Software Testing

Beginner path · lesson 25 of 35

BeginnerFundamentals4 min readPublished November 12, 2026

What is regression testing? Regression testing vs retesting

After a change, two questions matter: is the bug really fixed, and did the change break anything else? Retesting answers the first. Regression testing answers the second.

On this page

Key takeaways

  • Retesting (also called confirmation testing) checks that a specific bug is fixed, by running the steps that found it.
  • Regression testing checks that a change didn't break something that used to work.
  • Every change needs both. Regression suites run again and again, so they're the first candidates for automation.

Two questions after every change

A developer fixes a bug and hands it back. You now have two questions:

  1. Is the bug really fixed? That's retesting.
  2. Did the fix break anything else? That's regression testing.

Beginners often do the first and forget the second. But software is connected. A change in one place can break something in a place that looks unrelated.

Retesting (confirmation testing)

Retesting means running the exact steps that revealed the bug, on the new build, to confirm it's gone.

  • Use the same data, environment and steps as in the bug report.
  • If it passes, close the bug. If it fails, reopen it with details.
  • It's planned per bug, so you know exactly what to run.

Regression testing

A regression is when something that worked before stops working after a change. Regression testing looks for those.

Changes that can cause regressions:

  • Bug fixes and new features.
  • Library, framework or operating system updates.
  • Configuration and infrastructure changes.
  • Database changes and data migrations.

A regression suite is the set of tests you run to check that existing behaviour still works. It grows over time and is run again and again, which is why teams automate it.

Side by side

RetestingRegression testing
GoalConfirm a specific bug is fixedConfirm nothing else broke
Which testsThe steps from that bug reportTests of existing, previously working features
WhenAfter each fixAfter any change: fixes, features, updates
Expected outcomeThe test that failed now passesTests that passed still pass
AutomationOften manualStrong candidate for automation

When a test isn't repeated

Real-world example · NASA Mars Polar Lander, 1999A test that wasn't run again after a fix

Mars Polar Lander was lost during its landing attempt on 3 December 1999. The investigation's most probable cause: when the landing legs unfolded, sensors in them gave a brief false signal that the lander had touched down.

The software accepted the signal and would have shut down the descent engine too early, roughly 40 metres above the surface.

During system testing, the touchdown sensors had been wired incorrectly, so the test couldn't reveal the problem. The wiring was corrected, but the test was not repeated afterwards.

Lesson: when something changes, tests that depend on it must run again. A fix to the test setup counts as a change too.

How to choose what to regression-test

You rarely have time to run everything. Choose based on risk:

  1. Start near the change. Test the features that share code, data or screens with what changed. Ask the developer what they touched.
  2. Add the critical paths. Login, checkout, payments: the flows that must always work, after any change.
  3. Add the areas that broke before. Bugs tend to cluster in the same places.
  4. Automate what repeats. If you run the same regression check every release, it should probably be an automated test.

Do

After retesting a fix, spend a few minutes testing the area around it: related screens, related data, the same feature for another user role.

Don't

Close the bug the moment the original steps pass. The fix may have moved the problem somewhere else.

Check yourself

A developer fixes a bug where discount codes weren't applied to sale items. What do you retest, and what would you regression-test?

Show answer

Retest: apply a discount code to a sale item, using the original steps, and confirm the discount is applied. Regression-test: discount codes on regular items, orders with no code, expired codes, totals and taxes in checkout, and invoices, since all of them depend on how prices are calculated.