
Over the past couple of weeks, I have had opportunities to sit down and talk with risk management professionals from two perspectives; analysis and data. From the analytical perspective, I spoke with risk management practitioners and, on the data side, it was data warehousing experts. The question I posed was given the events of the last 18 months, what is the one issue that confronts you today in evolving your risk practices?
Without variation (which is risk defined), the answer was the quality and organization of data. Terms like rogue databases and single versions of the truth were bounced around. Sounds like an episode of the hit show "24" or perhaps more appropriately, "Mission Impossible". Risk professionals stated that while the cost of storage has come down and while the speed of computing has increased (along with the talent and knowledge of the risk community from a quantitative standpoint), the core issue remains the quality of the underlying data and consistent definitions of what that data means. An example in financial services would be the definition of what is a balance? In many institutions, a balance may have 2 or 3 or 4 definitions that various constituents not only swear by, but base exposure and risk calculations on. If we had a common truth across financial services organizations, it would enable technology driven risk solutions to be implemented more quickly and with more meaningful results over time.

Comments
Of course it's all about the data. When are we going to truly understand that data content is the primary factor of input into all business processes and required for all analytical processes. The problem is that the financial industry operates in multiple silos. Each acquires its own data (from multiple sources) and stores it using proprietary or inconsistent semantics (terms and definitions). Not surprising - vendors source data independently, transform it, rename it and stuff it into legacy technology environments. Then firms again source it business unit by business unit, transform it, rename it and stuff it into their environments. No wonder it's not comparable when we move from silo to linked applications.
If we are serious about systemic analysis (which means the ability to compare across multiple functions and companies according to many scenarios) then the data must be comparable. Every business that manages complexity (except the financial industry) understands that the first order of business is to get the language straight. And the language of the financial industry is based on "semantic tags and unique identifiers." Kinda boring (I know) but it is the foundation of being able to do analysis.