As storage costs decrease and user interface technology improves, many companies are launching business management dashboards to summarize key elements of a business for decision makers. Dashboards can be hugely effective in creating a common understanding of important issues and measures within a company. However, when supporting difficult decision making in complex environments, the wrong dashboard can be dangerous.

The problem is this: different pieces of information have different levels of value in the decision-making process. Before dashboard data was available, we were forced to rely on imperfect memory, judgment, and intuition for complex decisions. As described in Jonah Lehrer's How We Decide (Houghton Mifflin, 2009), a significant part of the brain is devoted to doing a good job with this kind of unconscious thought: "Dopamine neurons automatically detect the subtle patterns that we would otherwise fail to notice; they assimilate all the data that we can't consciously comprehend." In contrast to this unconscious brilliance, the rational part of the brain is incredibly limited in its processing power - Lehrer likens the difference to a modern computer versus a hand calculator. Study after study has shown that simple tasks interfere with rational processing in a significant way. For example, in a Stanford study, students asked to remember 7 digits made different choices about eating chocolate cake versus a bowl of fruit. In a classic 1974 experiment, Daniel Kahneman showed that a random number displayed on a roulette wheel influenced guesses about the percentage of African countries in the United Nations.

This second example illustrates a phenomenon known as the "anchoring effect", where irrelevant numbers influence decision making. As summarized by Lehrer, "The prefrontal cortex [the center of rational, conscious thought], it turns out, is easy to hoodwink. All it takes is a few additional digits or a slightly bigger candy scoop, and this rational brain region will start making irrational decisions."

The same is true about irrelevant information presented in a dashboard: we suffer from the illusion that more information is always better, and that a decision made with systematic attention to the maximum amount of data will produce the best business results. In many cases, especially when making difficult, collaborative decisions in complex and uncertain environments, this is wrong.

So what is to be done? The key lies in understanding how the human brain does its pattern recognition, and by so doing to enhance this ability with the right, natural, user interface display. As my graduate student Christy Medina and I wrote about in 1995, the computational mechanism used by the brain during pattern recognition can be mapped to the same approach used in modern decision trees, and indeed in many statistical methods. The brain's technique is simple: from all of the information presented to it, it determines the most relevant attributes of its environment (using, we believe, some of the same information-theoretic mechanisms we use to compress videos today), devoting extra synaptic weight to those neurons that code for the attributes that make the biggest impact on a decision. So if, for instance, you feel better when your factory workers are noisy than when they are quiet, you may have learned a correlation between their volume level and productivity, perhaps without even being able to explain why.

There's a rub, however. First, too much conscious thought interferes with this innate ability. Part of the reason: we are limited in the amount of information we can retain in the conscious decision-making part of the brain.

This is exactly the situation that dashboards are supposed to overcome: by offloading the brain from having to remember so many numbers, the idea is that it can think more clearly. Importantly, though, if the wrong numbers are displayed, they can significantly impede processing.

Fortunately, there is an answer. First, dashboards must do what the brain and what decision trees do: analyze the data to determine what elements are the most relevant to the specific decision being supported, and emphasize these attributes. Realize that other information, even if made available through some drill-down mechanism, may actually interfere with effective prediction (through a phenomenon well-known to statisticians and neural network people alike as "overfitting"). There are some well-accepted algorithms for performing this relevance calculation (see my 1993 paper Discriminability-Based Transfer between Neural Networks for one).

However, in many situations there is not enough historical data to make this relevance determination (this is Taleb's "Black Swan", discussed in his 2007 book by that name), so a second component becomes critical: maximize the visual nature of the display, specifically the degree to which the display supports the brain's natural approach to decision making. To support expanding the capacity of the rational thought process, show the causal structure that creates the data: causation is our natural way of thinking about complex systems. To support our ability to build better unconscious judgments, use a rich visual display that effectively trains the brain to improve its pattern recognition abilities within complex systems, and to overcome its natural irrationalities for the rational part of the analysis, such as ignoring the long- for the short-term, and weighting losses higher than equivalent gains. We have recently updated the Quantellia web site and produced a YouTube video of our software that uses this approach. See both on the software page of our web site, at http://www.quantellia.com/WMSoftware.aspx .