Overdoing IT

There are many ways Information Technology has been misused, it is as prone to fads and foibles as any other human endeavor. One of the chief temptations lies in over-reliance on quantitative approaches. It is only natural that IT started in the business world with the idea of automating bookkeeping - at that time it usually resorted under a CFO. Then came the time that it became fashionable to create a new "C" level position, the CIO, not to mention CTOs and CISOs, all under the banner that businesses began to understand that IT was indeed strategic to their success.

And presently we are coming full circle again, with Sarbanes-Oxley causing many CFOs to try to reassert their control, on the ostensibly logical grounds that they're the ones that go to jail if something goes wrong. Of course it is an illusion that those issues are resolved in any way by demoting or altogehter removing the CIO and putting IT back under the CFO again. In fact, logically that is the same error as thinking that your computers are your IT system, or that security exists by hosting your systems on-site. The illusion of ownership of the physical assets (or people), then masquearades to create the illusion of control over the system (or their work). One way or another there will be adjustments. Some will stick, some won't.

Personally, my experience included getting involved in IT from a base in business management, and taking various roles in IT development. I came to appreciate very quickly how afraid people are of the finite and quantitative aspects of IT, or conversely prone to overuse them once they think they understand them. The following anecdote may serve:

Sometime in the mid 80's there was another oil-price spike, and I had the good fortune of being in charge of corporate planning for a shipping company, and report to a president who was a self-proclaimed "oilman," having come to the job from an oil major. He knew just what to do: hedge our oil price risk, and yours truly was charged with implementing the program, the details of which would need board approval. Naturally also the job included confirmation and official endorsement of the "knowledge," and "expertise" of the "oilman" in charge.

After going to work with a reputable oil industry consulting firm, and obtaining the necessary pricing data, I quickly proved that there was no futures market at the time where we could hedge marine fuel price risks, because the kinds of oil and derivatives that were traded were not statistically correlated to marine fuel and/or diesel. Consequently I advised the President and also the CFO that unfortunately the hedging program (we had already taken positions), was a speculative venture, which increased business risk and consequently would have to be disclosed on the books as an incremental business risk, outside of our main line of business. Naturally this outcome was not acceptable to our "oilman," and he was not going to take it lying down.

I was immediately charged with bringing in some Wall Street talent with proposals for risk management. One of the parties who came through for us and made a presentation to our board, offering a feasible solution to manage this significant finacial exposure of our firm was Merill-Lynch, who showed up complete with powerpoints (in those days I think it was still Lotus 1-2-3), and with a Ph.D. in Mathematics in tow. Needless to say, they delivered an impressive presentation, and a clear solution, which was endorsed by the board, subject to final review by yours truly since I had expressed that I could not reproduce the results they presented, even though they were ostensibly based on the same data series I had used.

Off to the McGraw-Hill bookstore in New York for me, where I bought myself a stack of literature on time-series analysis and options pricing, and locked myself into a room with my own copy of Lotus 1-2-3, until I could reproduce the results of the Ph.D. I found it quick enough. He had dazzled our board with his higher math, by applying a Durbin-Watson correction to two time series that were clearly unrelated, and all the textbooks point out that it should only be used to filter out minor disturbances if there is no doubt otherwise that two time series are correlated. I called the Ph.D. and told him my findings, observing that in fact he had done the equivalent of prescribing acne medication for a case of advanced skin-cancer. The next day the proposal was withdrawn by Merill-Lynch. I would not doubt they, and everyone else on Wall Street did these things routinely, and banked on the fact that nearly no-one checked the math. About a year later Procter & Gamble won a settlement of ca $150 million against ChaseManhattan for being sold inappropriate risk management solutions, which were found on closer inspection to INCREASE business risk, and in fact caused serious losses. I suspect the brother of above Ph.D. probably worked for Chase Manhattan...

The above is one of the many examples where the temptation is to apply quantitative methods because it is easy, and not because it is right. War stories of this nature are endless, the infamous failure of LTCM being merely the biggest one to date, for which we're still waiting for the sequel to come out. Warren Buffet among others has pointed out it is only a matter of time, for these "risk management" instruments are so little understood, it's scary. LTCM essentially went wrong because the continued accuracy of some of the underlying assumptions in their models were never scrutinized, so while the world changed, the fancy mathematics did not keep pace. The result was massive failure and a near melt-down of the global financial system.

It is fair to say that overreliance on quantitative approaches is a very common problem and is perhaps far more dangerous, and potentially costly than the under-use of appropriate IT solutions in other areas. Yet there is a lot of investment here, because it appears to yield "easy money," and it pays film star salaries to a lot of "quants." At this point it is definitely in vogue, and so boards feel important voting for these types of solutions. I suspect that the lesson is to focus on the basics, and the KISS principle in order to deliver solutions that work, for real business problems, not made-up ones that are a distraction. Of course there are many areas where quantitative modeling can be extremely helpful, not to mention the fact that risk-management done well can really work too, but overreliance is a risk whenever the underlying assumptions of such methods are not continually verified, and/or the models are operated by people who do not really understand the process well enough to see if something is out of whack. Too often the reality is that with the illusion of a solution a problem merely goes out of mind for a while, and then comes back to haunt us later.


Copyright © 2005 Rogier F. van Vlissingen. All rights reserved.