The issue here is that in a Split URL test, control receives all or far more visitors than any variations.
The problem arises because in a Split URL test, when a query parameter submits to the test, it is submitted to control as well as forwarded to the variation. Now, since the query parameter is carried forward, the variation URL will not match the URL entered in the test as it has some extra parameters.
This results in the visitor count increasing for control but not for the variations, hence resulting in a large mismatch between the number of visitors seeing control and any variations.
To solve this problem, set up your control and variation URLs in the manner shown below:
- Control: *http://www.example.com/clothing/
- Variation 1: http://www.example.com/clothing/?queryparam=x*
- Variation 2: http://www.example.com/clothing/?queryparam=y*
Note: The number of asterisks (*) in Control URL pattern should be equal to the number of asterisks in Variation URL pattern. The opposite does not hold.