A/B testing on low traffic: what you actually can test

Classic A/B tests need thousands of conversions to reach significance; a small store almost never gets there. Test only big swings, run clean before/after periods, and let five-user tests do the work volume can't.

Most A/B advice assumes thousands of conversions a week, so on a low-traffic store a 50/50 split rarely reaches significance and 'winners' are usually noise. The realistic playbook: do a sample-size reality check first, test only large changes (offer, hero, price framing), run sequential before/after periods with guardrails when you can't split, and use qualitative methods (5-user tests, session recordings) that need five people, not five thousand. Never peek and stop at the first apparent win.