by Dan Power

Editor, DSSResources.com 

Information systems can be categorized in many ways, but historically business information systems began as tools to record and process transactions. It is still useful 

to distinguish between informational decision support and transaction processing systems. Transaction processing is divided into individual, indivisible operations, called 

transactions. More specifically, a transaction is a discrete unit of work that must be completely processed by a computer system or it fails. For example, entering a customer order is an example of a transaction. Decision support or informational systems summarize and report on transactions.

Continue reading at 

So what did Phillips say in his July 7, 2010 speech in London?

"We've taken the idea of an integrated suite and applied that to business intelligence because there are a lot of ways to extract information out of systems, but you need the information to be consistent."

"We architected this system so you have common metadata across all datasources, all calculation engines, all rules, all workflow." 

If only the architecture would do this.  People introduce the problems.

Supposedly BI 11g "... solves some of the largest problems IT faces, which don't go away just by hoping everyone works together," Phillips said. "They have to be engineered that way."

So now we are engineering away people problems associated with metadata. No chance.

Programmed trading is partly responsible for the extreme market instability of Thursday, May 6, 2010. According to Matt Goldstein, "High-speed trading, which uses sophisticated computer algorithms based on specific scenarios to automate transactions at speeds in the millionths of a second, now accounts for about 60 percent of U.S. equity volume." In one 10 minute span, the Dow Jones average fell 700 points. Some stocks lost 30% or more.

What is happening?  Traders develop rules that are monitored by software and then executed automatically using trading systems.  Let's say I create a rule that "if the price of stock X falls to $20, then sell 10000 shares."  Because the market moves so fast a person does not review the sell order and confirm, rather it executes automatically.  That is fine unless a panic occurs and then the programmed trading magnifies the panic.  What about counter trades?  For example, "if the price of stock X falls to $20, then buy 10000 shares." Market makers are facilitating trades but there is risk aversion on the part of traders that results in stop loss rules that seem to cause and magnify the downward spiral.

Can programmed trading be better regulated? Should it be regulated?  Do traders make errors when they create trading scenarios and develop trading rules? YES. The easiest solution is to require human confirmation of all trades.  Will this solve all the problems? NO.  Some human decision makers will push "OK" without monitoring what is happening in the market and will fail to look at the reasonableness of the trade.  Is requiring human authorization in real time better than automated? YES, a human will be more directly responsible for the consequences and results of the trade. Decision support is better than decision automation.

References

Goldstein, M., "Stock plunge raise alarm on algo trading," Reuters, May 7, 2010, URL http://www.reuters.com/article/idUSTRE64631Y20100507 .