You pressed publish and something broke. The fastest fix is often to put the old version back. That's a rollback.
Everyone knows the comfort of Ctrl+Z: one press and your last change is undone.
A rollback is Ctrl+Z for a whole website or app. When a new release causes trouble, you put the previous working version back live, quickly. Then you fix the problem calmly.
Why it’s so useful
- it's usually the fastest way to stop a problem
- customers get a working product again within minutes
- the team can investigate without pressure
The tricky part
Undoing code is easy. Undoing changes to stored data is not. If the new version already changed customer records, going back can be messy. Good teams plan for this before releasing.
What it isn’t
It isn’t a fix. The problem is still in the new version. A rollback just buys time.
Where the comparison breaks down
Ctrl+Z works on one document. A rollback can affect millions of users at once, so it’s done with care.
How the words fit together
A bad deploy gets a rollback. Then a patch fixes the real problem. Both protect uptime.
Roll back first, investigate second. Getting customers a working version quickly matters more than finding the cause right away.

When you'll hear it
“Let's roll back.” Put the previous version live again, to stop the problem now.
“Can we roll back the database too?” Can we undo the changes to stored data, which is much harder than undoing code?
“We rolled forward instead.” We fixed the problem with a quick new release instead of going back.
“If tonight's release breaks something, who can roll it back? How long would it take?”

Try it yourselfReady to use it? The exact commands are on my other site, devcmds: Undo a commit you already pushed ↗


