A Data Science Central Community
This is a review committee working group for practitioners and academics to establish a formal definition and set of classification criteria regarding business analytics model risk.
This group has been established based upon interest and feedback concerning a recent set of posts regarding business analytics model risk: http://tinyurl.com/ktabt3q
For those interested, you can join the group via this LinkedIn link:
The topic of model risk has rapidly come to the fore as a central concern for large organizations. Growing complexity in business decision making and an increasing reliance on IT-based decision systems, many of which become ‘black boxes’, has raised the stakes concerning model risk.
As broader industries and organizations, beyond banking and finance, are rapidly adopting complex model-based decision making methods, we are concerned with model risk more generally. In particular, the growth of ‘business analytics’ and ‘Big Data’ as structured approaches to complex business decision making has raised the stakes for improving decision model quality.
Model risk is here specified as ‘business analytics (BA) model risk’ to distinguish it from pure finance industry model risk. This recognizes that much of the literature output is focused on model risk for the finance industry, but that the scope of the model risk problem is larger and broader (across all industries) and thus deserves a more general discussion and treatment.
When speaking of model risk, we are referring to organizational decision making in large, complex organizations generally. We are particularly concerned with decision models as encoded into IT systems: business intelligence (BI), decision support systems (DSS), manufacturing control systems, predictive machine learning, etc.
In particular we are concerned here with highly complex ‘analytics’ decision models which become encoded in IT systems (algorithmically or otherwise in terms of automated computational data processing and procedures). This recognizes that large, complex organizational decision making is increasingly automated by IT systems which encode and embed decision models. The term ‘black box’ refers to the tendency for such systems to trap and hide potentially risky assumptions with models.