I deploy a plain PHP site to shared hosting, and I do it with SFTP. There is no CI pipeline, no build step, no container. For a long time I felt slightly embarrassed about that, as if a "proper" deploy had to involve a YAML file and a green tick somewhere. It does not. What matters is that the right files land in the right place, in the right order, and that I check the live site afterwards. Here is the checklist I actually run through every time I deploy ramrom.io.

1. Upload assets first

CSS, JavaScript, images and icons go up before any page that references them. If a page goes live a minute before its stylesheet does, a visitor in that minute sees a broken page. It only takes one unlucky visitor at the wrong second. Assets first, pages second, always.

2. Upload pages second

Next come the HTML and PHP pages themselves. I upload them in small batches rather than one giant zip, because a giant zip hides mistakes. If something goes wrong halfway through a batch, I know exactly which files were in it. I keep a copy of the previous working version of every file I change, so I can roll back one file instead of the whole site.

3. Cache-bust last

This is the step I used to forget, and forgetting it cost me real confusion. My pages reference scripts with a version query string, like app.js?v=14. If I change the JavaScript but leave the version number alone, visitors' browsers happily serve the old cached copy and my fix looks like it never happened. So the version bump is the very last thing I do, after the new files are in place. New file, then new version number, never the other way round.

4. Check the live site, not the file manager

The file manager saying "upload complete" proves nothing. I open the live site in a browser, in a private window, and I use it the way a visitor would: paste a link, tap download, open a guide page. On my phone too, not just on desktop. More than once I have found a problem this way that the file list never showed — a page that uploaded fine but referenced a file I forgot, or a script that worked on my machine and failed on the server.

5. Check the console and the network tab

While I am on the live site I open the browser console. Any red error is a deploy failure, even if the page looks fine. I also watch the network tab for 404s. The smallest file I ever chased was a missing favicon — a 2,771-byte file — and it was the only 4xx on the whole site, yet it was enough for Google to flag pages. Small files count.

6. Flush the cache if there is one

Some hosts sit behind a CDN or a server cache. After an upload, I clear it. Otherwise the fix is live for me and stale for everyone else, and I spend an hour debugging a problem that does not exist.

7. Write down what changed

Finally, I keep a plain-text note in the project folder: the date, what I changed, and why. Future-me has thanked past-me for this more times than I can count. A deploy note is not documentation for a team. It is a note to one person who will forget.

That is the whole pipeline: assets, pages, version bump, live check, console check, cache flush, note. No YAML, no runner, no green tick. It is boring, and boring is the point — a boring deploy is one I can repeat at midnight without breaking anything, and one I can undo file by file when something does go wrong. If you run a small site on shared hosting, you do not need a fancier system. You need a checklist you actually follow.