Team Pulse · illustrative sample

Team Atlas — payments platform

Last six sprints · 11 engineers · 9 of 11 responded to the pulse. All names, tickets and numbers here are invented.
A checkup has two halves: what the patient tells you, and what the panels show. Neither is the diagnosis on its own. The useful part is where they disagree.
Delivery predictability
62%
↓ from 84%
Unplanned work
31%
↑ from 12%
Reopen rate vs. org median
3.1×
↑ 6 sprints
Open items per engineer
2.8
↑ from 1.6
Pace felt sustainable
2.4/5
↓ from 3.8

The two instruments

The interview asks the team how it's going. The panels read what actually happened. Both are run before anyone draws a conclusion.

The interview — what the team said9 responses
Do you know what matters most this quarter?
4.1/5
Is the pace sustainable?
2.4/5
Do you have what you need to do your best work?
2.9/5
Are handoffs and ownership clear?
3.3/5
Would you feel comfortable flagging a risk early?
3.9/5
The panels — what the tickets show412 tickets
Points on stated priorities88%
Unplanned work, share of points31%
Cycle time · median → 90th4d → 22d
Ticket age spent blocked38%
Tickets touched by 3+ people41%
Reopened at least once58 of 412

What the ticket text revealed

These themes are not fields in Jira. Nobody labelled them. They come from reading the descriptions, comments and reopen reasons across 412 tickets and grouping what people actually wrote. Click any theme to see the evidence it was built from.

Evidence — what people wrote
PAY-2841“Blocked again — partner sandbox returning 502s since Tuesday. Can't verify the payout path end to end.”
PAY-2903“Waiting on a sandbox refresh from the partner. No ETA given. Parking this.”
PAY-2955“Third time this sprint the test environment dropped our webhook registrations.”
Why it matters: this is the single largest source of blocked time on the board, and it is invisible in every existing report because no field captures it. It reads as “slow team” in a delivery dashboard.
Evidence — what people wrote
PAY-2877“Is this ours or platform's? Moving to the platform board for now.”
PAY-2881“Moved back — platform says integration owns the mapping layer.”
PAY-2914“Nobody owns the retry contract between us and the connector team. Third time this has come up.”
Why it matters: the team scored ownership clarity 3.3/5 — middling, not alarming. The ticket text puts this second by cost. The team is under-reporting the thing costing it the most time.
Evidence — what people wrote
PAY-2790“Reopening — retries still duplicating on 5xx responses.”
PAY-2836“Same as the last two. We keep patching the symptom at the edge.”
Why it matters: these were filed as nine separate bugs across four sprints. Grouped, they are one design problem worth fixing once. This is the clearest example of AI reading text finding something that counting tickets cannot.
Evidence — what the pattern shows
Pattern28% of merged changes pass through one review queue, which is also the slowest step in the 90th-percentile cycle time.
Reported as a system risk, deliberately not as a person. The finding is that the process has a single point of failure, not that anyone is slow. Naming an individual here would turn a useful diagnostic into a performance review.

Where the story and the data disagree

This is the diagnosis. Agreement confirms; disagreement is where the work is.

The team said
“We know what matters most this quarter.” — 4.1/5
The tickets show
88% of points map to the stated priorities.
✓ Agreed — priorities are not the problem. Stop looking here.
The team said
“The pace isn't sustainable.” — 2.4/5, down from 3.8
The tickets show
Unplanned work at 31%, open items per engineer up to 2.8.
✓ Agreed — the strain is real and it has a measurable cause.
The team said
Ownership clarity is middling — 3.3/5, their fourth-ranked concern.
The tickets show
Boundary confusion is the second most expensive theme on the board, at roughly 14 days.
⚠ Gap — the team is under-reporting its second-biggest cost. Worth asking why in the next session.

So what

Three moves, ordered by what the evidence supports.

1
Escalate the partner sandbox as a dependency, not a bug
The largest single source of blocked time, and invisible in every current report. This needs an owner outside the team and a written SLA.
2
Name one owner for the integration boundary
Cheapest fix on the list. It costs a decision, not headcount, and it removes roughly 14 days of ping-pong per six sprints.
3
Treat the nine webhook reopens as one design fix
Fix the retry contract once instead of patching the edge a tenth time. Then re-measure reopen rate in two sprints.
How this is scoped, on purpose. Everything here is team- and system-level. No individual is named, ranked, or scored, and the pulse responses are reported only in aggregate. This is an early-warning tool meant to protect delivery and the people doing it. The moment it becomes a way to evaluate a person, teams stop answering honestly and the instrument stops working.