A large gaming company is developing a measurement strategy designed to shift its planning and buying on Facebook. The company believes that its ability to iterate on testing will provide more long-term growth than the short-term sales boosts from individual campaigns. The conversion rate is high enough that they should have NO issues with statistical power at any size of test/holdout groups. It is planning on running multi-cell Conversion Lift tests on a monthly basis.
What test/holdout proportions should the company use for each test cell?