TestFinch

BlogFlowQA

Smoke vs sanity vs regression testing: what each checks and when it runs

Smoke testing checks that a new build works at all, sanity testing checks that one change does what it should, and regression testing checks that nothing else broke. One release of a small booking app, followed through all three.

Smoke testing checks that a new build starts and its main paths open. Sanity testing checks that one specific fix or change does what it was meant to. Regression testing checks that everything the change was not meant to touch still works. All three can run in one release, in that order, and each one decides whether the next is worth running.

One release, three kinds of test#

The example is a booking site for a hair salon. A customer picks a service, picks a free time, enters their name and phone number, and gets a confirmation by text message. The owner sees the day's bookings on a calendar.

Last week two customers booked the same 10:00 slot. The developer has fixed it: a time now disappears from the list as soon as someone books it, and a second attempt at the same time is refused. The fix is deployed to a staging copy of the site.

Smoke testing: does the build work at all?#

A smoke test is a handful of checks across the whole app, each one shallow. On the booking site there are four.

  1. The home page loads and lists the services.
  2. Choosing a service shows a list of times.
  3. The owner can sign in.
  4. The calendar opens.

The smoke test books nothing and does not look at the double-booking fix.

If check 2 shows an empty page, testing stops there. The fix cannot be checked on a site that shows no times, so the build goes back to the developer with that one finding.

A smoke test runs on every build, first, and again on production straight after the deploy.

Sanity testing: did this change work?#

A sanity test goes to the place that changed and stays there. For the double-booking fix it has four checks.

  1. Book 10:00 on Friday, and the confirmation shows.
  2. Reload the list of times, and 10:00 on Friday is gone.
  3. In a second browser that still shows the old list, try to book 10:00 on Friday, and the site refuses and offers the next free time.
  4. Cancel the first booking, and 10:00 on Friday is back in the list.

Check 3 is the bug as it was reported, written so that it fails on last week's build and passes on this one. Checks 2 and 4 are the things right beside the fix.

The sanity test does not open the owner's calendar or send a text message, because the change did not touch them. It runs once the smoke test has passed, on each build that carries the change, and the person who fixed or reported the bug usually runs it.

If check 3 still lets the second booking through, the fix did not work and there is no reason to test anything wider yet.

Regression testing: did anything else break?#

A regression test covers what already worked. The booking site's suite has grown one test at a time.

  • A customer books a service and receives the text message.
  • A customer cancels from the link in the message.
  • The owner blocks out a lunch hour and those times are not offered.
  • The owner moves a booking and the customer is told.
  • The calendar shows the day's bookings in order.
  • A booking made at 23:30 lands on the right day.

None of these is about double booking. The fix changed how free times are worked out, and the owner's lunch hour uses the same code. In this release the regression run finds that blocked lunch hours are being offered to customers again. The smoke test passed and the sanity test passed, and the suite still caught a bug that the first customer to book at 12:30 would otherwise have found.

A regression suite runs before every release, after every deploy and on a schedule. It is long and identical each time, so it is the one to automate first. A failure in it is a bug, or a test to update because the app changed on purpose. Check 3 of the sanity test joins it now, so double booking cannot come back unnoticed.

Smoke, sanity and regression testing compared#

Smoke Sanity Regression
Question Does the build work at all? Did this change work? Did the change break something else?
Covers The main paths of the whole app One fix or feature and what sits beside it Every feature people rely on
Depth Shallow Deep, in one place Deep, everywhere
Runs On every build and after every deploy When a build carries a specific change Before every release and on a schedule
Usually run by A script The person who made or reported the change A script
If it fails Stop and send the build back The change is not done A bug to fix before release
Kept afterwards Yes, unchanged The reported case moves into regression Yes, and it grows

Where the names blur#

Some teams call the sanity test a "smoke test of the fix", and some call a short regression run a "sanity run". Teams that write unit tests often say "smoke test" for the first few that run in CI. The ISTQB glossary, read on 18 November 2026, defines smoke testing and has no entry for sanity testing, and some testers use the two names for the same quick check of a build. The same glossary calls check 3 of the sanity test, the re-check of a fixed defect, confirmation testing. When a colleague uses one of the names, ask which of the three questions they mean.

Sanity testing is sometimes described as a subset of regression testing. That holds for the checks beside the fix, which re-check things that used to work. The check of the fix itself is new, so it has nothing to regress from until it joins the suite.

Templates and the other types of test#

The free templates have worked rows for each: the smoke test template, the sanity test template and the regression test template. Types of software tests places these three beside functional, end-to-end, integration, unit, visual and exploratory testing, with a table of which question each answers.

What FlowQA runs#

FlowQA writes and runs the automated two. Its AI agent explores your web app and writes end-to-end tests in plain steps; a suite of a few of them, run after every deploy, is the smoke test, and all of them on a schedule or from CI are the regression suite. The sanity check of a fix stays with a person. Once the fix is confirmed, a bug filed in Snag can become a FlowQA test from the steps Snag captured, which is how check 3 of the sanity test would join the regression suite. FlowQA does not write unit tests, and its tests drive pages in Chromium only.

FlowQA explores your site, writes the tests, keeps each one only once it fails on a page broken the way its checks describe, and runs them after every change.

See FlowQA