Why did our cycle time jump this month?
Ran cycle time breakdown on the Engineering health dashboard, Aug 30 to Sep 29
Cycle time is up 39% this month, and most of the increase is in review. PRs spend 26.6 hours there on average against a 5.9 hour median, so most reviews clear quickly while a subset sits stuck.
Cycle time breakdown
Pickup is next, at 12.4 hours on average before a first review. Want me to show which reviewers these PRs are waiting on?
Kraken cut review time by more than half in four months, from 2 days 15 hours to 1 day 3 hours at the 90th percentile.
Size and complexity accounted for maybe 5 to 10% of the slowdown. 70 to 80% came from time zone differences, team silos, and reviewer availability.
Nik Sudan, Engineering Operations Lead, Kraken
Trusted by the teams transforming software delivery
A team lead asks about this week's PRs, and a VP of engineering asks about the quarter. LinearB answers both from the same data, limited to the teams each person can already see.
Platform team · Sep 22 to 28
Which of our PRs are stuck in review?
LinearB AI assistant
#9185 add read replica for orders-db
Claude Code agent · no reviewer
#9171 retry idempotency keys
Claude Code agent · no reviewer
#9168 bump grpc to 1.66
Dana K. · reviewer out until Thursday
Three PRs have waited more than two days, and two of them have no reviewer assigned.
All teams · Aug 30 to Sep 29
Which team's cycle time moved most this month?
LinearB AI assistant
Cycle time per team
Payments moved most, up 18% to 31.2 hours, and the increase sits in review. Platform improved most, down 11%.
Whole organization · Q3
Is AI paying off this quarter?
LinearB AI assistant
Cost per effective PR
PRs merged rose 28% while AI spend held flat, so cost per effective PR fell about 15%, to $330.43.
The LinearB AI assistant names the widget, dashboard, date range and teams behind every part of an answer, so you can open the same view and check it yourself.
Answers come from the same widgets and metric definitions that power your dashboards.
Admins, editors and viewers get answers about the teams they can already access, and individual metrics follow your organization's privacy setting.
Cycle time breakdown
Cycle time is up 39% this month, and most of the increase is in review. PRs spend 26.6 hours there on average against a 5.9 hour median, so most reviews clear quickly while a subset sits stuck.
Want me to show which reviewers these PRs are waiting on?
Describe the metric you need in Agentic Studio. LinearB writes the query, charts the result, and saves it as a widget on a dashboard your whole team sees.
Every widget comes with its SQL, so you can edit the query, widget type and axes before you save.
SELECT toStartOfWeek(merged_at) AS week, avgIf(review_hours, author_type = 'agent') AS agent_review_h, avgIf(review_hours, author_type = 'human') AS human_review_h FROM pull_requests WHERE merged_at BETWEEN {start_date:Date} AND {end_date:Date} GROUP BY week ORDER BY week
Admins, editors and viewers. LinearB scopes every answer to the teams that person can already see, so a team lead asking about delivery gets answers about their own teams.
From the same widgets and metric definitions behind your LinearB dashboards. Each answer names the widget, dashboard and date range it used, so you can open the same view and compare.
Answers stay in the chat, and you can copy them. To keep a metric on a dashboard, switch to Builder, describe the widget, and save it to Studio.
Yes. The Agentic Studio shows the SQL behind every widget. You can edit it, run it again, and change the widget type and axes before you save.
Yes. If your organization has turned off individual-level metrics, the assistant doesn't return individual-level data.
LinearB grounds every answer in your LinearB data and names the widget it ran. As with any AI-generated answer, check important numbers against the dashboard before you act on them.
