incurred to achieve such returns. Risk is ultimately a very human concept comprised of many human dimensions (e.g., emotion, psychological response to uncertainty, fear of underperfor-mance, etc.). Since no two human beings are identical, no two risk assessments are identical. To measure risk and return most comprehensively, we have seen that a panoply of tools (e.g., historical simulations, liquidity awareness, Monte Carlo methods, etc.) can be helpful in order to gain the most complete understanding of the risk present in a portfolio. If the tools yield materially different forecasts, the onus is on the risk professional, working together with senior management and portfolio managers, to apply judgment to determine the most appropriate forecast under the circumstances. How to Improve the Meaning!illness of Performance Measurement Tools Performance tools are especially robust when they confirm a priori expectations regarding the quality of returns. If we can identify a disciplined and effective process, we should expect that the process will generate superior risk-adjusted returns. The tools provide a means of measuring the extent of the process's effectiveness. The tools should confirm our belief that the process is indeed functioning the way it was designed to. For example, risk decomposition analysis should show that small cap managers are in fact taking most of their risk in small cap themes. Similarly, a manager with a particular industry specialization should be able to demonstrate that most of that risk budget is spent in securities in that industry. And so on. For a process to be present, one must be able to define "normal behavior." If normalcy is not identified, the process is likely to be too amorphous to be quantified. Simply put, a process cannot exist without well-defined expectations and decision rules. Normal behavior suggests that behavior should be predictable. If a process is effective, continued normal behavior (i.e., trading in a manner consistent with the established process) should give us reason to conclude that high-quality returns observed in the past are likely to replicate themselves in the future. Later on in this chapter, we will introduce some commonly used performance tools. Before discussing these, however, it is worth noting that performance tools, while necessary, are not a substitute for timely management intervention when there is an indication of abnormal behavior. By the time that abnormal behavior manifests itself in the form of poor performance statistics, the damage might al- 13In cases where a portfolio holds illiquid assets, returns are the product of human judgment to some degree. It is conceivable that two individuals looking at the same positions could arrive at materially different valuations-this phenomenon occurs because there can be a material divergence between value and price in illiquid markets. In contrast, for liquid securities, the low bid/ask spread is an indication that price is a good approximation of value.