Run A/B tests and personalize your store without code. 14-day free trial, no credit card.
Start free trial →In-depth guides and case studies where this concept is put to work.
The alternative hypothesis (H₁ or Hₐ) in A/B testing is the claim that a real, non-random difference exists between the control and the variant on the target metric. It is the hypothesis you are trying to gather evidence for. When statistical significance is reached and the null hypothesis is rejected, the alternative hypothesis is accepted — meaning the data supports the conclusion that the observed difference reflects a genuine effect of the change being tested.
Alternative hypotheses can be directional or non-directional:
Two-tailed (non-directional): The variant conversion rate differs from the control in either direction (better or worse). Most A/B tests use two-tailed tests.
One-tailed (directional): The variant conversion rate is specifically higher (or lower) than the control. Used only when you have strong prior evidence about direction.
One-tailed tests require a smaller sample size to reach the same significance level, but they are only valid when the direction of improvement is genuinely predictable — using them opportunistically inflates false positive rates.
The alternative hypothesis is where your business logic lives. A well-written H₁ connects the CRO hypothesis to a real customer behavior change: "Adding a free-shipping threshold progress bar will increase average order value because customers will add items to qualify for free delivery." Specificity in the alternative hypothesis prevents teams from retrofitting a narrative onto ambiguous results after the test ends. It also forces clarity about what "better" means — is the alternative hypothesis about conversion rate, AOV, RPV, or checkout completion rate? The answer shapes the entire test design.
The Man Company's CRO team designed an experiment on their subscription product page. Their alternative hypothesis: "Displaying a per-unit cost comparison (₹X/day vs. ₹X one-time) will increase subscription plan selection rate compared to showing only total price." The null hypothesis was that plan selection rate would be the same. After 24 days, the per-unit display variant showed an 18% higher subscription selection rate at 97% confidence. H₁ was accepted: the framing change had a real, measurable effect on the metric specified in the hypothesis before the test started.
In A/B testing, the alternative hypothesis is accepted when the p-value falls below the pre-set alpha threshold. Critically, accepting H₁ does not prove it is true in an absolute sense — it means the data provides sufficient evidence to reject the assumption of no effect. The strength of that evidence is quantified by the confidence level and the observed effect size.
Run smarter A/B tests with CustomFit.ai — 14-day free trial, no credit card required.