BLOG
BlogGuides

Release checks for a startup with no QA team

No QA engineer, two deploys a day, and a manual click-through that gets skipped. A minimal release check that a small team can keep: five tests, one gate, one channel.

The honest situation#

Most early teams have no one whose job is testing. Developers test their own change, someone clicks around before a big release, and most deploys go out with neither. It works until the day sign-up breaks for six hours.

The usual advice (a full regression suite, formal test plans, a QA sign-off) is written for teams that have time. A minimal process has to be smaller than the thing it protects, or it gets abandoned.

Five tests#

Pick the five journeys whose failure would cost you a customer this week. For most products:

  1. A new user can sign up and confirm their email.
  2. A returning user can sign in and see their data.
  3. The core action works and saves (create, edit, reload).
  4. The money path works, as far as is safe (checkout in test mode, or the upgrade page loads with the right price).
  5. One page a customer relies on daily renders its key value.

Write each as a plain-language test. Anyone on the team can, including the founder who knows how the product should behave:

Sign in with the stored account "customer".
Open the dashboard and check the "Revenue this month" card shows a
number with a currency symbol, not "NaN" or a dash.

One gate#

Run the five after every deploy, from CI: GitHub Actions, GitLab or any CI. The job fails when a check fails. Five runs per deploy is a cost you can see on the Plan and usage page, project by project.

One channel#

Send failures to the channel where the team already talks: Send test failures to Slack. Each message has what was seen and a link to the replay, so nobody has to reproduce it to understand it.

One habit#

When a bug reaches production, add a test for it before you close it. That is how five tests become fifteen useful ones without anyone planning a suite: one bug, one test.

What to skip for now#

  • Exhaustive edge cases. Add them when they bite.
  • Pixel-level visual checks.
  • A test for every page. Five that run every time beat fifty that run never.
A project: tests grouped by their last result, with Run on each and Run all for the project.
A project: tests grouped by their last result, with Run on each and Run all for the project.

This post is part of Release testing for product teams.

Filed underWebReleases and regressionsCI, schedules and agents