A popular risk analysis technique used for capital investment projects is Monte Carlo simulation of net present value (NPV). Typically, the analyst makes an assumption about probability distributions for parameters in each period over the life of the project and then uses these distributions to generate the probability that NPV is positive. We argued that the conventional approaches to multiperiod uncertainty may be unrealistic for some parameters. Our solution modeled the evolution of a parameter as a Martingale with a shrinking error variance.