The objective of this study was to investigate the effect of composition and size of the reference population in imputation efficiency of INDUSCHIP v2 in Indian HF crossbred cattle. Data set consisted of a total of 869 cattle from 14 Indicine breeds, 2 crossbreds (HF and Jersey crossbreds) and 2 exotic breeds (HF, Jersey) genotyped with Illumina BovineHD (Illumina, San Diego, CA) panel. Post QC, 846 animals and 449955 SNPs remained for imputation study. 3 test groups each with randomly selected 25 HFCB animals with subset genotype of INDUSCHIP v2 were created, whereas with HD genotyping data of remaining animals, 3 different categories of reference groups were created namely reference 1 (HF, Jersey, all 14 Indicine breeds, HF and Jersey crossbreds), reference 2 (HF, HF crossbred, Sahiwal, Gir and Kankrej) and reference 3 (pure HF, Sahiwal, Gir and Kankrej). Imputation efficiency of INDUSCHIP v2 was expressed in terms of concordance rate and Dosage R2 (DR2). Reference groups 1 and 2 were found to be better than Reference group 3. Further, the size of the reference population had an impact on imputation efficiency. The concordance rate and DR2 decreased with decline about population size. However, a reference population with 280 animals was found to be sufficient to obtain a concordance rate of around 95% or more and DR2 around 0.93. More number of HF, HF crossbred, Sahiwal, Gir and Kankrej animals need to be HD genotyped and incorporated in the reference population to improve the imputation efficiency of INDUSCHIP v2.