To add to this, the differential part is important; it's not as simple as a binary 'bound/not bound' state, since you are normally assessing DNA isolated from a population of cells which can be in various states depending on the experimental design. So, in effect you are assessing whether there is a significant switch from one state or the other based on the experimental conditions, and what you are measuring is an averaging of the states in that condition. This also underscores the need for experimental replicates.

And also, its not only in "promoters", it could be anywhere in the genome, especially enhancers.