What is a control group?

A control group is a common tool that researchers use. It allows them to prove a cause-and-effect relationship with an independent variable. This variable does not change for the control group. In this sense, the control group is the status quo. Researchers compare the effects in the experimental group against the control group.

The independent variable is the thing the researchers are testing. They are trying to determine whether it’s responsible for any change that occurs in the experiment. The research control group is key for this as it allows them to isolate the independent variable’s effect on the experiment.

Splitting the audience you’re testing into two identical groups will give you a control group and an experimental group.

Nothing will change for the control group during the research. For example, this group would receive a placebo in pharmaceutical research.

In contrast, one key variable changes for the experimental group. In a pharmaceutical experiment, researchers might administer a different drug. In advertising research, this might involve increasing the experimental group’s exposure to ads.

When evaluating the results, researchers will compare those obtained from the experimental group against the control group. The control group is the baseline.

In research where the two groups are truly identical, seeing different results between the groups suggests they were caused by the independent variable—the only thing that changed.

Examples of control groups in research exist in a wide range of business contexts. For example:

• You want to test whether a 15% loyalty discount for repeat purchases would positively impact retention and revenue. So, you send a discount email to 50% of your customers who were randomly selected. The other 50% of customers are your control group.

• You want to test whether a personal sales call will increase your chance of a sales conversion. You add this step to your existing nurturing campaign for a randomly selected portion of leads. Those who don’t receive a phone call are your control group.

• You want to test whether different product packaging can change brand perceptions. To do this, you change the packaging for a randomly selected portion of customers. Customers who receive the same packaging as before are your control group. Sending a survey to all customers about their brand perceptions before and after the experiment will reveal the impact of the new packaging.

These are just some of the countless examples of control groups. Perhaps the most well-known example is in the medical field, where placebos treatments are used. Control groups receive placebo treatments under the exact same conditions as the experimental group to determine the treatment’s effects.

Control groups matter in research because they act as the benchmark to establish your results’ validity. They enable you to compare the results you see in your experimental group and determine if the variable you changed caused a different outcome. 

Control groups and experimental groups should be identical in their makeup and environment in every possible way. You’ll be able to draw more definitive conclusions as long as the research process is identical for both groups. In other words, working with control groups improves your research’s internal validity.

Control groups are most common in experimental research, where you’re trying to determine the impact of a variable you’re changing. You split your research group into two groups that are as identical as possible. One receives a placebo, for example, while the other receives a treatment.

In this environment, the identical makeup of the group is essential. The most common way to accomplish this is by randomly splitting the group in two and ensuring that any variables you’re not testing remain the same throughout the research process.

You can also conduct experiments with multiple control groups. For example, when testing new ad messaging, the split between two control groups and one experimental group may be as follows:

• Control group 1 receives no advertising

• Control group 2 receives the existing advertising

• Control group 3 receives the new ad messaging

This more complex type of experiment can test both the overall impact of ads and how much of that impact you could attribute to the new messaging.

Control groups are less common in non-experimental research but can still be useful. They most commonly occur in the following process designs:

Matching design

In this research process, every person in the experimental group is matched to one other person based on their environmental and demographic similarities.

This is most common when randomly selecting two groups on a broader scale would not result in them being equal. It can help you ensure that the control group or individual continues to act as the baseline for the variable you are studying.

Quasi-experimental design

This is where multiple groups are part of the research, but they are not randomly assigned to test and control conditions.

Quasi-experimental design is most common when the groups you are studying already exist, like customers being shown new ad messaging versus non-customers. The control group in this example is made up of your non-customers, as the variable did not change for them.

While control groups tend to be similar across research contexts, they generally fall into two categories: negative and positive control groups.

Negative control groups

The independent variable does not change in a negative control group. This group represents the true status quo, and you would test the experimental group against it.

Examples of negative control groups include many of the experiments listed above, like only changing product packaging or only offering a discount for one group of customers.

Positive control groups

In positive control groups, the independent variable is changed where it is already known to have an effect. You would compare this group’s results against those from the experimental group receiving a variation of the same independent variable. This would enable you to determine if the effect changes.

In the example of a multi-control group experiment seen above, control group 1 (receiving no advertising) is a negative control group, while control group 2 (receiving the current level of advertising) is a positive control group.