Buried in The Economist's report on securitisation last week (February 16 issue) was the mention of investors clamoring for more information, in order to better understand financial instruments. The report's authors noted: "Yet reams of information already accompany mortgage-backed securities sold in public markets. [...] So some interpret calls for greater disclosure as whimpering by investors who did not do their homework." That reminded me of a Buttonwood column that I had set aside over the summer (Too much information, July 14, 2007) - about information vs noise in a news-intensive world and how to distinguish one from the other. A study by an investment bank even "found that the average forecasting error [on analysts' forecasts of company profits] was 43% over 12 months and 95% over two years." Striking is also the fact that "most of the statistics are revised in subsequent weeks but the revisions rarely have as much market impact as the original figures." Doesn't that make you think of headless chicken?
The column also suggests that increasing the amount of data available can decrease the quality of the decisions (a new type of Braess' paradox, maybe), and provides a proxy that will not surprise anyone but for which I did not expect to find any hard evidence: "An academic study has found that American mutual fund managers are more likely to weigh their portfolios in favor of shares in companies in which one of the senior officers went to the same university as they did. The effect is strongest when they were there at the same time and on the same course." (And how am I going to convince my students to pay attention in class rather than making new friends now?)
Given the risk of information overload, I am curious to see how promising innovations such as RFID will evolve - in particular whether they will avoid overwhelming managers with data. Successful data-driven management will help businesspeople make superior decisions by giving them a fuller picture of customer behavior (for instance identifying quickly which product is performing above expectations), but making it successful might not be a trivial matter. In the end, the answer might be to outsource data analysis to firms that specialize in it. This has indeed become a booming business, in particular in fields such as supermarket retail that don't sound cutting-edge but have those dear membership cards that keep track of everything you buy.
According to a December 2007 article (Watching as you shop), "Big shops are using elaborate technology to monitor and influence the behavior of their customers." In a project called PRISM, "sensors recorded data on customer-traffic patterns" for Coca-Cola, Procter & Gamble and Wal-Mart, and in yet another trial, a system called BehaviorIQ was "used by retailers to gather data on where their customers go, where and how long they stop, and how they react to different products." Some stores even record the conversations of their shop assistants to track "employee performance and customer behavior." The companies analyzing all this data insist that their algorithms work wonders, and maybe they do. (After all, they just have to work a bit better than the store's current strategy to yield substantial profits.) If so, the true future of super-computing might not lie in the finance industry after all, and data-driven management has beautiful days ahead.