Experimentation loop
How UXLAB runs diagnose → hypothesis → test → ship so conversion gains compound instead of resetting every quarter.
Tools used
How it works
01
Pick one leak
One journey, one metric, one owner.
- Rank drop-offs by revenue, not by pageviews.
- Write the hypothesis in one sentence.
- Define the kill criteria before launch.
02
Ship the smallest change
The test is a live product change, not a mock.
- Build in production with a flag or A/B tool.
- Instrument the same events on both variants.
- Cap runtime so the next loop can start.
03
Keep only winners
Losers go away. Winners become default.
- Promote the variant and remove the flag.
- Log the learning in the playbook.
- Start the next leak from the new baseline.
