We survey a class of decision problems characterized by open-ended gain for recommender systems and how the delay on actions can affect the gain. We believe that these problems differ from the much surveyed delayed action cost problems in terms of the interactivity and the time frame. We conjecture that the estimated cost of delayed action (ECDA) model in cost problems can be used in this scenario. We survey some existing methods on making the model less dependent on the formulation of utility and probability function. We also expose an open problem on determining the right time for a certain action to be executed.