Thanks for stopping by!

Do you mean equal population when the variations are broken down by segments? If so, yes, the breakdowns should be similar between the variations otherwise it could point to some issues with randomization.

If you mean having a 50-50 split between variations, that is ideal but it depends on the risk appetite etc. The important thing is to not change the split wildly while the experiment is running.

Both the examples are somewhat extreme but hopefully convey the point well.