Abstract We revisit the forward algorithm, developed by Irle, to characterize both value function and stopping set for a large class of optimal problems on continuous-time Markov chains. Our objective is renew interest in this constructive method showing its usefulness solving some constrained that have emerged recently.