What Sonar watches for
All 17 signals — and what each one refuses to conclude.
A signal is not a score and never a verdict. It is one sentence saying something changed for one person, measured against how that person normally works — never against a team average, and never against anybody else.
- Signals in all
- 17
- Groups
- 3
Signals in all
Groups
The shape of one
Every signal has four parts.
Here is one of the 17, opened all the way. The fourth part is the one that matters, and it is written into every single signal.
What sets it off
- Online past 11pm six nights running
- Meeting load doubled vs. their normal week
- Response time on Slack creeping up day over day
- Uninterrupted work time collapsing — no gaps left in the calendar
Why it matters
Burnout is the longest-leading signal we track. By the time someone hands in notice, the calendar pattern that predicted it was usually visible six weeks earlier. Catching it at week one means a workload conversation; catching it at week six means a backfill.
How to open the conversation
Don't lead with the pattern. Lead with the person. "How are you holding up?" lands differently than "I noticed your calendar has been packed." Use the data to know who to ask, not what to say.
What it refuses to conclude
Not an assessment of output quality. Not a performance flag. Nothing is sent to HR. Sonar identifies a load pattern — what caused it and whether it's a problem is a question only a conversation can answer.
That fourth part is not a disclaimer added to this page. It ships with the signal, it is on the screen your team sees, and every one of the 17 has one.
The whole vocabulary
Everything it can ever tell you.
Not a sample, and not a category we might add to later — Sonar cannot say anything that is not one of these 17. Open any one for the whole anatomy.
Individuals
Behavior change in one person against their own normal.
11 signals
1OverloadedRunning too hot for too long.
How it shows up in the data
- Online past 11pm six nights running
- Meeting load doubled vs. their normal week
- Response time on Slack creeping up day over day
- Uninterrupted work time collapsing — no gaps left in the calendar
What this signal does not mean
Not an assessment of output quality. Not a performance flag. Nothing is sent to HR. Sonar identifies a load pattern — what caused it and whether it's a problem is a question only a conversation can answer.
Why it matters
Burnout is the longest-leading signal we track. By the time someone hands in notice, the calendar pattern that predicted it was usually visible six weeks earlier. Catching it at week one means a workload conversation; catching it at week six means a backfill.
How to open it
Don't lead with the pattern. Lead with the person. "How are you holding up?" lands differently than "I noticed your calendar has been packed." Use the data to know who to ask, not what to say.
2Weekend work creepRecovery time quietly eroding, weekend by weekend.
How it shows up in the data
- Saturday and Sunday activity climbing three weeks running
- Weekend commits and messages now well above their own normal
- The pattern building gradually, not a one-off crunch week
- Still calm on weekdays — the strain is showing up on the edges first
What this signal does not mean
Not a measure of dedication or output, and never a reason to praise weekend work. Someone may simply prefer a shifted schedule. Sonar sees the rising trend against what's normal for them — whether it's a problem is a conversation, never a verdict, and nothing goes to HR.
Why it matters
This is the earliest burnout tell there is — it shows up weeks before weekday after-hours ever spike. Someone steadily giving up their weekends is spending down a reserve. Catch the creep and it's a workload-and-rest conversation; miss it and it becomes the acute Overloaded signal, then a resignation.
How to open it
Be gentle and specific about the rest, not the work. "I've noticed you've been online the last few weekends — is everything getting the time it needs during the week?" Make it safe to say no.
3No recoveryWeeks on end with no real break.
How it shows up in the data
- Worked every weekend for a month or more, unbroken
- No PTO taken across the whole stretch
- A flat, relentless pace — not a spike, just no let-up
- The kind of run that doesn't trip an overload alarm because nothing is changing
What this signal does not mean
Not a judgment of work ethic or output, and never the opposite — relentless availability is not something to reward. Sonar sees an unbroken stretch with no PTO; the reason and the remedy belong to a conversation, not a record, and nothing goes to HR.
Why it matters
Burnout isn't only about intensity — it's about the absence of recovery. Someone holding a steady but unbroken pace with no day off is depleting a reserve that doesn't refill on its own. This is the pattern an acute-overload or rising-trend signal will miss precisely because it's flat. The fix is often simple and overdue: make sure they actually take time off.
How to open it
Lead with permission, not concern. "You've been going hard for a while with no real break — I'd like you to take some time. What would make that possible?" Then help clear the path.
4Meeting overloadTheir calendar is eating the week.
How it shows up in the data
- Two weeks or more of meeting time well above their own normal
- More than twenty hours a week sitting in meetings
- Real work pushed to the edges of the day or to nights
- A sustained pattern, not one busy planning or offsite week
What this signal does not mean
Sonar reads calendar event times and lengths — never what was discussed, never who called the meeting, never whether it was worth holding. Lots of meetings can be exactly right for a given week. Whether the load is a problem is a conversation, never a verdict, and nothing goes to HR.
Why it matters
An engineer who spends half their week in meetings has no stretch of quiet time long enough to do the work they're measured on. The strain hides for a while — they make it up after hours — until the after-hours catch up with them too. This one has the most concrete fix of any burnout signal: it's a lever you can actually pull. Decline a recurring invite, hand off a standing meeting, protect two mornings a week.
How to open it
Make it about their focus time, not their attendance. "Your calendar's been packed for a few weeks — which of these meetings actually need you, and which could we clear so you get real heads-down time back?" Then help cut them.
5Checking outPulling away over a longer window.
How it shows up in the data
- Message volume sharply down and staying down for about a week straight
- Talking to fewer distinct people across Slack and GitHub, not just less overall
- 1:1 attendance rate dropping — skipping or shortening standing check-ins
What this signal does not mean
Not evidence of poor attitude or disengagement from the job. A drop in communication pattern could mean a personal situation, a focused sprint, or dissatisfaction — Sonar can't tell which. No flag goes on their record.
Why it matters
Different from a bad week. Checking out is a sustained shift — the person is still on the team but the connection is gone. It often precedes a quiet job search by 30–60 days.
How to open it
An honest 1:1 — not a performance conversation. Ask what they're excited about right now. The answer (or the silence) is the data.
6Gone quietOut of the pattern, suddenly.
How it shows up in the data
- No Slack messages, PR activity, or calendar engagement for 10+ business days
- Or three 1:1s missed in a row
- No PTO marked for that stretch
What this signal does not mean
Could be illness, focused work, a personal matter, or just a quieter week. Sonar flags the absence against a fixed threshold — it does not assume why, and it doesn't need a history of this person's normal to fire. No automatic escalation happens.
Why it matters
Sometimes it's a sick kid. Sometimes it's a deadline focus mode. Sometimes it's something they want to talk about but haven't found the moment. Ten business days of silence with no PTO on record is long enough to be worth a check-in, whatever the reason turns out to be.
How to open it
Direct check-in, not a Slack message. Five minutes, in person or on a call. "Hey — haven't seen you around this week, everything OK?"
7StalledWork that's been sitting for too long.
How it shows up in the data
- PR sitting open for over a week with no review activity
- Reviewer hasn't replied; author has moved on to other work
- Linear / Jira ticket in the same status for far too long
- Comments thread has gone cold mid-discussion
What this signal does not mean
Not a reflection on the author or the reviewer. Stalled work is almost always a queue or priority problem, not a person one. No individual gets flagged as underperforming.
Why it matters
This is the cheapest signal to act on. A blocked PR is rarely a person problem — it's a queue problem. Five minutes of unblocking saves a week of frustration on both sides.
How to open it
Don't ask the author. Go to the reviewer: "Anything I can do to help you get to PR #482?" Most of the time the answer is "oh — I forgot."
8Stalled ticketA high-priority ticket assigned to them has been sitting untouched for over a week.
How it shows up in the data
- Linear issue in 'In Progress' for more than seven days with no activity
- Urgent or high-priority issue — not a backburnered nice-to-have
- No comments, sub-issue updates, or field changes in that window
- The person has other active work — this one just stopped moving
What this signal does not mean
Not evidence that the person is underperforming or avoiding the work. A stalled ticket almost always means a hidden dependency, a context switch, or a priority that shifted but wasn't updated. No individual is flagged as underperforming.
Why it matters
A high-priority ticket that's been In Progress for a week with no activity is either blocked and the person hasn't said so, or it slipped through the cracks. Either way, a manager asking one question often unblocks it in minutes. Left alone, it becomes the thing that delays the sprint review.
How to open it
Come in curious, not corrective. "I noticed the auth-service ticket has been In Progress for a while — is anything blocking it, or has it shifted priority?" Most of the time the answer reveals a dependency nobody had named.
9Slow feedback loopEvery time they ship, they wait days for a look.
How it shows up in the data
- Their pull requests routinely sit a couple of days before anyone reviews
- The wait shows up again and again, across many PRs — not one unlucky week
- Some PRs eventually merge fine; the problem is how long the first look takes
- Counting only weekdays, so a Friday PR reviewed Monday isn't held against anyone
What this signal does not mean
Sonar reads when PRs were opened and when the first review landed — never the code, never the review comments, never quality. Slow feedback has many causes: a thin reviewer bench, a hard change, a crunch week. Whether it's a problem is a conversation, never a verdict, and nothing goes to HR.
Why it matters
Where Stalled is about one PR stuck right now, this is the pattern underneath: someone who waits days for feedback every single time. It's one of the most corrosive things in a developer's week — you've moved on by the time the review lands, and the lag quietly says "your work isn't a priority." It's also one of the most fixable: give them a reviewer buddy, set a turnaround norm, balance the review load.
How to open it
Treat it as a pipeline problem, not a people one. "Your PRs are waiting too long for a first look — let's fix who reviews your work and how fast." The person waiting is rarely the one who can fix it.
10Too much in flightLots started, nothing landing.
How it shows up in the data
- Five or more pull requests open at the same time, held for days
- New work opened before the older work closes
- The pile staying high all week, not a one-day batch they cleared
- Branches going stale while attention is split across all of them
What this signal does not mean
Sonar counts how many PRs were open at the same time from open and merge times — never the diffs, never the content. A high pile can be exactly right during a big migration. Whether it's a problem is a conversation, never a verdict, and nothing goes to HR.
Why it matters
Every open PR is a thread someone is still holding — rebasing to stay current, tracking review comments, fending off merge conflicts. When the pile stays high, nothing reaches "done," and the more they juggle the slower each one moves. It's demoralizing and it's risky — stale branches and big simultaneous merges. The fix is coaching, not pressure: help them land work in order before opening more.
How to open it
Help them sequence, don't tell them to hurry. "You've got a lot open at once — which two should we get over the line first, and what's blocking those?" Clear the oldest before starting the next.
11RisingWorth saying out loud.
How it shows up in the data
- Code-review volume up sharply this sprint — well above their own normal
- Collaborating with a wider circle of teammates than usual
- Message activity up while staying steady, not spiky — a sustained shift, not a one-off busy week
What this signal does not mean
Not a promotion recommendation or a compensation trigger. Sonar shows a pattern worth recognizing — it doesn't make any organizational decision for you.
Why it matters
Recognition that's specific lands; recognition that's generic doesn't. The biggest reason quiet stars leave is they think nobody noticed. Sonar's job here is to make sure you noticed before they did.
How to open it
"I noticed you've been reviewing a lot more of the team's PRs this sprint, and helping more people than usual — that's made a real difference." Specific, public, this week — not at next quarter's review.
Team & process
Pair-, codebase-, and review-level patterns the team's own metrics miss.
4 signals
1Collaboration riftTwo people who used to work closely have gone to zero interaction.
How it shows up in the data
- Two engineers who cross-reviewed each other weekly now have zero interactions for 21+ days
- DM volume between two people dropped to zero after being daily
- One person consistently absent from threads the other initiates
- Collaborative commit patterns between the two have stopped entirely
What this signal does not mean
Not proof of a conflict or a relationship problem. Two people's collaboration patterns may have naturally diverged as work changed. Sonar identifies the absence; only a conversation determines what it means.
Why it matters
Relationship problems between engineers rarely come up in 1:1s. People don't volunteer that they've stopped talking to a colleague — they just quietly stop collaborating. Sonar detects the absence before you notice the slowdown in delivery.
How to open it
Don't bring up the other person by name. Start with context: "I've noticed your work on X has been more independent lately — is there anything blocking the collaboration there?"
2Invisible carrierReviewing everyone's PRs, answering every blocker — while their own output looks flat.
How it shows up in the data
- Reviewing far more PRs than they're opening themselves — three times as many, sustained over two weeks
- Answering first-responder questions in shared channels well above their own normal weekly rate
- Own commit output flat or below their normal, while the review and unblocking load stays high
What this signal does not mean
Not a performance problem. This person's own output may look flat because they're spending time unblocking others — that's a contribution, not a deficiency. Nothing here implies underperformance.
Why it matters
The invisible carrier is the person your team's velocity depends on — and the person most likely to burn out without anyone noticing because their own output metrics look flat. They're not underperforming. They're paying a hidden tax that doesn't show up in any dashboard.
How to open it
Acknowledge the contribution before anything else. "I've been looking at how the team has been unblocking each other — you're carrying a lot of it. I want to make sure that's recognized, and also make sure it's sustainable for you."
3Bus factorOne person owns the majority of a critical service. If they leave, production is at risk.
How it shows up in the data
- One engineer holds the overwhelming majority of commits to a critical service
- No other team member has reviewed or touched that code in months
- No documentation commits from anyone other than the sole owner
- That engineer is also showing Overloaded or Checking Out signals
What this signal does not mean
Not a criticism of the codebase owner. Knowledge concentration usually results from good work over time, not gatekeeping. Sonar flags it as an org risk, not an individual failing.
Why it matters
Bus factor isn't a people problem — it's a risk problem. It becomes urgent the moment the owner shows any sign of departure risk. Sonar flags the combination: concentrated ownership plus behavioral warning signals in the same person.
How to open it
Frame it as team resilience, not criticism. "I want to make sure we're spreading knowledge on the payments service — not because anything is wrong, but because I don't want you to be the only person who can work on it."
4Scope driftA branch with 40+ commits and no PR. Something grew beyond its original scope.
How it shows up in the data
- Branch with 40+ commits over 12+ days and no open PR
- Commit messages shifting topic mid-branch — feature to refactor to unrelated fix
- No updates in the associated Linear or Jira ticket for 10+ days
- Author's PR output has dropped to zero while this branch stays open
What this signal does not mean
Not an assumption of wasted effort or poor discipline. A long-running branch often means an engineer doing more than asked. Sonar flags that it hasn't reached a review gate — it makes no judgment about the work inside.
Why it matters
Scope drift is almost always well-intentioned — an engineer who found real problems and kept fixing them. But the longer it runs, the harder it is to review, the harder it is to reverse, and the more likely it lands in production as a single ungovernable change.
How to open it
Come in curious, not corrective. "Hey, I can see you've been deep in that branch for a while — what's it turning into? I want to make sure we get eyes on it before it gets too big to review in one sitting."
You as a manager
A mirror — patterns in how you engage your reports. Private to you.
2 signals
1Attention gapYou respond 3× faster to some reports than others. You probably don't know.
How it shows up in the data
- You reply to one report in minutes, to another in hours — and the gap has been steady for weeks
- 1:1 duration with one report averages 48 minutes; with another, 18 minutes
- One report initiates almost all their interactions with you; another never has to
- Meeting rescheduling rate significantly higher for some reports than others
What this signal does not mean
Never shared with your reports. Private to you only. Not a score — a mirror. No one at Sonar reviews it. No one else in your org sees it.
Why it matters
This is usually unconscious. Affinity bias in manager attention is one of the most consistent patterns in employee attrition research — people who feel less visible to their manager start looking elsewhere. Sonar makes the pattern visible so you can correct it before someone notices it for you.
How to open it
This signal is for you, not a conversation with your report. Use it as a self-audit. Review the pattern, adjust deliberately, and notice whether the gap closes over the next 30 days.
2Credit gapHigh output, no recognition. Three weeks. They're starting to feel invisible.
How it shows up in the data
- High output, high PR throughput — zero @mentions or public acknowledgment for weeks
- Not recognized in sprint retros, standups, or shared channels despite measurable output
- Previously recognized regularly — recognition dropped off sharply
- Rising signal active (output up) but no corresponding acknowledgment from anyone
What this signal does not mean
This person didn't file a complaint. Sonar read their output pattern, not their feelings. They may not consciously feel overlooked yet — the signal exists so you act before they do.
Why it matters
The quietest exits are often the most preventable. High-output engineers who feel invisible don't complain — they update their LinkedIn. Specific, timely recognition is the cheapest retention tool available to a manager, and it's the one most likely to be delayed until a quarterly review when it's already too late.
How to open it
Public, specific, this week — not at next quarter's review. Name the work. "Devon shipped the data pipeline rewrite solo and kept the team unblocked through the whole sprint. That's not invisible to me." Say it in the channel where the work happened.
What Sonar can see
It sees when work happened. Not what was said.
Not a filter applied after the fact. The words never arrive in the first place, so there is no later step that could read them.
What arrives
- When something happened — the timestamp, and nothing else about it
- Who talked to whom, by account, in which channel or repository
- That a pull request was opened, reviewed, merged — never what changed
- That a meeting happened, how long it ran, how many people were in it
What never arrives
- Message text, in any channel, in any direct message
- Code, or any diff, or a commit message
- Pull request titles, descriptions and comments
- Meeting titles, agendas, invitations and notes
- Documents, pages, tickets — the contents of anything
One exception, and it is a real one
Sonar has a meeting notetaker. Switch it on and it writes down what was said in that meeting, which is genuinely the words — the only part of the product that ever reads any. It is off until someone turns it on, it asks every meeting, and anyone in the room can say no. We would rather tell you here than have you find it later.
Everything else on this page holds for every tool Sonar connects to today and every one it connects to next, because it is a property of how the data arrives rather than a rule applied per tool. What we hold, and how to get it back →
Ready when you are
17 signals. One short email. Weekday mornings.
17 signals, each one readable by the person it is about. Free for up to four people, no card.
No card · Cancel any time