Log in
← Blog

Smoke, Sanity and Regression Testing: What’s the Difference, and When to Run Each

Three kinds of testing that get mixed up in every team and every interview. What each one checks, how big it should be, and when it runs.

Vihari Panyam

April 23, 2026 · 5 min read

An engineer on a ladder checking an aircraft propeller before flight
Photo: “170306-F-IY918-158” by U.S. Department of Defense Current Photos, Public Domain Mark 1.0

In short

  • A smoke test asks "is this build alive enough to test?" It is small, fast and runs on every build.
  • A sanity test asks "did this specific fix or change work?" It is narrow and runs after a change, before anything bigger.
  • A regression test asks "did anything that used to work stop working?" It is the widest of the three and grows with the product.
On this page

A new build lands on the test environment at 9am. Before lunch, three different people will run three different kinds of tests on it, and at least one of them will call theirs by the wrong name. That’s fine in conversation. It stops being fine when a team plans its week around "we’ll do sanity on Thursday" and half the team pictures a five-minute check while the other half pictures a full day.

The three terms answer three different questions. Once you see the questions, the differences stop being trivia.

01Smoke testing: is this build worth testing at all?

The name comes from hardware: switch the board on and see if smoke comes out. In software, a smoke test checks that the build starts and its most important paths work at all. Can you log in? Does the home page load? Can you reach checkout? Does the API answer?

  • Size: small. Ten to thirty checks is typical.
  • Speed: minutes, not hours. If it takes longer than making a coffee, it’s doing too much.
  • When: on every build or deploy, before anyone spends time on deeper testing.
  • What a failure means: stop. The build is rejected and goes back. Nobody tests further on it.

A smoke suite is the best place to start automating, because it runs so often. A flaky smoke test is worse than none, so keep it to checks that are boring and reliable.

02Sanity testing: did this change do what it should?

A developer fixes a bug where the discount code field accepted expired codes. A sanity check tries an expired code, a valid code and a code that expires today. It doesn’t retest the whole checkout. It answers one question: is this fix sane enough to bother with a full regression run?

  • Size: narrow, focused on the changed area and its nearest neighbours.
  • Speed: usually quick, and often done by hand, because it is specific to one change.
  • When: after a bug fix or small change, before wider testing.
  • What a failure means: the fix is rejected. No point running regression on a change that doesn’t work.

03Regression testing: did we break something that used to work?

Every change can break something that has nothing obvious to do with it. Fix the discount code, and somehow the order confirmation email stops including the discount line. Regression testing re-runs checks on features that already worked, to catch exactly that.

  • Size: the biggest of the three, and it grows with every feature you ship.
  • Speed: from minutes for a well-automated suite to days for a manual one.
  • When: before a release, and ideally on every merge for the parts that are automated.
  • What a failure means: something that used to work is broken. Find out whether the test or the app is wrong, then fix one of them.

04Side by side

SmokeSanityRegression
QuestionIs the build alive?Did this change work?Did anything else break?
BreadthWide, shallowNarrow, focusedWide and deep
Typical size10 to 30 checksA handfulHundreds to thousands
WhenEvery buildAfter a fix or small changeBefore release, and on merges
UsuallyAutomatedManual or quick scriptMostly automated
If it failsReject the buildReject the fixInvestigate and fix

05How they fit into one release

  1. A build is deployed to the test environment. Smoke runs automatically. If it fails, the build goes straight back.
  2. Testers check the specific stories and fixes in this build. That is sanity for fixes, and feature testing for new work.
  3. Once the changes look right, regression runs: the automated suite first, then any manual checks for risky areas.
  4. After release, a shorter smoke suite runs against production to confirm the deploy went out cleanly.

06Keeping the regression suite from growing out of control

Regression suites only ever get bigger, until they take too long to run and people stop waiting for them. A few habits keep them useful.

  • Run what the change touched first. Most tools can select tests by the files or features a change affects. Run those on every merge and the full suite nightly.
  • Delete duplicates. Ten tests that walk through the same checkout with slightly different products prove the same thing.
  • Keep negative tests. The tests that expect the app to refuse bad input are the ones that catch rules quietly disappearing. Don’t trim those to save time.
  • Fix or remove flaky tests quickly. A suite people don’t trust is a suite people skip.

Quick check for your team

  • Everyone uses the same meaning for smoke, sanity and regression
  • Smoke runs on every build and finishes in minutes
  • A failed smoke test rejects the build automatically
  • Fixes get a sanity check before the full regression run
  • Regression runs affected tests on every merge and everything nightly
  • Flaky tests are fixed or removed within a week

Frequently asked questions

What is the difference between smoke and sanity testing?

Smoke testing checks that a whole build basically works, across many features, very briefly. Sanity testing checks that one specific change or fix works, in a little more depth, before wider testing.

Is sanity testing a subset of regression testing?

It is often described that way, because it re-checks part of the system after a change. In practice it is narrower and happens first: sanity confirms the change works, then regression confirms nothing else broke.

Should smoke tests be automated?

Yes. They run on every build, so automation pays back quickly. Keep them few, fast and reliable.

How often should regression tests run?

The automated part on every merge for the affected areas, the full suite at least nightly, and the complete set including manual checks before each release.

Is smoke testing the same as build verification testing?

Yes. Build verification test (BVT) is another name for the same idea: a short set of checks that decides whether a build is stable enough to test.

Written by

Vihari Panyam

QualityAIQ

Sources and further reading

  1. ISTQB Glossary: smoke test, regression testing
  2. ISTQB Certified Tester Foundation Level
  3. Martin Fowler: The Practical Test Pyramid

More in Testing Fundamentals