Cadence tests that stay fair to users and to the data
More messages are not a strategy. How to design frequency experiments that your holdout group can actually teach you from.
A cadence experiment collapses when the “heavy” arm receives better creative or better offers. Keep copy constant. Change only spacing and count within a fixed period.
Define stop rules before you start: unsubscribe spikes, complaint rates, or a drop in same-day action that exceeds your tolerance. Timing analytics is not only about opens—it is about whether people still want to hear from you next week.
Share the readout in plain language: what changed, who was held out, and what you will schedule going forward. Stakeholders remember decisions; they forget charts.