Pradeep Kumar, practice head for “Marketing Mix Modeling” at Draftfcb and colleague of mine, referred me to a very interesting and surprisingly old article about “Decision Calculus” by John Little. “Decision calculus” describes a methodology that combines statistical analysis with expert opinion to solve a particular problem, primarily in marketing. John Little introduced this concept first in 1970 when he tried to develop an approach to get managers more involved in analytically derived decision making. The AMA dictionary defines “Decision calculus” as…:
“…the quantitative models of a process that are calibrated by examining subjective judgments about outcomes of the process (e.g., market share or sales of a firm) under a variety of hypothetical scenarios (e.g., advertising spending level, promotion expenditures). Once the model linking process outcomes to marketing decision variables has been calibrated, it is possible to derive an optimal marketing recommendation”.
Over the last 40 years, these attempts to stronger formalize the expertise and human knowledge of managers into analytically derived decision has had as many supporters as opponents. I think that something is lost in this quite often heated debate. Pradeep Kumar says it best when he says: “Understanding is more important than predicting”.
The underlying genesis of “Decision Calculus” resides in the desire to not necessarily create the most perfect analytical model but to build model based processes and systems that get used more often. Too often statisticians try to build more and more perfect mathematical models that are then used by analytical ignorant managers to the detriment of smart decisions in a fast changing business environment. The better approach might be to invest more time in building potentially less complex models but incorporating the user’s expertise on the path of solving a particular problem. The question of how much do I need to invest in marketing programs for a particular new product launch could be dealt with purely academically and mathematically by using the most sophisticated analytics. Or it can combine analytically derived insights with the heuristically based and incorporated insights of the expert and ultimately budget responsible marketing manager.
Over the last years data visualization has been the key application to “democratize” data. Now, we might be at the point of “democratizing” more complex statistical models. Visualization only get us so far, since marketers can use dashboards and visualized models without understanding the real foundation and strong limitations that any analytical derived model has. The concept of “Decision Calculus” continues to show tremendous promise, even 40 years after its birth. We will need to continue utilizing and expanding it in real life situations to improve its impact and define better rules for its successful application.