The volume problem
A meaningful conversion rate test needs more traffic than most stores send in a month. Calling a winner early is how people confidently ship the worse page.
What to do instead
- Fix the obvious first — speed, shipping information, price visibility, mobile layout. These do not need testing.
- Test big changes, not button colours. Bigger differences need less traffic to detect.
- Use qualitative signals: session recordings, support questions, and where people stop scrolling.
If you do test
Run it long enough to cover a full weekly cycle, and do not change campaigns during it.