Smoke vs Sanity vs Regression Testing — Differences, CI/CD Flow & Examples

Smoke testing is a broad, shallow check that verifies a new build is stable enough to test. Sanity testing is a narrow, focused check that verifies a specific change or bug fix works as intended. Regression testing re-runs the wider test suite to confirm the change did not break existing functionality.

In a CI/CD pipeline they run in that order: Smoke → Sanity → Regression → Deploy. Smoke gates the build, sanity scopes the regression, and regression gates the release.

Use this table as your featured-snippet reference and interview cheat-sheet:

Related: 25+ types of software testing · Software Testing Life Cycle (STLC).

1. Commit → Build. Developer pushes; CI builds and produces a deployable artefact.
2. Smoke stage (under 5 min). 20–50 critical-path tests tagged @smoke. If any fail, the pipeline stops and the build is rejected — this is the fail-fast gate.
3. Sanity stage. Runs a subset of tests scoped to the changed module (e.g. only checkout tests when the pricing service changes). Decides whether the full regression is worth the compute cost.
4. Regression stage. Full suite tagged @regression, usually sharded across parallel runners. Nightly for main, and always before a production deploy.
5. Deploy. Green pipeline → canary → 100% rollout. Post-deploy synthetic smoke tests monitor production.

In practice, teams use tags/annotations to filter which subset runs at which stage. Example with Playwright + TypeScript:

1test('user can log in @smoke @regression', async ({ page }) => {
2  await page.goto('/login');
3  await page.fill('#email', 'qa@example.com');
4  await page.fill('#password', 'secret');
5  await page.click('button[type="submit"]');
6  await expect(page).toHaveURL('/dashboard');
7});
8
9test('EU tax applies to Ireland orders @sanity', async ({ request }) => {
10  const res = await request.post('/api/cart/tax', { data: { country: 'IE', total: 100 } });
11  expect((await res.json()).tax).toBe(23);
12});

Then in GitHub Actions:

1- name: Smoke gate
2  run: npx playwright test --grep @smoke
3- name: Sanity (only if smoke passes)
4  run: npx playwright test --grep @sanity
5- name: Nightly regression
6  if: github.event_name == 'schedule'
7  run: npx playwright test --grep @regression --shard=${{ matrix.shard }}/4

Prefer JUnit or TestNG? The same pattern applies via @Tag("smoke") and mvn test -Dgroups=smoke. See our Playwright and Selenium Q&A hubs for framework-level detail.

• A new build arrives from dev. → Smoke. Verify build stability before spending compute on deeper suites.
• A hotfix patches a tax-calculation bug. → Sanity on the tax module, then a targeted regression on checkout.
• A framework upgrade (React 19 → 20). → Full regression, sharded, plus visual regression.
• A config-only feature flag toggle. → Smoke only; skip regression.
• Pre-production release. → Regression is the mandatory gate. No green regression, no deploy.

• Bloating the smoke suite. Once smoke exceeds ~5 minutes, developers stop trusting the gate and start bypassing it.
• Skipping regression on urgent hotfixes. The one-line fix that broke everything is a weekly incident in teams that do this.
• Not tagging tests. Without @smoke/@sanity/@regression tags you cannot filter suites — every stage runs everything.
• Confusing sanity with smoke in stand-ups. They are different — clarify in your interview answers using the table above.

In a QA or SDET interview, the golden answer is: “Smoke is build-acceptance, sanity is change-acceptance, regression is release-acceptance.” Then reference the CI/CD flow — Smoke gates the build, sanity scopes what to re-run, regression gates the deploy.

Practice this live in the AI Mock Interview, and drill 300+ related questions in the Software Testing Interview Q&A hub. Landing an SDET role? Update your CV with the ATS Resume Review and browse live openings in the QA Jobs Radar.