Our Journey to Fully Automated Weekly Releases
summary
Company replaces biweekly manual validation with an automated pipeline that runs end to end tests on PRs via the GitHub merge queue and Flank, verifies UI with Paparazzi screenshots, and uses Kakao Espresso for resilient UI flows. Release build submits Saturday, rollout increments 1 → 15 → 75 → 100% by a croned GitHub Action calling BigQuery and Google Play API, halts hourly on bad SLOs, posts daily 9:00 status, and cuts time-to-100% from ~21 to 9 days while saving ~1.5–2 hours per engineer per two weeks.
AI-written summary · about
chapters (16)
- 0:00Weekly release automation intro
- 1:00Manual release struggles
- 3:4021-day release pain points
- 7:37New automated weekly system
- 8:01Automation goals
- 9:26End-to-end testing strategy
- 13:27Kakao Espresso & Flank reliability
- 15:34Paparazzi screenshot testing
- 18:27Full test automation confidence
- 20:12CI fast feedback loops
- 22:49Automated rollouts & monitoring
- 24:54Release health dashboards
- 25:38SLO fast rollouts
- 26:43Fully automated results
- 27:46Automation ROI
- 28:49Q&A
related talks
50mRelease & Iterate Faster: Stop Manual Testing
25mJourney to Automated Releases
32mClimbing the Mountain of CI/CD: A Tester’s Tale
23mAchieving Weekly Releases: A Technical Deep Dive into Our Development Process
20mStreamlining Release Management for DevOps Success
54mFrom Code to Customer
47mFrom bi-annual to fortnightly releases in 4 months for 15 teams and one monolith
39mOur journey to automated releases