To understand how “best practice” is used productively, we need to step back and realize that all problem domains are not alike (analysis stolen from Cynefin):

Chaotic problems have no clear link between action and effect. Best practices are expected to have predictable effects, so they fail when applied to chaotic problems. Better to act first and analyze later. With a chaotic problem, almost any action will produce useful feedback.
Complex problems display a connection between action and effect, but only in retrospect. Applying a best practice will not get you what you expect. Instead, your best strategy is to invent an action based on first principles (like “move fast and break things” or “make decisions reversible”), and observe the results.
Complicated problems have a, well, complicated connection between action and effect, but at least they are predictable. A best practice is not effective, but choosing among a pallette of known-good practices is.
Simple problems have a clear, linear connection between action and effect. When solving simple problems, effort used to invent or choose between actions is a waste of time. Applying the best practice is efficient.