Startup without QA: the first 5 quality habits
No QA team, developers shipping fast and bugs reaching your customers? Five simple habits to put in place this week, without hiring or slowing down.
In a startup, quality is rarely the topic of the day. You ship, you fix, you ship again. Until the day a bug blocks payments, or loses you a big customer during a demo.
The good news: you don’t need a QA team to avoid most of these accidents. Here are the five habits I recommend first.
1. List the 5 journeys that must never break
Spend an hour with product and the developers, and answer a single question: if it breaks, do we lose money or a customer?
Usually the list fits on one hand: sign-up, login, payment, the core feature, and the export or integration your customers depend on. These journeys become your top priority, for manual testing as well as for automation.
2. Write an acceptance checklist before every release
Before automating anything, a shared checklist already works wonders. Ten lines, no more, one per critical journey, with the expected result. Someone runs through it before every release, and it’s never the person who built the feature.
It’s low-tech, and that’s exactly what makes it effective from day one.
3. Automate the critical journeys first, end to end
When automation arrives, start with the journeys on the list, and only those. A few end-to-end tests with a modern tool like Playwright cover most of the risk. Unit tests still matter, but they won’t tell you that the “Pay” button has stopped responding.
Golden rule: an end-to-end test must be stable. A test that fails one time in five for no reason will be ignored within a week, and it does more harm than good.
4. Plug tests into CI from the very first test
A test that only runs on a developer’s laptop doesn’t exist. From the first automated test, plug it into your CI (GitHub Actions, GitLab CI…) so it runs on every pull request. A red PR doesn’t go to production: it’s simple, visible, and it builds a quality culture without speeches.
5. Make bugs easy to report, and to reproduce
A badly described bug costs three round trips. Create a ticket template with four mandatory fields:
- the steps to reproduce;
- the actual result;
- the expected result;
- the environment (browser, device, test account).
Add a screenshot or a video, and you’ll cut the time to fix in half.
What about AI?
These five habits are the foundation on which AI can then speed things up: generating test cases from your user stories, agents testing in CI, automatic bug triage. But an AI agent plugged into a team with no identified critical journeys will only automate the chaos.
