Where the time went
Every ticket that overran its estimate or was flagged blocked, classified by cause and attributed to who could have prevented it.
Most frequent blocking reasons count · quarter
Classified from ticket comments, Slack threads and PR state. Click a bar to filter the impediment list.
Attribution days lost
Who was in a position to prevent the delay.
Days lost per sprint, by attribution
Partner sandbox outages drove the spike in sprint 14; internal causes are trending down since the environment work shipped.
Estimate accuracy estimated vs actual
Each dot is a ticket. Above the line means it overran. Hover for the reason.
Impediments
Every blocked or overrun ticket this quarter, with the evidence it was classified from.
Impediments
| Ticket | Team | Reason | Attribution | Source | Est → Actual | Lost |
|---|
Preventable, within our control
Internal causes grouped into the change that removes them, ranked by the days they cost. This is the list the platform exists to produce.
Recommended changes ranked by impact
Internal causes, by sub-reason
Days lost this quarter.
What changed after the lower-environment work
Sprints 10–12 vs 15–17, internal causes only.
Connected systems
The evidence trail is only as good as its inputs. The platform reads three systems and requires a reason on every blocked or overrun ticket.
Jira
Tickets, estimates, status transitions, and the comment left when a ticket is flagged blocked or misses its window.
Rule enforced
Blocked or overrun → a structured reason comment is required, naming whether the cause is internal, client, partner or external.
Slack
Delivery automation channel plus the shared client channel: blocked notifications, deadline changes, competing priorities, outage announcements.
Signals captured
Deadline moved · priority swap · partner sandbox status · production incident
GitHub
Pull request state and review latency. A PR waiting on review past threshold becomes an impediment on the linked ticket.
Threshold
Review requested and idle > 2 business days
Classification
Each impediment is classified by an LLM pass over the ticket comments and linked Slack thread, then attributed. Low-confidence classifications are queued for a human check.
This quarter
118 impediments · 106 auto-classified · 12 reviewed by hand