The overwritten homepage nobody could explain
A small law firm's website manager logged in one morning to find the homepage reverted to an old version from months earlier, with a staff bio for someone who'd left the firm. Nobody could say who made the change or when, because the site was managed by connecting to the server over FTP and uploading files directly — there was no record of who changed what, no way to see a difference between versions, and no way to undo the mistake except by manually re-uploading the correct files from whatever backup someone happened to have saved locally.
What a git-based deploy actually replaces
In a git-based workflow, the website's files live in a version-controlled repository rather than existing only on the live server. Every change is committed with an author, a timestamp, and a description of what changed. Deploying isn't manually copying files over FTP — it's pushing a commit, which triggers an automated process that copies the updated files to the server. The server is never the single source of truth; the repository is, and the server is just wherever the repository's current state happens to be published.
- Every change has a clear author and timestamp, ending the 'who changed this' guessing game
- Any previous version can be restored exactly, not approximately, because the full history is preserved
- Multiple people can work on different changes without overwriting each other's uploads
- Automated checks can run before a change goes live — catching a broken page before it reaches visitors instead of after
How the automated part works
The mechanism that connects a repository to a live server is typically a continuous-integration and deployment pipeline: when new code is pushed to the designated branch, an automated job runs, optionally executes tests or validation checks, and then transfers the approved files to the hosting server, commonly over secure FTP for shared hosting environments. This removes the manual step of remembering which files changed and uploading exactly those, which is itself a common source of partial, inconsistent deployments where some files update and others don't.
- Set up a git repository containing the site's files, with a clear main branch representing what should be live
- Configure a deploy pipeline that watches that branch and triggers on every push
- Store server credentials as encrypted secrets in the pipeline configuration, never in the repository itself
- Have the pipeline transfer only changed files where possible, and log the outcome of each deploy
- Add a basic smoke test after deploy — a simple check that key pages return a successful response — so a broken deploy is caught immediately rather than discovered by a visitor
If you can't say exactly what changed on your site last week, you don't have a deploy process — you have a series of accidents.
Where this pays off most
The benefit scales with how often a site changes and how many people touch it. A single-page brochure site that's edited twice a year gets less from this than a site with an active blog, seasonal pricing updates, and multiple contributors — which describes most active small businesses across NetWebMedia's service niches, from tourism operators updating seasonal packages to law firms publishing regular content. The underlying discipline — every change tracked, every deploy reversible — is the same infrastructure NetWebMedia uses across its own portfolio of properties, all of which ship through version-controlled deploy pipelines rather than manual file transfers.
For a business evaluating whether its current site management process has this kind of gap, the practical test is simple: try to answer exactly what changed on the site in the last month, and who made each change. If that's hard to answer, the workflow is worth revisiting — see https://netwebmedia.com/services.html for how this fits into ongoing site management.
Does your business show up when AI answers?
ChatGPT, Claude, Perplexity and Google's AI Overviews are already answering the questions your customers ask. The $49 AI Visibility Scan shows you where you're cited, where you're invisible, and the three changes that move you first — a written report in your inbox within 48 hours. If nothing in it is actionable, you don't pay.
Run the $49 AI Visibility Scan →Share this article
Comments
Leave a comment