how to test three formats without wasting the budget.
> pick three formats that are structurally different, not three hooks on the same format. talking head, text-over-reaction, and silent demo are three formats. three different opening lines are one.
> same product moment in all three. if the demo changes you are testing two variables.
> same posting window, different accounts.
> run until each has enough posts that one flop does not decide it. below ten you are reading noise.
> judge on installs, not views. sort the finished set by installs and it is a different ranking almost every time.