← All features
Commit effort

Every commit summarized and rated 1 to 10.

A label fix should not weigh as much as a new feature. Each commit gets a one-line summary and an effort score for the real work involved. A person’s effort is the sum of their scores.

Commits · Kofi Oseieffort 4.2 / 10 avg
9/10Queue incoming GitHub events with retriesacme/api-gateway · +612 −248 · 2h ago
8/10Customers can export their invoices as PDFacme/billing · +412 −38 · yesterday
5/10Declined payments are retried automaticallyacme/api · +90 −13 · Oct 6
2/10Fix the sign-up button alignment on mobileacme/web · +6 −4 · Oct 6
1/10Fix the label on the settings pageacme/web · +2 −2 · Oct 5
1Scored on work, not sizeGenerated files, lockfiles and reformatting count little. Logic, migrations and new behaviour count a lot.
2Effort is the sumThree fixes rated 1 make 3 points. One feature rated 9 makes 9. Rankings follow the work, not the commit count.
3Visible everywhereScores appear on the timeline, the member profile, the reports and the Who ships ranking.
1 to 10 scale1 is a typo or a label. 10 is exceptional: a new subsystem, a complex migration.
No code keptOnly the summary and the score are stored. The diff is discarded after reading.
Not a gradeThe score describes a commit. It is never turned into a rating of a person.
In depthWhy counting commits is wrong, and what we do instead5 min read

Commit counts measure habits, not work

Some developers commit every ten minutes, others once a day. Counting commits rewards the habit, not the work. Lines changed are no better: a generated lockfile or a reformat can add thousands of lines without any thought behind them.

Commit effort replaces the count with a reading. Each commit is summarized in one or two lines and scored from 1 to 10 for the work it represents: a typo is a 1, a configuration change a 2 or 3, a new endpoint with tests a 7, a feature touching several layers an 8 or 9. A 10 is exceptional.

Effort is the sum

A person’s effort over a period is the sum of their scores. Three label fixes rated 1 make 3 points; one feature rated 9 makes 9. The ranking on the dashboard and in reports follows the work, not the number of commits. Generated files, vendored dependencies and pure reformatting are recognised and score low whatever their size.

How the score is produced

When a commit lands on a tracked branch, its diff is loaded in memory, read by Squadlog’s own model, summarized and scored. The diff is then discarded; only the line and the score are kept. The same model and the same scale apply to every repository, so scores are comparable across a team.

What the score is not

It is not a rating of a developer. It describes a commit. It is shown next to the commit, averaged per person for context, and summed for the ranking, but there is no threshold on people, no alert based on a low score and no automated decision of any kind.

Who it is for

For teams whose work does not fit a PR count

Commit-first teamsMeasure what ships to mainNo pull requests, no problem: effort is computed from commits, and dashboards lead with it.
Managers of mixed seniorityStop comparing output volumesA senior who lands one hard change a week is visible next to a junior who lands ten small ones.
Teams with heavy generated codeIgnore the noiseLockfiles, snapshots and migrations score low; the human work stands out.
Questions

About commit effort

All answers →
  • The commit content is read once and rated 1 to 10 based on the real work involved: scope, logic, tests, risk. Generated files and reformatting count little. Only the summary and the score are kept.
Other features