Insight Extraction (3E Framework)

Insight Extraction

Operationalization is the heart and soul of conceptual frameworks. There are multiple reasons for that, with perhaps the most important one being the ease of adoption, which is illustrated by some conceptually compelling ideas that nonetheless failed to deliver the anticipated benefits, manifestly because of lack of clear how-to. A case in point is the widely cited (in the field of risk management) COSO’s Enterprise Risk Management framework, the goal of which, as implied in its name, is to offer organizations clear direction and guidance for managing the totality of internal and external threats. And while the framework puts forth compelling arguments in support of the desirability of individual risk level measurement, it offers no appreciable guidance regarding how that is to be done, which ultimately greatly limits the conceptualization’s practical value. Though a glaring deficiency, the lack of clear measurement guidelines has a logical explanation in the form of situational risk invariance, the recognition of which likely compelled the architects of the framework to push the measurement specification responsibility onto users of the framework. When considered from that vantage point, the decision to not include explicit risk measurement details might seem quite reasonable, but since it led to undesirable outcomes in the form of diminished framework’s usability, it was clearly a questionable choice.

As graphically depicted in Figure 7.3, the how-to of insight extraction is a key part of the 3E framework. Building on the earlier discussion of the within-meta-category, -category, and -subcategory communalities, and cross-meta-category, -category, and -subcategory distinctiveness, the goal of optimal measurement approach identification was to pinpoint methodologies capable of capturing the informational essence of evidence comprising the individual groups. In other words, at which of the three levels of evidence abstraction (meta-categories, categories, and subcategories) was an analytically singular measurement approach most appropriate?

The mix of evidence comprising the two broad meta-categories – empirical and experiential – is too diverse to lend itself to a singular analytical approach because while operational data require considerable curation, feature engineering, and analyses, theoretical research studies typically yield post-analysis findings. At the other end of the evidence categorization continuum, some of the multiple subcategories delineated in Figure 7.3 called for analytically non-distinct approaches (e.g., transactional details and descriptive attributes entail the same or comparable data analytic steps), whereas other ones required analytically distinct operationalizations, as was the case with research summaries and industry benchmarks. Ultimately, the four distinct evidence categories of operational data, theoretical research, norms & standards, and expert judgment were deemed to be most analytically cohesive.