A scientist building a house might use an experiment to find what kind of mortar is best for holding one brick to another brick. He must, however, use reason to deduce that a mortar that can hold two bricks together is also capable of holding a thousand bricks together.
Then he can combine it with a bunch of other observations about the world... for instance that mortar which works fine on a Sunday will also work fine on a Monday. How do we know this? Well, because no other material has been observed to change its properties depending on the day of the week, and because there is no logical connection between physical properties of materials and the arbitrary cycling of the calendar which is a human invention. On the other hand, some materials do have different properties in December vs July (eg the water in the lake might completely change state depending on where you live). Anyway, I seem to have gone off on a tangent but basically, observation and reasoning are inextricably intertwined if you want to actually figure out anything about the state of the world.
How does a scientist build a house? Since apparently he can not calculate the statical properties in advance, the only option seems to be to build a random house and then measure if it crumbles down or not? That doesn't seem very effective. In fact, several houses would have to be built to get statistical significance. Then nobody would be allowed to move in for 10 years to be reasonable sure it is stable. As soon as somebody moved in, the conditions of the experiment would change and the house might still topple down. I can't see a way for this to work. If science is like that, it is useless.