Current approaches to managing an organizations' portfolio of predictive analytical models are currently ad hoc and under-informed by theory. In time-sensitive campaigns, analytical models may be constructed, applied and maintained by departments or individuals working in isolation. This silo-based approach fails to take into account complex model inter-relationships, possible model correlation and covariance, and other types of interdependencies that can confound and limit the effectiveness of campaigns. In this paper, we view a suite of predictive analytical models from a portfolio theoretic-based vantage point, we state the nature of a multi-stage campaign management bottleneck effect in portfolio terms, and we establish cases where informing campaign managers about the state of portfolio under-performance can be expressed and tracked. For the most complex case, we develop a Cross-stage Dispersion Index (CDI) that can be used as a benchmark and tracked over time.