The survey research in social sciences, such as, psychology, pedagogy, often contain various dimensions, resulting in excessive questions in one questionnaire. For the consideration of better quality and lower cost, the long, complex questionnaire will split into several small, simple questionnaires depending on the research dimensions. Then, the survey will be conducted by gathering data from different groups of participants with various small, simple questionnaires. After completing the survey, the data interpolating techniques will be used to complete the dataset. But, due to the survey method, the possibility of measuring the correlation between two research dimensions in different small questionnaires are no longer existing. In this article, we propose a new method called 'Two- stage incomplete survey design'. First, the questions in the long, complex questionnaire will split into two small questionnaires, each of them will contain common questions of personal information and a part of questions in long questionnaire. In addition to the different questionnaire dividing method, the survey procedure is also different. The first phase of the survey, we gather data with one of the small questionnaire, named ‘dataset1' and identify several key questions of the small questionnaire from dataset1. The second phase, we add key questions in another small questionnaire, gather data and complete the survey. After the survey, the information of key questions will be used to data interpolation. In this paper, a real world survey case, 'The 2006 World Values Survey', is used to test the performance of the proposed method. We also compare the performance of our method and the method proposed by Liu and Tsai (1997). It shows that the proposed method is better.