About this app

Two-proportion power calculation with an inverse-normal approximation, taking a baseline rate, a relative lift, confidence and power, and returning observations per variant plus the calendar time at your traffic. A log-scale curve shows the requirement exploding as the effect shrinks, and a table gives the smallest detectable lift at common run lengths. Makes the argument that sample size scales with the inverse square of the effect, which is why small teams should test big changes rather than button colours.