As usual, Debbie Williams is right on target. At a fundamental level, risk is a portfolio concept. As such it requires consolidated information on the full portfolio under consideration. Overcoming the massive data fragmentation that has accompanied migration of computing power from mainframes to minicomputers to PCs over the past 35 years is one of the biggest challenges to achieving effective enterprise risk management. As Debbie points out, flexibiity and transparence is the key. The necessary data elements needed for effective risk calculations will evolve continuously both because of new businesses and new products and because of a desire to perform more complex calculations.

My one caution is to avoid the temptation to attempt to build an all encompassing data warehouse with every scrap of information in the entire organization. That kind of project will become a black hole and will never be updated in a sufficiently timely fashion or be performant enough in the computational sense to provide the expected benefits. A far better approach is to build fit-for-purpose data marts that are designed to meet well defined analytical objectives. This will:

1) Provide a clear framework for what data elements are required.

2) Allow database designs that meet the retrieval needs of he intended analysis in a computationally efficient manner (which will not be the case in a fully normalized - i.e. maximally flexible - table layout.)

3) Provide early evidence of valuable output that will strengthen organizational support for the continuing cost and effort involved in data consolidation.