In my experience reporting to management, I have found that decision-makers have a strong preference for speed: speed of initial results and speed in revising. They will opt for "good and fast" over "slow and perfect." If you offer 80 percent accuracy in five minutes over 99 percent accuracy next week, they will always elect 80 percent accuracy. They also want analyses that can be quickly revised to respond to new input from them.
After thoroughly investigating other alternative analytical approaches, I decided to use Monte Carlo Simulation (MCS) to act as a second-tier decision aid. In effect, I used MCS to approximate the results of an agent-based model, placing bounds around the possible results. When looking for a product, I chose @RISK 5.0 by Palisade Inc.
The product provides correlation-driven simulations, data distribution fitting and dashboards of summary statistics on simulations completed.
Due in large part to its graphical abilities, including the ability to overlay multiple distributions, @RISK provides intuitive results for even the nonstatistically proficient. Dozens of variables can be transformed from statistic model parameters to distributed ones, making significant degrees of sensitivity analysis possible. Without @RISK, it would take a significant amount of custom coding in Visual Basic just to replicate the most basic functions. The tool provides an easy to use interface that fits seamlessly into Excel. I have yet to experience any crashes or errors in using the product. Users who require a broad output space, including a full spectrum of potential outcomes will find @RISK to be a very valuable tool.
One word of caution: for the more astute researcher, who may require millions of iterations before feeling comfortable with the results, they may find @RISK to be too "practitioner-oriented." Although there are plenty of example files that come with the package, there is no case study-based tutorial.
Consultant David Lengacher is a Six Sigma Master Black Belt who specializes in measurement performance, simulation and decision support systems.
Be the first to comment on this post using the section below.