Contextual Complications in Analytical Modeling: When the Problem is Not the Problem

Abstract

In this essay, I describe 10 critical complicating factors that directly affect the six basic modeling components of problem definition, assumptions, decision variables, objective functions, constraints, and solution approach. The proposed 10 contextual complicating factors are (1) organization, (2) decision-making processes, (3) measures and key performance indicators, (4) rational and irrational biases, (5) decision horizon and interval, (6) data availability, accuracy, fidelity, and latency, (7) legacy and other computer systems, (8) organizational and individual risk tolerance, (9) clarity of model and method, and (10) implementability and sustainability of the approach. I hypothesize that the core analytical problem cannot be adequately described or usefully solved without careful consideration of these factors. I describe detailed examples of these contextual factors’ effects on modeling from six published applied prescriptive analytics projects and provide other examples from the literature. The complicating factors are pervasive in these projects, directly and dramatically affecting basic modeling components over half the time. Further, in the presence of these factors, 23 statistically significant correlations tend to form in three clusters, which I characterize as culture, decision, and project clusters. Unrecognized, these factors would have hampered the implementation and ongoing use of these analytical models; in a sense, the models themselves were wrong, absent consideration of these contextual considerations. With these insights, I hope to help practitioners identify the effects of these common complications and avoid project failure by incorporating these contextual factors into their modeling considerations. Future research could seek to better understand these factors and their effects on modeling.