When Marilyn Manson sang of “The Beautiful People,” I think it was about how symmetry of facial features makes a person more visually attractive. At least that was the message I got out of the song. But unfortunately, even facial symmetry cannot compensate for an insufferable personality, as both Marilyn and I have learned.
Like its role in beauty, symmetry plays an important role in forecasting, too. What are we to do, for example, when the cost of forecasting errors is not symmetric – as when the impact of underforecasting costs more than overforecasting? Should we purposely bias our forecast in one direction or another, to account for the asymmetric costs? I would argue no.
The forecast should still represent an unbiased best guess of what is really going to happen. If the cost of forecast error is not symmetric, you would take this into consideration in your planning process. For example, if there is a severe penalty for failing to ship an order, then you might authorize carrying extra inventory. Or, if inventory carrying costs are high, there is risk of product obsolescence, and customers can readily find substitutes if you go out of stock on a particular item, then you may plan to carry low inventory. In either situation, you still forecast what you really think is going to happen—what you are really going to ship or sell—but you compensate for asymmetric error costs in your planning decisions.
New Blogs
Wanted to bring your attention to a couple of new analytics - related blogs:
Data Mining World - Run by data analysis professional Burcu Kalender, this blog is intended to provide educational content on data mining through free internet resources. In today's (September 3, 2010) posting, Burcu gives a callout to Analytics magazine (register for free online subscription), and provides a nice summary of the recent article "Worst Practices in Business Forecasting" by me and my colleague Udo Sglavo.
The Do Loop: Statistical Programming with SAS/IML Software - by my colleague Rick Wicklin, author of the new book Statistical Programming with SAS/IML Software. Per his initial posting:
I will present tips and techniques for writing efficient SAS/IML programs for data analysis, modeling, simulation, sampling, matrix computations, regression, and data visualization. ...I will try to make the posts appeal to a wide range of SAS programmers by increasing the difficulty through the week.
•Mondays will feature introductory ideas and tips for beginners.
•Wednesday posts will discuss intermediate techniques and statistical ideas such as modeling and simulation.
•On Fridays, I’ll address topics for experienced SAS/IML programmers and data analysts, such as data visualization techniques that are implemented in SAS/IML Studio.
On most days, I’ll post programs. This is a blog for programmers, by a programmer, so that programming shall not perish from the earth.
