Control Group Study (a.k.a. Randomized Controlled Study)

A model of evaluation in which the performance of cases who experience an intervention (the treatment group) is compared to the performance of cases (the control group) who did not experience the intervention in question. In medical studies where the intervention is the administration of drugs, for example, the control group is known as the placebo group because a neutral substance (placebo) is administered to the control group without the subjects (or researchers) knowing if it is an active drug or not. Typically, the intervention is considered successful if its performance exceeds that of the control group's by a statistically significant amount. When assignment to control and treatment groups is made at random, and no other factors enter into the assignment into control or treatment, any differences between the two groups are due either to the treatment or to random variation. When a given difference between the two groups is observed, say in spending on a particular set of items, it is possible to calculate the probability of this difference arising purely by chance. If the probability of an observed difference is very small (generally less than 5 percent but more stringent rules can be adopted) the observed difference is said to be due to the treatment.


Control Group Study: (a.k.a. Randomized Controlled Study) A model of evaluation in which the performance of cases who experience an intervention (the treatment group) is compared to the performance of cases (the control group) who did not experience the intervention in question. In medical studies where the intervention is the administration of drugs, for example, the control group is known as the placebo group because a neutral substance (placebo) is administered to the control group without the subjects (or researchers) knowing if it is an active drug or not. Typically, the intervention is considered successful if its performance exceeds that of the control group's by a statistically significant amount. When assignment to control and treatment groups is made at random, and no other factors enter into the assignment into control or treatment, any differences between the two groups are due either to the treatment or to random variation. When a given difference between the two groups is observed, say in spending on a particular set of items, it is possible to calculate the probability of this difference arising purely by chance. If the probability of an observed difference is very small (generally less than 5 percent but more stringent rules can be adopted) the observed difference is said to be due to the treatment.


Coupon Example: instead of receiving a coupon(s) that they are eligible to receive, a predetermined fraction of eligible customers are assigned to a coupon's control group and will receive either an alternate coupon (placebo), or no coupon at all. The determination of whether a customer receives a coupon or is assigned to the control group is made at random. Afterwards the performance on the marketing objective, usually a function on revenue over a set of items, is measured for the group that received the coupon and for the control group and the statistical significance of the difference in performance is calculated. For example a Tide coupon targeted at customers who have bought in the cleaners category but have not bought Tide, with a 20% control group setting will mean that one fifth of the customers in the segment will be placed in the promotion's control group. Control Group Size: the size of the control group needs to be sufficiently large to support a statistical comparison. Ideally the control group includes over two hundred cases, but may be as small as fifty. See t-test.

Back to Glossary