For example, an upscale restaurant, mid-market clothing store, hardware store, and general store could all be sampled, and over the course of four days shoppers could be asked to respond to a survey at the check-out counter about their perceptions of the store and shopping experience. The data could then be analyzed demographically, in terms of consumer perceptions, and also precisely what the different shoppers bought on different days, as opposed to what they said they would buy. This would also allow repeated testing -- using the same four-day musical format over different periods of the year at different stores. The article's study takes the form of a posttest only study design: the independent variable of the music (or in the case of the control group, the lack thereof) is applied, and then subjects record their responses.

In other words, the same subjects, forced to eat in a controlled environment with different types of music playing in the background might easily 'pick up' on the fact that their perception of the situation, depending on the music being played, was being manipulated in a very controlled fashion and they were expected to have specific responses. However, other studies would seem more optimal in design, given the stated goal of the researchers.
Reference

North, Adrian C. & David J. Hargreaves. (1998). The effect of music on atmosphere and purchase intentions in a cafeteria. Journal of Applied Social Psychology. 28 (24):