The work aims to use the SKU merging optimization method of delivered with original package (DWOP) to achieve the purpose of saving material costs by reducing the types of stock keeping units (SKUs). K-means clustering analysis and combinatorial optimization theory were used comprehensively to establish the SKU merging model of DWOP. Afterwards, the genetic algorithm based on non-uniform mutation operator was designed to solve this problem. With the test data set of randomly generated SKU size information and its corresponding order quantity of DWOP as an example, the effectiveness and feasibility of the optimization method were verified by comparing the number of SKU types and the material cost of DWOP before and after merging. The results showed that the average reduction rate of the SKU types in the DWOP SKU merging scheme after optimization was 33.13%, and the average material cost reduction rate was 2.84%. The research results can enrich the related research in the field of DWOP and have guiding significance for optimizing the supply chain structure and promoting the serialization and intelligent development of packaging.