Workflows nobody had written down
Every team relied on daily routines in the old system that existed only in people's heads.
All work
Mytheresa, early 2023
Shop migration: 6M+ customer accounts
Mytheresa's biggest project of the year was a new shop system, Magento to Sylius, for more than 6M customer accounts. I owned production readiness for about six months.
- Role
- Product owner for production readiness
- Timeline
- About six months, early 2023
- Team
- About 20 business teams, engineering, experts and POs
- Scope
- Magento to Sylius, 6M+ customer accounts
The problem
About 20 teams depended on a shop system that was about to be replaced, and nobody had written down how they used it.
Customer IDs that didn't match
Legacy customer IDs didn't line up with the new system, putting 6M+ accounts at risk.
A long bug list before launch
A month before the pilot, about 160 bugs were still open.
What I did
Turned workflows into acceptance criteria
Every team mapped its daily workflows into acceptance criteria and UAT cases, tracked in 12 weekly check-ins with experts and POs.
Fixed the data match with engineering
We matched records on another unique field instead of the legacy ID, and mismatches fell from 3% to under 0.1%.
Triaged by revenue impact
I ranked the open bugs by revenue impact, effort and launch criticality. All 25 launch blockers were cleared in 3 weeks.
The call
Results
The pilot went live on schedule with every launch blocker cleared.
6M+customer accounts migrated
~20teams made launch-ready
<0.1%record mismatches, from 3%
25launch blockers cleared in 3 weeks
What worked
- Writing each team's workflows down as acceptance criteria turned unknowns into testable cases.
- Triage by revenue impact focused the last month on what could actually block launch.
What didn't
- About 160 bugs were still open a month before the pilot, which made the final stretch tight.
What I'd do differently
I'd start the workflow mapping before the build rather than alongside it, so testing could begin earlier.