No matter how successful your campaigns get, there's always a possibility you could improve some of its elements. A popular way to reveal the strong and weak points of your campaigns is A/B testing. You need to prepare two versions for one element of your program. Then, by including a percent split object, you can randomly send contacts to different step content, or down different paths.

Using percent split to establish a control group

The basic idea of a control group is to determine the effectiveness of your campaign. By adding a percent split, you can purposely exclude a random sample of contacts from a program to prevent them from receiving messages. After your program completes, compare the activity or level of engagement of the control group (0) against the contacts that participated in the program.

1. Connect a percent split at the point where you want to test the impact of your program. You can start from the Audience icon or after further steps to analyze a particular program stage. If you are using an event-triggered autoresponder program, place the control group after the first step.
2. Set the split to 90 and 10. Allow the 90% to continue through the program and 10% to be held back or exited. 
3. After messages are sent, you can review open and click rates for each message to determine which version is more successful or see how many contacts move down a specific path.

Example: test email subject lines in a welcome program

1. Create a Welcome program, entering contacts as they opt in to your database as an event-triggered entry.
2. After the first message, connect a percent split to divide the contacts 50/50 as they arrive, sending a message with Subject line A to 50% of the contacts, and a message with a different Subject line B to the other 50%.
3. After a few days, check the email reports to see if email A or B had the most opens and clicks.
4. If you determine a clear winner based on email metrics, deactivate and adjust your program to send all contacts to the winning version, by changing the percentage split to 100% and 0%. After re-activation, subsequent contacts follow the 100% split.
5. Percentage splits can be added to decision paths; either rules-based paths or remainder paths to test various scenarios.

Example: a re-engagement program

This example describes how to set a control group for contacts that participate in the program, so you can measure the success of the program or its specific path.

1. Target contacts who did not open an email in the last 6 months.
2. Of the contacts that qualify for the program based on that rule, when they enter, hold out 20% and allow 80% to move forward through the program.
3. You can then compare the open rate of the 80% that proceed through the program to the ones in the control group that receive no messages.