The objectives of this study were to estimate the genetic parameters and genetic trends for different measures of persistency (PM) for milk yield in Iranian buffaloes using a random regression test day model. Test day records of the first three lactations of buffaloes within 715 herds comprising 43 014, 38 941 and 33 937 records, respectively, were obtained from the Animal Breeding Center of Iran during 1992 to 2012. Legendre polynomial functions were chosen to fit the lactation curves in the framework of a random regression test day model for estimating (co)variance components. Three measures of persistency were modified based on the lactation curve conditions of buffaloes: (1) The average of estimated breeding values (EBVs) for test day milk yields from day 226 to day 270 as a deviation from the average of EBVs from day 44 to day 62 (PM1), (2) A summation of contribution for each day from day 53 to day 247 as a deviation from day 248 (PM2), (3) The difference between EBVs for day 257 and day 80 (PM3). The highest heritability estimates were observed for PM2 over the first three lactations (0·29, 0·80 and 0·42, respectively). Estimates of genetic correlations between persistency measures and milk yield varied from −0·48 to 0·29 (first lactation), from −0·12 to 0·48 (second lactation), and from −0·62 to 0·63 (third lactation). Estimate of genetic trend for first lactation PM2 was negative (−0·0088 ± 0·0043; P < 0·05). It is essential to include milk yield persistency in the breeding goal of Iranian buffaloes along with other important traits such as milk yield or reproductive traits.