Explanation
It is the discipline that makes volume worth anything. Posting twenty videos without varying anything deliberately produces twenty results and no information; posting twenty that differ on one axis produces an answer you can act on. The axis is almost always the opening, because it carries the most variance and is the cheapest thing to change.
The practical rule is to change one thing at a time and to give each variant enough posts to stop being noise. A single post is a coin flip. Five posts in the same shape is a signal. Most accounts declare a winner far too early and then build a quarter of content on a result that was a fluke.
The reason this discipline used to be rare is that every variant had a production cost, so testing was something only large advertisers could afford. When the marginal variant is close to free, testing stops being a luxury and becomes the default way to work.
The common mistake
Changing the hook, the script, the footage and the caption at once. When the winner emerges you will not know which change caused it, which means you cannot repeat it.