How many posts do you need to test before you know what works?
Fewer posts than you think for a first read, more than you think for a reliable answer.
3 posts: a first read
Three different posts on the same account, judged at the same moment, tell you which idea people responded to most. It is a ranking, not proof. One post can do well by luck.
That is what our free test is for: 3 posts, 10 days, and a clear answer on which one did best.
20 posts: a pattern
With around 20 posts across a few accounts, luck starts to even out. If the same idea keeps winning across accounts and weeks, you can trust it enough to spend on it.
The pilot runs 20 posts in weekly groups of 5, so each week can build on what the last one showed.
Test ideas, not small tweaks
Changing a filter or one word rarely moves the numbers enough to see. Test different ideas: a different first second, a different setting, a different reason to care.
Once one idea wins, then try small changes on that idea.
Decide the bar first
Write down what counts as a win before anything goes live. Ours is 1,000 views and a 1.5% save + share rate at 7 days. If you set the bar after seeing the results, every test looks like a success.