Experiments are used in design and research to measure the effects of actions or features by showing the causal relationship between those actions or features and specific outcomes (e.g., behaviours). Typically, a hypothesis is posed about such relationships and then the actions or features are manipulated in controlled ways while the outcome is measured. In the simplest case (say low sweetness vs high sweetness in a beverage), the comparison of outcomes (amount drunk, preference rating) between the two conditions provides a simple model of the impact of the manipulation of sweetness.
A/B testing is used to compare two versions of a design to see which performs better against a specific outcome (return coupon, sales, click through, etc). The methodology is that of a randomised experiment with two variants, A and B (control and test). Such testing is common in website optimisation and experience design in order to maximise desired outcomes and identify changes that can increase sales (where thousands of users can mean that even marginal increases in specific behaviours can have a significant impact on success). A/B testing is also called split testing. When there are multiple versions of a design, such testing is called multivariate or bucket testing. Read more »