Bottom-up or top-down understanding?

The past few days I've been modeling the dynamic light scattering (DLS) signal from particles embedded in a complex fluid. The following question keeps floating around in my mind, and I don't quite have a good answer to it.

Is it better to first find an empirical model to explain an observed phenomenon and then infer the physical, microscopic reasons for why the model works? Or is it better to first understand the microscopic makeup of a system to arrive at an empirical model that explains the data?

I'm inclined to think that establishing an empirical model first is better. We often have little-to-no idea of the microscopic details of a system. However, empirical models often apply to systems with similar microscopic behaviors, so once the correct model is found, we can establish a stronger intuition for the details.