
Summary cards
Reviews by bots, including cubic, don’t count as human reviews.
Delivery
The Delivery chart shows merged PRs as bars and median time to merge as a dashed line. Switch between Organization, By team, and By repository. The team and repository views show the five groups with the most merged PRs and combine the rest. A PR whose author is on several teams is split evenly between them.Work output and work mix
Work output estimates how much reviewed work shipped. Each file in a merged PR scores 1, 2, or 4 points by complexity, based on cubic’s review, so PRs that cubic didn’t review don’t count. Switch between Organization, By team, and By engineer. Bots are left out. Work mix shows how merged PRs split by type. The headline, Feature work, is the share of merged PRs that are features.
cubic classifies each PR from its title and description. The type reflects what the PR says it
does, and cubic doesn’t check it against the code.
Review activity
Review activity charts Reviews given over time. Use Group by to switch between Engineer, Team, and Organization. Every review counts, so a PR reviewed twice counts twice. Bots are left out. Next to the chart:- Revert PRs: merged PRs that undo earlier changes. A PR counts as a revert when its title starts with “Revert” and its description names the reverted commit or an earlier merged PR.
- Average review rounds until merge: for each reviewed PR, cubic counts the approvals, change requests, and dismissed reviews from the reviewer who submitted the most, then averages across PRs. A reviewer who requests changes once and then approves makes two rounds.
Contribution and collaboration
The table shows metrics per person or per team. Switch between Members and Teams, filter by name, and select a column header to sort. To export the table as CSV for the selected period and filters, select the download button.Lines of code measure activity, not productivity. A refactor that deletes more than it adds can be
worth more than a feature with thousands of new lines.