GitHub connector: datasets and metric definitionsBeta
Bring in pull requests, reviews and default-branch commits for each developer: PRs merged, time to merge and to first review, reviews given, work type, areas of the code, and how much was AI-assisted. Updates daily.
When you connect GitHub, Quiriz imports the datasets below and adds the metric definitions to Company Context, the one place your company’s definitions live. Every answer that uses one of these metrics uses the same definition, in the app, in Slack and in the Excel, Sheets, PowerPoint, Slides, Word and Docs add-ins. An admin can edit any of them.
DatasetsMetric definitionsQuestionsFAQ
Datasets
| Dataset | What it holds |
|---|---|
| [GitHub] Pull requests | One row per pull request merged, opened or closed in the last 13 months (by merge date when merged, else creation date), across the connected repositories. owner is who gets credit (the merger when an AI agent opened it). Lines exclude lockfiles and generated files, and are a size, not a productivity score. AI-assisted counts only AI that marks its work (agent accounts, co-author trailers, tool footers), so it is a minimum. Promotion PRs (a long-lived branch merged into another, such as dev into main) are kept but left out of the metrics. Outside contributors and bots are labeled and left out of the metrics. When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count.40 columnsrepo, pr_number, title, state, author, owner, developer_kind, author_association, team_member, is_promotion, head_branch, base_branch, created_at, merged_at, closed_at, date, month, merged, lines_added, lines_deleted, lines_changed, lines_added_raw, lines_deleted_raw, files_changed, size, work_type, work_type_source, area, top_folder, ai_label, ai_tool, ai_assisted, review_count, reviewed, hours_to_first_review, hours_to_merge, changes_requested, is_revert, reverted_later, commit_count |
| [GitHub] Commits | One row per non-merge commit on each repository's default branch in the last 13 months. A commit from a pull request carries that PR's work type, area and owner; direct_push = 1 means it reached the default branch without a reviewed PR. Lines exclude lockfiles and generated files, and are a size, not a productivity score. AI-assisted counts only AI that marks its work (agent accounts, co-author trailers, tool footers), so it is a minimum. Promotion PRs (a long-lived branch merged into another, such as dev into main) are kept but left out of the metrics. Outside contributors and bots are labeled and left out of the metrics. When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count.23 columnsrepo, sha, committed_at, date, month, developer, login, owner, developer_kind, team_member, pr_number, direct_push, lines_added, lines_deleted, lines_changed, files_changed, work_type, work_type_source, area, top_folder, ai_label, ai_tool, ai_assisted |
| [GitHub] Reviews | One row per submitted review on a pull request in the Pull requests dataset; developer_kind and team_member describe the reviewer. Self-reviews and pending reviews are left out. Lines exclude lockfiles and generated files, and are a size, not a productivity score. AI-assisted counts only AI that marks its work (agent accounts, co-author trailers, tool footers), so it is a minimum. Promotion PRs (a long-lived branch merged into another, such as dev into main) are kept but left out of the metrics. Outside contributors and bots are labeled and left out of the metrics. When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count.12 columnsrepo, pr_number, pr_author, reviewer, developer_kind, team_member, state, submitted_at, date, month, review, hours_after_open |
Metric definitions
Each metric is a total over the period you ask about. A metric built from other metrics, such as a rate, divides the totals for the whole period; it does not average daily rates. Dated by is the date column that decides which period a row belongs to.
[GitHub] Pull requests
| Metric | How it is calculated | Dated by | Notes |
|---|---|---|---|
| GitHub PRs merged | Sum of merged, where state = 'merged' AND developer_kind <> 'bot' AND team_member = 1 AND is_promotion = 0 | merged_at | Pull requests merged, by merge date. Promotion PRs, bots and outside contributors are left out. When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count. |
| GitHub PRs opened | Count of rows, where developer_kind <> 'bot' AND team_member = 1 AND is_promotion = 0 | created_at | Pull requests opened, by creation date, whatever happened to them since. When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count. |
| GitHub lines changed | Sum of lines_changed, where state = 'merged' AND developer_kind <> 'bot' AND team_member = 1 AND is_promotion = 0 | merged_at | Lines added plus deleted in merged PRs, lockfiles and generated files left out. A size, not a productivity score. When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count. |
| GitHub merge hours | Sum of hours_to_merge, where state = 'merged' AND developer_kind <> 'bot' AND team_member = 1 AND is_promotion = 0 | merged_at | Total hours from opening to merging, over merged PRs. Use GitHub average hours to merge for a typical PR. When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count. |
| GitHub average hours to merge | GitHub merge hours ÷ GitHub PRs merged | merged_at | Average hours from opening a PR to merging it. When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count. |
| GitHub reviewed PRs | Sum of reviewed, where developer_kind <> 'bot' AND team_member = 1 AND is_promotion = 0 | created_at | Pull requests that got at least one review from someone other than the author (bots left out), by creation date. When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count. |
| GitHub first review wait hours | Sum of hours_to_first_review, where reviewed = 1 AND developer_kind <> 'bot' AND team_member = 1 AND is_promotion = 0 | created_at | Total hours PRs waited for their first review (from opening, or from leaving draft), over reviewed PRs. When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count. |
| GitHub average first review wait | GitHub first review wait hours ÷ GitHub reviewed PRs | created_at | Average hours a PR waited for its first review. When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count. |
| GitHub PRs sent back | Sum of changes_requested, where state = 'merged' AND developer_kind <> 'bot' AND team_member = 1 AND is_promotion = 0 | merged_at | Merged PRs that had at least one review asking for changes. When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count. |
| GitHub rework rate | GitHub PRs sent back ÷ GitHub PRs merged | merged_at | Share of merged PRs that were sent back for changes, as a decimal (0.2 = 20%). When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count. |
| GitHub fix PRs | Sum of merged, where state = 'merged' AND work_type = 'fix' AND developer_kind <> 'bot' AND team_member = 1 AND is_promotion = 0 | merged_at | Merged PRs whose work type is fix (from labels, the title prefix, or the AI labeler). When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count. |
| GitHub fix share | GitHub fix PRs ÷ GitHub PRs merged | merged_at | Share of merged PRs that were fixes, as a decimal. When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count. |
| GitHub AI assisted PRs | Sum of ai_assisted, where state = 'merged' AND developer_kind <> 'bot' AND team_member = 1 AND is_promotion = 0 | merged_at | Merged PRs made by an AI agent or with an AI co-author marker. This is a minimum: only AI that marks its work is counted. When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count. |
| GitHub AI share | GitHub AI assisted PRs ÷ GitHub PRs merged | merged_at | Share of merged PRs that were AI-assisted, as a decimal. This is a minimum: only AI that marks its work is counted. When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count. |
[GitHub] Commits
| Metric | How it is calculated | Dated by | Notes |
|---|---|---|---|
| GitHub direct pushes | Sum of direct_push, where developer_kind <> 'bot' AND team_member = 1 | committed_at | Commits that reached the default branch without a reviewed PR (pushed straight to it, or only through a promotion PR). When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count. |
[GitHub] Reviews
| Metric | How it is calculated | Dated by | Notes |
|---|---|---|---|
| GitHub reviews given | Sum of review, where developer_kind <> 'bot' AND team_member = 1 | submitted_at | Reviews submitted, by the reviewer (approvals, change requests and comments); self-reviews left out. When ranking developers, leave out anyone with fewer than 3 merged PRs in the period and break ties by PR count. |
Questions to ask
- Rank developers by PRs merged last month
- Average hours from opening a PR to merging it, by developer, this quarter
- Who waited longest for a first review last month?
- Who gave the most reviews last month?
- Work mix (feature, fix, refactor, tests, docs, dependencies) by developer this quarter
- Which areas of the code changed most this month, and who changed them?
- What share of each developer's work was AI-assisted last month?
Related guides
Ask GitHub in plain words
Try the questions on sample data with no signup, then connect GitHub read-only and ask your own.
Try Quiriz free →Frequently asked questions
Does Quiriz change anything in GitHub?
How often does the GitHub data refresh?
Can I change how a metric is calculated?
Can I combine GitHub with other sources?
All connectors: Quiriz connectors. Definitions on this page are generated from the product’s own connector code.