Choosing a Git branching workflow

Goal

Pick a branching model that matches release cadence, review practice and team size, and write it down so contributors do not have to guess.

Prerequisites

A shared repository, an agreed integration branch (often main), and a review or CI gate that every change passes before it reaches that branch.

Steps

1. Decide how long changes may live outside the integration branch. Short-lived topic branches (hours to a few days) keep merges small; long-lived branches accumulate conflicts.
2. Decide whether you need long-running branches beyond main. Pro Git describes a stability ladder (for example main for releases, develop for integration) that only pays off when releases must be supported separately from ongoing work.
3. Decide how a change enters the integration branch: fast-forward merge, merge commit, squash or rebase. Choose one and document it; mixed styles make history hard to read.
4. Decide how releases are cut: a tag on main, or a release branch that receives only fixes. Release branches are justified when a version must be maintained for a long time.
5. Write the decisions into the contributor guide with the exact commands, and enforce what tooling can enforce (protected branches, required checks).

Expected result

Every contributor can answer three questions without asking: where do I branch from, how long may my branch live, and how does it get back. Merges stay small enough to review in one sitting.

Limits and test basis

This is a decision procedure, not a benchmark. Teams that ship continuously usually need fewer branches than the classic model; teams maintaining several supported versions need more. Revisit the choice when release cadence or team size changes.