A Method for Quickly Establishing Personas

The use of personas can help teams better understand the characteristics of users, which leads to more accurately discovery the problems and real pain points that users face. At present, there are two main ways to establish personas. One is to generate personas qualitatively or quantitatively through interviews, questionnaires, etc. These processes are related to the experiences of analysts and the statistical methods used, usually resulting in different conclusions and spending much time. The other is that the technical teams directly obtain the users’ operation data on the products and use algorithm models to automatically generate personas. But this method is only suitable for mature products or existing functions, while the questionnaire method has nothing to do with mature products and functions. In this paper, we present persona segmentation through K-Means and PAM clustering algorithms in machine learning for questionnaire data, including mixed data, as an objective, quick, low-cost method for establishing personas. The method consists of four steps: first, design questionnaire. Second, transform the multi variables caused by multiple choice questions into a single variable. K-Means clustering algorithm is used for the continuous data of multi variables. The rule-based clustering method is used for the classified data. Then, cluster the processed data by PAM. The fourth step is to create personas, which are labeled in this paper. In the end, we demonstrate that the method is appropriate to create useful personas by machine evaluation and expert evaluation.