Null hypothesis

The null hypothesis is the default position a statistical test starts from, usually that nothing interesting is happening: two groups have the same conversion rate, a treatment has no effect, or a coefficient is zero. It is written as H0 and is what a test is actually designed to argue against, not for.

hypothesis testing works by assuming the null hypothesis is true and asking how likely the observed data would be under that assumption, expressed as a p-value. A small enough p-value is treated as evidence against the null, leading analysts to reject it in favor of an alternative hypothesis that some real effect exists. Crucially, failing to reject the null is not the same as proving it true; it may simply mean the study lacked the statistical power to detect a real but smaller effect.

The null hypothesis matters because it sets the burden of proof: without a clearly stated null, "the data looks different" is not a testable claim. The most common misreading is treating a non-significant result as evidence that there is definitely no effect, when it may only reflect too small a sample. This distinction is formalized in the framework of Type I and Type II errors, where wrongly failing to reject a false null is a Type II error.