Split testing (also see A/B testing) is the simplest form of an A/B test, a method for testing where the traffic is split in equal parts between the two variations, unchanged or control one and changed one - the variation. Split testing is conducted by randomly assigning visitors to control or variation, thus ensuring that the experiment is conducted under controlled conditions. Users are thus exposed to one or the other variation and their interaction gauged. The more variations are tested, the larger sample is necessary to ensure significance of the test. If you conduct split testing and start reducing the amount of traffic allocated to each variation as it appears to be losing, it is called bandit testing.