Why this matters
A result nobody can repeat is a result nobody should trust. If your manager asks in three months why the forecast changed, you need to show exactly which data and code produced each version. Adverts name Git, clean code and maintainable software, and teams that share work depend on it. Reproducible habits also protect you: when something breaks, you can go back to the version that worked.
Git in practice
Git records snapshots of your project. Each snapshot, a commit, has a message saying what changed and why. The everyday loop is short:
git init
git add clean.py README.md
git commit -m "Clean district names before joining"
git log --onelinePush to a free hosted service such as GitHub or GitLab to back up your work and share it. Use branches to try an idea without breaking the main version, and merge it back once it works. Write commit messages a colleague could understand: "Fix duplicate farmers after plot join" beats "update".
A project layout others can follow
farm-yields/
README.md
requirements.txt
data/raw/
data/clean/
notebooks/
src/clean.py
reports/- data/raw holds data exactly as received. Never edit it by hand; every change happens in code.