Emotions have a significant impact on our decision-making, learning, awareness, social interactions, and mental and physical health. Even though large efforts have been put into quantifying human emotions, their subjectivity, context-dependence, and complexity render them as being almost unpredictable. However, while different streams of research in pervasive computing and psychology have made significant progress in the quantification of emotions, the most successful research results come out of controlled laboratory studies. In this article, we present a retrospective of a series of in-the-wild studies through the lens of human emotions. We are looking at the weaknesses and strengths of traditional research methods and present lessons learned. We furthermore call for a readjustment of research rigor and describe potential new research designs specific to out-of-the-lab studies.