“Why do things look as they do?” (Koffka, 1935, p.75). The law of Prägnanz gives a hypothetical, abstract explanation: “[…] psychological organization will always be as “good” as the prevailing conditions allow” (Koffka, 1935, p.110). This also suggests that we will perceive a certain organization of the (visual) input because it is the ‘best’ organization possible. The law of Prägnanz has often been criticized for being circularly defined (e.g., Neisser, 2002): what we see will be as good as possible, and what is good is what we see. We argue that this circularity, which impedes a proper causal explanation, can only be avoided ( a ) by providing a clearer definition of goodness or Prägnanz; and ( b ) by providing a way to measure, predict, or manipulate Prägnanz. We propose a refined definition of the law of Prägnanz, in which Prägnanz or goodness is defined as process and outcome efficiency. In other words, we will perceive a certain organization of the input because that organization is easiest to achieve (i.e., process efficient) and most useful (i.e., outcome efficient). Often there is a trade-off between the two: the organization that is most useful is almost never the one demanding the least resources. The ‘winning’ organization will be the one for which the combination of process and outcome efficiency is maximal under the given conditions. A solution for the second problem, namely providing a way to measure, predict, or manipulate Prägnanz, is in the making. Currently, we base our manipulations of Prägnanz on a hierarchical predictive coding framework. This framework includes hypotheses about ( a ) effects of the ‘prevailing conditions’ (i.e., input, individual, context) on Prägnanz (i.e., process and outcome efficiency); ( b ) effects of Prägnanz on which organization will be perceived; and ( c ) effects of Prägnanz on other psychological processes like categorization, discrimination, recognition memory, or aesthetic appreciation. How to test the hypothesized causal relationships within this framework? As the causal agent in ( a ) is ever-changing (i.e., the combination of individual, visual input, and context), exact replications are nearly impossible. As the causal agent in ( b ) and ( c ) is still not exactly quantitatively defined yet (i.e., Prägnanz or goodness), investigating this second type of causal relations is also challenging, and dependent on ( a ). Some suggestions concerning how to empirically test the Prägnanz framework of perception will be given, including experiments on effects of functional relevance and working memory capacity on the perceived organization as well as experiments on the effects of Prägnanz on categorization, discrimination, recognition memory, and aesthetic appreciation.