The principal reason for using a "randomization" technique is to avoid predictability in the treatment assignment; i.e., neither the patient nor the investigator knows or controls the treatment assignment before it is made. Successful randomization eliminates selection bias on the part of the study subject or the investigator, thus ensuring that characteristics of the patients do not influence the treatment that is assigned. Successful randomization also tends to create comparison groups that are balanced in terms of both known and unknown factors that may be related to outcome, however, because of the random nature of treatment allocation, imbalances in baseline factors may exist.

Because randomization does not guarantee that the treatment groups will be identical in terms of baseline characteristics, stratification before randomization (e.g. by gender, age group, disease stage, or other prognostic factor) may be used as an additional means of ensuring balance in the treatment groups for know variables. Stratification is usually used when there is an important known prognostic factor, such as tumor stage, that can be measured before randomization.