As an economic and market‐transparent program, weather index insurance is expected to mitigate asymmetric problem. Capturing the relationship between yield and weather factor(s) is the basis of index insurance, but remains a challenge for weather index schemes. Meanwhile, composite weather index insurance is needed by farmers when their agricultural activities involve several risks, but is rarely studied. We aim to design a composite weather index insurance model and evaluate its efficiency in hedging yield risk by using the case of rice production in China. We divide the whole growth cycle of rice into six stages on the basis of agronomic knowledge, and use the average value of each weather factor in each stage to design a weather index. Then, the efficiency of composite weather index insurance is evaluated by mean‐semivariance and value‐at‐risk methods. First, we find that subdivision of the growth cycle helps to better capture the subtle relationship between rice yield and weather factors. Second, composite weather index insurance evidently reduces yield risk. Our findings help further adoption of weather index insurance in agricultural fields.