I agree with Erik completely. Design and consideration of the parameters that go into the design are the only way to ensure that the resulting product will have a chance to demonstate its reliability!!

I nearly disagree with Donald's statement of "Designing carefully does not prove it will work" because you are not designing carefully if you do not ensure correctness at every step. I say only "nearly disagree" because I do admit it is possible to make a contraption that "works" correctly within its design parameters but may fail for other reasons outside of that set of parameters. For example, a vented brake rotor may work great from new through up to 0.030" of wear and resurfacing, but may be of dubious use after exposure to 10 Minnesota winters of salt covered metro freeways. Another example could be an aluminum grain shovel, which works wonderful in the grainery for shoveling oats, but used to shovel snow on a concrete driveway and the product fails miserably!! (And the reasons why may not be obvious to all readers here! I can elaborate if necessary).

"Simple observation of an arbitrary entity without an understanding its constituent parts cannot be used to predict its behavior or its ability to reliably repeat previously observed behavior".

Michael Karas