When data are collected in a survey (an observational study), relationships between the variables may be observed, but it is difficult to conclude that either variable causally affects the other. On the other hand, data that are collected in well-designed experiments do give information about causal relationships.

If an experiment is poorly designed, the data may not contain the information necessary to answer questions posed by the investigator. Many experiments have been conducted at great expense and over a long period of time, from which conclusions could not be drawn due to poor statistical design.

More positively, good statistical design of experiments will provide more accurate answers to questions of interest and can save a considerable amount of money.