One trick if you, like me, start with the figures and their poses, and work in the buildings later:

Once you've gotten the figures in, and posed, use parenting. Suppose I want to have someone on a motorcycle stopped at a light staring and two people wrestling on the ground while a dog is busy eating the sandwhich one of the wrestlers has dropped. I would get to the point where I had the figures clothed and posed, and textures adjusted. Then I would want to bring in an urban set (such as one of Stonemason's wonderful scenes), and want to move everyone to a convenient corner.

So, here is what I would do. I would parent the driver to the motorcycle (or vice versa), so that I could make adjustments of the two together (by adjusting the parent.) I would parent the sandwhich to the dog, so that if I moved the dog, the sandwhich would move. I would then bring in a primitive, such as the cube, and parent the cycle, dog, and the two wrestlers to the cube.

Next I would load the urban scene in. I could then move the cube around (using the translation dials) and everyone would move together. (I could also scale all of them by scaling the cube.) I could use the rotation on the cube to line them up with the road. I would probably use y-translation on the cube to make sure the wrestlers were on the ground, then on the dog to make sure it was, then on the motorcycle.

That's my workflow, anyway.