Glossary · Testing · also A/B test
What is a split test?
A split test randomly assigns visitors to two or more variants of an experience and compares an outcome between them. Randomisation is what makes the comparison causal: any difference in the outcome can be attributed to the variant rather than to differences between the people who saw it.
- Reviewed 6 October 2026
In more detail
The failure mode on smaller stores is not the test design, it is the arithmetic. A test judged on purchases needs many conversions per variant to separate a realistic effect from noise, and many stores cannot supply that in a sensible timeframe.
The response is not to abandon testing but to choose the metric with care. Judging a product-page change at add-to-cart rather than at purchase raises the event rate substantially, at the cost of measuring a narrower question; a guardrail on revenue keeps the narrower read honest.
Key points
- Randomisation is what makes the result causal.
- Tests judged on purchases are often out of reach for low-traffic stores.
- A power check before the test says whether it can conclude at all.
How Liftable relates
Related terms
Guides that go deeper
- Testing · 5 minHow to split-test when you don’t have the trafficMost stores can’t read a purchase-level A/B test in a sensible time. Why, and the adjustments that make testing work at the volumes stores actually have.Read
- Testing · 7 minHow to test a fix without breaking your live storeTest through a route that never edits your theme’s code, read the change first, run one test per surface, keep the rollback to one switch. The safe sequence.Read
Last reviewed against the product on 6 October 2026.
Scan your store.
See where your store is losing revenue and what to fix first. It is free, takes about a minute, and needs no install.