The research goal is to develop an accurate learning evaluation instrument in Social science (IPS) subject in elementary school. The method used is Research and Development, following the 4 main steps of the Richey RDPE model, namely: research, design, production and evaluation. Field trials were carried out 2 times, with the number of respondents increasing each stage. Analysis of item characteristics uses classical test theory. The results of the development are in the form of achievement tests for Social science subjects at the fifth grade of elementary school, consisting of 35 multiple choice items and 5 essay items. The test was developed by following the standards for writing a proper test, all procedures taken have been approved by experts from a qualitative perspective. Item analysis is performed for multiple choice tests with 94.3% of questions have moderate difficulty level, the range of index values is 0.53-0.67. 80% of the items have good differentiation, the average index value is 0.30-0.50. Optimization of teacher-made achievement tests produces more qualified test kits in performing measurement functions, as evidenced by the scores obtained by students which accurately describe their real abilities during the learning process