Purpose -The purpose of this paper is to propose a method to compute RAMD indices to measure and improve the performance of skim milk powder production system of a dairy plant under real working conditions. Design/methodology/approach -The present work is carried out by developing performance model based on Markov birth-death process. The skim milk powder production system consists of six units. The first order governing differential equations are derived using the mnemonic rule and further solved to calculate RAMD indices i.e. reliability, availability, maintainability, dependability, MTBF, MTTR and dependability ratio for each subsystem of the system. Findings -The subsystem SS1 comprising of chiller and cream separator is the most critical from maintenance point of view, as the reliability, availability, maintainability, dependability, MTBF and dependability ratio indices are low as compared to those of other subsystems of skim milk powder production system of the dairy plant. Originality/value -The RAMD indices of the present work is very useful for finding the critical subsystem and its effect on the performance of the system working under real working conditions. Further, based on findings the maintenance priorities for various subsystems can be decided.