Kinda late to the party on this one, but I thought that this was an interesting topic for discussion
For me, it's probably a mix, as many of the others have also stated. If I'm looking for a strength program, I want to look at what the big and strong guys are doing. I'll evaluate the programs scientifically if I can (similarities between programs, scientific basis for a technique, etc, etc) and try to "make sense" of what the rest of the field is doing. If I'm looking at a sport-specific program, I'll again look at what other coaches are doing, evaluate the programs and techniques scientifically to see if they're logical and adhere to basic science and clinical research if available, and go from there.
The overall planning and periodization is a mix: I use a lot of anecdotal evidence and experience with my periodization, although the basis of periodized programs has been evaluated and for the most part justified in the literature.
With exercise selection itself, I'm far more scientific: I use biomechanics and kinesiology extensively in my programming, looking at joint mechanics, positioning and postural evals, length tension relationships and force production potential, sport-specific analysis if applicable, functional roles, etc, etc. Science doesn't have the answer to everything, though, so even here certain exercises will be selected based on experience and practical application, especially in certain areas of rehab/prehab where the research is limited.
Overall, like the others have said, if something works, I use it, and if something doesn't work, I drop it. That may or may not be based in scientific literature or not. Science may also lead me in new directions in my programming, or verify something that I already suspected, too. The fact is, i would consider any evaluation of a technique by scientific means to be 'a role of science' in a program, and so in that sense it probably counts for a great portion of what I do on any given day.
But the bottom line is that my clients are paying money for results, period. Science can help this, but at the end of the day it still boils down to practical results.