Scenario optimization considers decisions made on the grounds of past experience. A scenario decision is characterized by two fundamental elements: one is the performance the decision is expected to achieve and the other is the risk that this expected performance will not be met in a new case at hand. A mathematical theory offers an opportunity to study the trade-off between these two elements and in this presentation we shall try to highlight the principal conclusions of this theory. The significance of these results in the practical use of scenario optimization will also be discussed.