel issue. It was a coordination issue your data was telling you about for two months.
What a checked-out engineer looks like on a calendar
Engineering leaders read calendars as availability. Sarah is in three meetings tomorrow, so she has two hours of focus time. True. Also missing the point.
A calendar is behavioral data. It tells you what an engineer is choosing to do, what they are choosing to decline, and what they are choosing to keep open. Two engineers with identical meeting loads have completely different stories the moment you read the cancellations, the declines, the reschedules, and the freshly-blocked focus windows.
When a senior engineer starts checking out, the pattern shows up in their calendar before it shows up anywhere else. Five signals, all visible to anyone reading the data:
Internal coffee chats decline. The casual 30-minute 1:1s with peers stop getting scheduled. The ones already on the calendar get rescheduled, then declined, then quietly removed.
External meetings appear. Coffees with no internal attendee. 45-minute blocks during work hours, attendees marked Optional, no agenda, no recurring pattern. Recruiter conversations look like this.
Focus blocks shrink. Engineers who used to protect 4-hour focus blocks start letting them get split. They stop fighting for them. Then they stop scheduling them at all.
1:1 cadence drops. The weekly 1:1 with their manager gets pushed to bi-weekly without an explicit conversation. The skip-level 1:1 stops happening. Nobody noticed because nobody owned it.
Off-hours commit patterns emerge. GitHub activity shifts from 10am to 6pm to 10pm to 1am. They are not working harder. They are working at the only time interruptions cannot reach them, which is the only time the work is still rewarding.
Any one of those signals on its own is noise. Senior engineers reschedule meetings. Focus blocks get split. People work late sometimes.
All five of them showing up in the same person, over a 30-day window, is not noise. It is a checked-out senior engineer telling you, through their calendar, that they have already made a decision they have not yet announced. The signal exists because the decision is real. The engineer is not hiding it. They have stopped investing in relationships they do not expect to be in three months from now. The calendar is the leading indicator of that withdrawal.
It shows up six to twelve weeks before the resignation Slack.
What the brief surfaces on Monday morning
The Weekly Brief reads this pattern across the team continuously. Cross-tool. Identity-resolved. Confidence-scored. The board-grade form looks like this:
Sarah Chen (Auth team, Senior Engineer, joined Jan 2023) is showing the coordination shape that historically precedes a senior departure.
Calendar: Internal 1:1 frequency dropped from 4.1 per week to 1.2 per week over the past 21 days. Three external coffee-style meetings accepted in the same window. Two focus blocks declined or rescheduled.
GitHub: Commit cadence shifted from 10am to 6pm (last 90 days) to 10pm to 1am over the past 14 days. Review participation on Auth PRs dropped 73 percent.
Slack: Direct-message volume in #auth-team down 41 percent. Public posts down 28 percent. Response time to mentions doubled.
Confidence: 78 percent. Predicted resignation window: 45 to 90 days.
Board-grade implication: The Auth identity migration committed in the Q3 board update has a key-person dependency on Sarah Chen. Replacement at this seniority is a 90 to 120 day hire plus three months of onboarding ramp. Estimated productivity gap if departure lands inside the migration window: 80 thousand dollars. Renegotiating the Q3 commit before the slip lands at the board is materially cheaper than after.
That is what the brief produces. Not a vibe. Not “Sarah seems checked out.” A name. A confidence band. A predicted window. A board-grade dollar figure.
Five mediums of data. One person. One executive output. Cited.
Do not act on Sarah. Act on the role concentration.
Reading the brief and pulling Sarah into a closed-door conversation is the wrong move.
The brief is not telling you Sarah is leaving. The brief is telling you Sarah is showing the pattern that historically precedes someone leaving. Confronting her with that signal collapses the very relationship you would need to retain her, if retention is the goal.
The brief is not for talking to Sarah. The brief is for restructuring the role around the risk.
Three actions a CTO takes in the same week the brief lands.
Move the review queue. Sarah is currently the primary reviewer on 73 percent of Auth PRs. Within a week, Marcus is doing 40 percent of those reviews. Knowledge starts flowing out of Sarah and into the team. This happens regardless of whether Sarah leaves.
Start the documentation conversation. Not “Sarah, write up the migration architecture in case you leave.” Instead: “We are formalizing the Auth migration architecture as an internal doc for the next two hires. Can you lead it?” Sarah writes the playbook for her own succession without knowing she is doing it.
Renegotiate the Q3 board commit. The Auth identity migration that was committed in the Q3 board update gets a 30 to 60 day buffer added. Not publicly. In the CTO’s internal forecasting. The CFO is told. The CEO is told. The board update goes through with the new timeline, before the slip, not after.
None of these actions require Sarah to know what is happening. None require any conversation about the departure. All of them reduce the company’s exposure to the loss whether or not the loss happens.
If Sarah leaves on day 67, the company is two months ahead of where it would have been. If Sarah stays, the company has formalized the playbook on a strategic system and built a second senior engineer in the role. Both outcomes are net positive. Neither requires a conversation that risks accelerating the departure.
The brief tells you what to do, not what to feel.
What this costs you if you do not read it
Most companies find out a senior engineer is leaving when the engineer tells them. That is the structurally most expensive moment to find out, because at that point the engineer has already mentally left, the runway is whatever they choose it to be (usually two weeks), and the conversation about retention happens with someone who has already decided.
A continuous read on calendar density, GitHub cadence, and Slack participation gives you 60 days of advance notice on roughly 70 percent of senior departures. The remaining 30 percent are decisions made under duress, family events, sudden recruiter wins where the calendar pattern never had a chance to form. You will not catch all of them. You do not need to.
Six in ten senior departures are visible in the data 45 to 90 days out. The cost of acting on the signal, in the cases where you are wrong, is that you formalized documentation and reduced single-point-of-failure risk on a strategic system. That is a positive outcome under any condition. The cost of not acting on the signal is that you find out the company has a key-person risk on a board commitment when the key person has already accepted the next offer.
Your board update next quarter will either name this risk or get surprised by it. There is no third option.