Excluding Users from an Experiment

You may have noticed an additional setting at the bottom of the rule configuration area when we set up our experiment. There is a dropdown menu below our distribution percentages next to the phrase "The remaining X% will be served":

This is because experiments can be provided to fewer than 100% of our users. If we change the distribution to send 10% of users to each variation, we will see that 70% of users won't be included in our experiment data at all:

It's a common practice in experimentation to exclude a percentage of your users. You may choose to exclude users when your site's traffic volume is very high. Say the feature you're experimenting on is interacted with hundreds of millions of times a day. In these scenarios, including just 1% of your total traffic would still result in a situation where your experiment achieves significant results very quickly.

If you're going to be excluding users from an experiment, we recommend determining the percentage you'd like to exclude prior to running your experiment. Changing the percentage of excluded users on a running experiment means that your users have a chance of receiving a new variation, or being excluded from the experiment after being included at a prior time. In the event that you do need to update the traffic distribution for a running experiment, it's best practice to Start a new iteration of your experiment from the Experimentation tab:

Congratulations! You've finished the lesson. We hope you've learned a thing or two about experimentation in LaunchDarkly. If you're interested in learning more, continue on to the take home challenge in Lesson 5.