Control Variant
The control variant is the unchanged, original version in an A/B test against which all new variants are measured.
In depth
The control exists to isolate the effect of a change. Because traffic is split randomly and simultaneously between the control and each challenger, external factors like seasonality, ad spend, or day-of-week patterns affect both groups equally, leaving the difference in outcomes attributable to the variant itself. Without a concurrent control, you would be comparing this week to last week, where countless confounders make any conclusion unreliable.
In practice the control is also your safety net: if every challenger underperforms, you keep the proven baseline and lose nothing but the testing window. A frequent pitfall is quietly editing the control mid-test or running it at a different traffic share, which breaks the apples-to-apples comparison. On a quiz funnel, the control might be your current lead-capture form while challengers test fewer fields or a progress bar, and only a stable control lets you trust that any lift in completions is real.
Example in practice
Frequently asked questions
Why do I need a control variant?
The control gives you a concurrent baseline, so any difference in results can be attributed to your change rather than to timing or traffic shifts. Without it, comparisons across time periods are easily distorted by outside factors.
Can the control change during a test?
No, editing the control mid-test breaks the comparison and invalidates your results. Lock the control for the full duration and only update it once a winner has been promoted.
Should the control get equal traffic?
In a standard A/B test the control and challenger usually receive equal splits for the fastest, cleanest read. Unequal splits are sometimes used to reduce risk but require more total traffic to reach significance.