Given a multistage stochastic program, its solution (if it exists) consists of the first decision vector, as well as all the recourse decision vectors for all possible scenarios of an event tree. But how do we use these recourse decision vectors in practice? In the beginning, we use the first decision vector (here and now decisions). Then, in the second stage and the following stages, we have real-world information that is revealed, and I have to choose, according to this information, what is the scenario that is the closest to reality and use the precomputed recourse decision vector corresponding to this scenario. Is it the way the recourse decisions are used in practice? If so, how can I determine the scenario that is the closest to the real-world information once it has been revealed?