The present study was carried out using total 63,098 first lactation monthly test-day milk yield (TDMY) records of 7,419 crossbred Holstein Friesian (CBHF) animals sired by 209 sires. Data were collected from Sabarmati Ashram Gaushala (SAG) CBHF Progeny Testing (PT) project under the National Dairy Plan Phase-I for three districts viz. Sabarkantha, Panchmahal and Surat of Gujarat state from Information Network for Animal Productivity & Health - Management Information System developed by the National Dairy Development Board. Additionally, information on 300 animals and owners/ milkers was collected from 82 villages of the same districts under SAG PT project to study the error pattern, a farmer may make, while reporting performance of animal. Three factors were found to have significant effect (P<0.05) on error pattern using Least squares analysis viz. total number of animals with farmer, possibility of milk recording at regular interval by farmer and possibility of milk recording twice a day at monthly interval by farmer. The breeding values of CBHF sires for test-day milk yield were estimated under three different milk recording strategies viz. existing standard milk recording strategy (i.e., morning – evening milk recording at monthly interval by PT project milk recorders:ST1), bimonthly milk recording strategy (i.e., morning – evening milk recording at bimonthly interval:ST2) and milk recording by farmer (morning – evening milk recording at monthly interval:ST3). Data set for ST2 were drawn from ST1 by removal of alternate months, while data set for ST3 was simulated data after imputing error estimated in recording by farmer. Random Regression model (RRM) with Legendre polynomials of order three was fitted on all three data sets. The breeding values were estimated by best linear unbiased prediction (BLUP) method, using variance and covariance components estimated by Average Information Restricted Maximum Likelihood (AIREML) algorithm. The effectiveness of three strategies and decision on optimum strategy for sire evaluation were arrived on the basis of heritability (h2) of test-day milk yield, Log likelihood function (logL), Akaike’s information criterion (AIC), Bayesian information criterion (BIC), Percentage squared bias (PSB) and Spearman rank correlations. The average h2 value for ST1 and ST2 were nearly similar (0.31 and 0.30, respectively) while for ST3, it was almost half (0.16). The logL, AIC and BIC values of ST3 were almost double that for ST1. PSB value for ST3 was more than ten times greater than that by ST1 and ST2. Poor rank correlations were observed between rankings of sires for their 305-day milk yield breeding values (with more than 70 % reliability) estimated by ST3 with those by ST1 and ST2 (0.789 and 0.799, respectively) compared to that between ST2 and ST1 (0.881). ST3 suffered lowered exploitation of genetic variance, biased results and poor prediction of sire genetic merit relative to ST1 and ST2. As compared to ST1, ST2 indicated slight tendency to exploit lower genetic variance and showed variations in ranking of sires on the basis of genetic merit to the tune of around 21%. ST1 is thus recommended as the optimum strategy for milk recording, followed by ST2 which should be considered with caution under conditions wherein deploying ST1 is not feasible, while ST3 strategy is not recommended for large-scale field-based animal breeding projects in India.