I think mere observation does count as a form of non-formalised and occasionally unreliable science. Certainly if you're going to classify "science" and "reason" as the two ways of finding out about the world, then you have to bung "observation" in there with science.

In engineering you often do want to generalize to a whole class of examples, and can hardly help doing so. If you want to nail two pieces of wood together, and it's never been done before, you can't help but discover general principles along the way that are applicable to all wood-nailing scenarios. Then one day you'll go to nail two bits of different wood together, they'll behave differently and heck, you'll have discovered some additional facts about different types of wood.

Eventually you'll hit your finger with the hammer and discover that it really hurts to do so. Now I come to think of it, I would classify "hitting your finger with a hammer really hurts" as a perfect example of those true facts about the world which has nonetheless never been the subject of a proper experiment and certainly wasn't derived from reason, either.