Web app

Cutting a 6-hour regression suite to 22 minutes

A B2B SaaS team was testing every release by hand. We rebuilt the suite in Playwright, sharded it across CI runners and made releases boring again.

Client
B2B SaaS platform (NDA)
Role
QA Automation Engineer
Timeline
8 weeks · 2025
Stack
  • Playwright
  • TypeScript
  • GitHub Actions
  • Allure
  • 22 min Regression runtime (was 6 h)
  • 148 Tests automated end to end
  • 0 Flaky tests in 90 days

The challenge

Two testers spent most of every sprint on a six-hour manual regression checklist. Releases slipped, flaky results were ignored, and nobody trusted the last green run.

The approach

Mapped the five revenue-critical user journeys, then built a Playwright + TypeScript framework with fixture-based test data and a deterministic login helper. The suite runs sharded across four GitHub Actions runners on every pull request, with Allure reports linked in the PR and a quarantine rule for any test that flakes twice.

The outcome

Regression time dropped from six hours to 22 minutes, and the team moved from weekly releases to daily deploys. The manual checklist now covers exploratory and edge-case testing only.

Details

The interesting constraint was test data: every spec needed a new tenant without slowing the run down. We seeded tenants through the API, cached them per worker, and cleaned up in a global teardown. That single decision removed most of the flakiness.

Need a test suite your team actually trusts?

Send me your stack and your biggest testing headache. You will get a scoped plan with timelines and a fixed price — usually within one business day.