On the Experiment summary page you find the option to change the traffic distribution of an experiment. You can only do that per experiment and not per variation.
The distribution allows you to control what percentage of visitors you like to test. In the example above you can see that 25% of the visitors will not participate in the experiment and will see the original version.
Their behavior will not be monitored. The test group of 75% is equally divided over 37.5% and 37.5% (displayed here as 37%).