A/B test
An A/B test splits traffic between two groups: variant A (original) and variant B (modified). Statistical comparison of performance determines which converts better before rolling out the winning version to 100% of users. Proper sample sizes are required for reliable results.
In practice
An A/B test requires sufficient traffic volume to reach statistical significance — below 1,000 visitors per variant, results are unreliable. Common tools include VWO, AB Tasty, and Convert (Google Optimize was discontinued in 2023). The highest-impact tests on e-commerce pages typically involve the product page title, the color and text of the buy button, and the sort order of internal search results.