Mental stress has become a prominent contemporary public health concern, exerting profound influences on both physical and mental well-being. Evaluating the psychological and physiological states under stress scenarios poses a challenging task. Detecting mental stress states is of utmost importance. In this study, we present a characterization approach to quantify the complexity of electrodermal activity during mental stress scenarios during a midterm exam to a group of university students, and use it to predict students' performance. We measure the complexity of wearable skin conductance signals using permutation entropy. The results enable discrimination of signal complexity into three groups based on exam grades: High, Mid, Low. These findings offer valuable insights for advancing the characterization of wearable signals to predict stressors in various real-world scenarios.