Predictions

Write down, before anyone looks, the exact number each rival explanation expects on one shared measure. If two rivals expect the same result, the test cannot tell them apart, so keep rewriting until their forecasts diverge.

Write down, before anyone looks, the exact number each rival explanation expects on one shared measure. If two rivals expect the same result, the test cannot tell them apart, so keep rewriting until their forecasts diverge.

Example

In the onboarding pilot: "new flow works" predicts signups up 8 to 15 percent in the test group and flat in the control. "Holiday season" predicts both groups rise together and both fall back by January. Same number, different forecasts.