The situation
A B2B SaaS platform with 60 engineers was spending two days of manual testing per release and still producing an average of nine production hotfixes a month. Nobody could state what was covered, so release scope negotiations were guesswork.
What we did
- Wrote a test strategy and mapped the 31 journeys that produced revenue or support load.
- Built a Playwright regression suite in their repository, running on every pull request.
- Added k6 load tests with budgets that fail the build rather than warn.
- Introduced defect templates requiring reproduction steps and evidence, which cut triage time sharply.
The results after two quarters
| Metric | Before | After |
|---|---|---|
| Release testing time | 2 days | 35 minutes |
| Production hotfixes per month | 9 | 1 |
| Automated journey coverage | 0% | 84% |
| Escaped defects per release | 14 | 2 |
What mattered most
Ordering the automation by cost of failure rather than ease of implementation. The first eight journeys automated covered the paths responsible for most of the previous year of incidents, so the defect curve moved before the coverage number looked impressive.
Case studyQASaaS