What is regression testing, and when should you run it?
Regression testing checks that new changes have not broken what already worked. Here is what it covers, how to choose what to retest, and when to automate it.

A developer fixes a small bug on the profile page. The fix is three lines. It is tested, it works, and it goes live. The next morning, nobody can reset their password.
The two things seem unrelated, but they shared a piece of code that the fix changed. Nobody checked password reset because nobody had touched password reset.
Regression testing exists to catch exactly this.
What regression testing means
A regression is something that used to work and no longer does. Regression testing is rechecking existing features after a change, to make sure the change did not break them.
It answers a different question from ordinary testing. Testing a new feature asks whether the new thing works. Regression testing asks whether everything else still does.
Why changes break unrelated features
Software is more connected than it looks from the outside.
- Shared code. One function formats dates for the invoice, the dashboard and the email. Change it for one and you change it for all three.
- Shared data. A new field added for one screen changes what another screen receives.
- Updates to libraries. A routine update to a third-party component alters behaviour somewhere you were not looking.
- Fixes that undo fixes. A change restores an old bug that was fixed months ago, because nobody remembered why the code was written that way.
The larger and older the product, the more of these connections there are, and the less anyone can keep them all in their head.
What to include
You cannot retest everything after every change, so the skill is in choosing. Build the list from four sources.
- The things that must never break. Sign-in, payment, and the main action your product exists for.
- Whatever is near the change. Features that use the same screens, data or code as the thing that was modified.
- Past bugs. Every bug that reached customers earns a permanent check, so it cannot return unnoticed.
- The most-used features. What most customers do on most days.
A small product might have twenty checks on this list. A mature one might have several hundred.
When to run it
- Before every release, at least for the "must never break" group.
- After a bug fix, around the area that was fixed.
- After updating libraries, servers or third-party services.
- After a large change such as a redesign or a new payment provider. Run the full list.
Manual or automated
Regression testing is the strongest case for automation in all of QA, because it is the same checks repeated.
A person running the same sixty checks every week gets tired, and skips the ones that have always passed. Those are the ones that fail next. A script runs all sixty every time, in minutes.
Automating has costs too:
- Writing the tests takes time up front.
- Tests break when screens change, and someone has to maintain them.
- A test only checks what it was told to. It will not notice that a page looks wrong.
The usual answer is to automate the stable, critical paths and keep a person for the rest. Our comparison of manual vs automated testing goes into where the line falls, and why we use Playwright explains the tool.
Regression testing compared with retesting
The two are often confused.
| Retesting | Regression testing | |
|---|---|---|
| What it checks | The specific bug that was fixed | Everything around and beyond the fix |
| Question | Is the bug gone? | Did the fix break anything else? |
| Scope | Narrow | Wide |
You need both. Retesting confirms the fix. Regression testing confirms it was safe.
Signs you need more of it
- Bugs that were fixed come back.
- Each release brings a complaint about something nobody changed.
- The team is nervous about releasing on a Friday.
- Changes are being written faster than they can be checked, which is common now that AI writes much of the code.
A sensible way to start
If you have no regression testing today, do not try to build a full suite at once.
Write down the ten actions your business cannot survive without. Check those ten by hand before each release. When that list stops changing, automate it, and add a new check each time a bug gets through.
Within a few months you have a suite built from real failures, which is worth far more than one written from guesses.
We set up and maintain regression suites as part of Playwright automation. If releases have started to feel risky, tell us about your product and we will suggest where to begin.