● Identify root causes.
● Test and confirm using simulations of the effect of the root cause.
● Develop recommendations for fixes as part of kaizen.
● Revert to previous stages to reresearch, retest, and reimplement the fixed trading
solution.
● Begin again Stage 4 to manage the fixed trading/investment system.29.1. Failure Mode and Effects Analysis
Failure mode and effects analysis organizes brainstorming, the first activity toward finding
root causes, by systematically identifying potential points of failure in a trading/investment
system. Failure modes are the ways in which a trading/investment system has or might fail
in the future. Effects analysis refers to studying the consequences of those failures. Failures
are prioritized according to the severity of their consequences, how frequently they occur
(or may occur), and how risk managers may detect them. Furthermore, effects analysis
should document current knowledge of and actions taken to mitigate risks. Which is to
say, the product can use effects analysis to guide strategy design in Stage 1. In this stage it
is used for process control, before and during ongoing operation of the trading/investment
system. Optimally, though, it should begin during the earliest stages of research, the fuzzy
front end, and continue throughout the trading/investment system life cycle.29.2. The Five Whys
While simple in its brainstorming approach, Sakichi Toyoda ’ s Five Whys technique (later
used at Toyota) asks questions to uncover root cause and effect relationships between inputs
and nonconforming outputs. By asking “ Why? ” five times or more, risk managers dig past
symptoms of nonconformance to root causes. Answers to “ Why? ” questions lead to other
problems and further questions. Sometimes, however, risk managers may only be able to
answer questions with existing knowledge, when new knowledge may hold the true solution.
Asking the right questions is not always obvious; different people may arrive at different root
causes for the same problem. Each cause or contribution of the causes needs to be tested until
the problem is explained 100%. Nevertheless, as a first pass toward determining root causes,
the Five Whys method can be a powerful technique to guide more statistical analysis.29.3. Fishbone Diagram
Kaoru Ishikawa ’ s cause and effect diagram, usually now called a fishbone diagram, helps
identify special causes of variation. A fishbone diagram provides a framework for adding
and subtracting different inputs and analyzing the effect of inputs on your model ’ s out-
puts. The results in this case are returns, standard deviation, Sharpe ratio, Sortino ratio, as
well as other performance metrics. Fishbone diagrams work by testing the input variables
on the branches and the outputs in the result area, then using a blending algorithm to
modify the inputs and record the outputs.264 CHAPTER ◆ 2 9 Determine Causes of Variation