Cross breeding is one of the major tools used by breeders for exploiting heterosis to develop useful commercial hybrids. Being most simple and easy to handle, diallel cross hybrids are the most commonly used hybridization method. But, triallel cross hybrids are genetically more stable and consistent in performance than diallel cross hybrids. Here, a method of constructing design for breeding trials involving complete/partial triallel crosses has been developed. This method is based on triangular association scheme. The variance of contrasts pertaining to estimated general combining ability (gca) effects of full parents as well as half parents are of two types in accordance with the association scheme used for its construction, indicating that this design is partially variance balanced. The efficiency factor in terms of information per cross pertaining to general combining ability effects of half parents as well as full parents in comparison to a complete triallel cross plan, assuming the error variance to be same for both the plans, has been computed and is found to be fairly high. The method has been illustrated through a proper example.