Various standardized milk yield prediction methods have been developed and used. The objective of this study was to compare two methods for the estimation of 305-day milk yield inthe Holstein breed, in terms of breeding values and their accuracy. Genetic evaluations of milk yield were compared using: 1) adjusted total lactation yield for days in milk, month, and age at calving (adjusted TY305) or 2) adjusted305-day milk yieldestimated by fitting test-day(TD) records to the Wood model(adjusted WY305).The method with better ability to predict standardized milk yield was used to identify a Tunisian cow reference population toward genomic evaluation of milk trait. Three datasets were used. The first data contains 380’709 TD records corresponding to 34’281 three first lactations of 20’758 cows collected between 2008 and 2018 in 33 herds. The second dataset contains 11’175 total first three lactation yields recorded between 2012 and 2017 from 6251 cows belonging to 33 herds.The third data is a pedigree file of 27’487 males and females. The predictive ability of the two methods was assessed by Spearman’s rank correlation between predicted breeding values for 305-day milk yield (PBV305) from the full dataset and reduced dataset in which the records from the last calving year were masked. The two methods were compared in terms of rank correlation between PBV305 and the percentage of selected animals in common when different selection intensities were applied based on PBV305.The average gain in accuracy was calculated and a Tunisian reference population was identified. The results showed that heritability estimates were 0.11 (±0.02) and 0.13 (±0.01) for adjusted WY305 and TY305, respectively. The highest correlation for PBVs between full data and reduced data was achieved in TY305 dataset. Rank correlations between PBV305 estimated for adjusted WY305 and TY305 were 0.67. The percentage of animals selected in common was 11% or 21%, respectively, when 1 or 5% of cows were chosen as future dams of bulls, according to PBVs. An average gain in accuracy of 15% was observed for cows when using adjusted TY305 to estimate PBVs for milk yield trait. The obtained results showed that adjustments applied to the total milk yield records could be appropriate for 305-day milk yield prediction and genetic evaluation of milk production in the Tunisian Holstein population. Based on two main designs (extreme yield and top accuracy), a total of 1000 cows were selected to form the Tunisian female reference population using adjusted TY305 records.