Variant

Variant is any version of a webpage, feature, or element being tested in an A/B or multivariate test, including both the original control version and any modified treatment versions.

Meaning & Context

In the simplest A/B test, you have two variants: the control (A) and one treatment (B). In A/B/n tests, you might have three or more variants (A, B, C, D, etc.), each representing a different hypothesis about what might improve performance. Each variant should differ in specific, measurable ways so you can identify which changes drive performance differences. All variants should be tested simultaneously under identical conditions to ensure valid comparisons.

Why It Matters

The number and design of variants directly impacts your testing strategy, required sample size, and time to statistical significance. More variants require larger total sample sizes and longer test durations, as traffic is split across more groups. Each variant should represent a meaningfully different approach rather than minor tweaks, and you should limit the number of variants to avoid overly long tests or insufficient power. Well-designed variants test distinct hypotheses that can provide actionable insights regardless of which wins.

Example

You're testing pricing page layouts with three variants: Variant A (control) shows monthly pricing prominently, Variant B emphasizes annual pricing with a discount badge, and Variant C displays both equally with a comparison table. You'll split traffic equally and measure which variant generates the most paid conversions.