Tom Davenport makes an excellent point in his most recent post about the challenges of decision making:

We have lost much of the connection between the supply of information and the demand for it in decision-making. Despite the fact that companies often justify IT projects on the basis of better decisions, there is seldom a direct tie between the information a particular system produces and the decisions that are supposed to be based on it.

Tom then outlines three common reasons for the disconnect between supply and demand:

1. Systems implemented at the enterprise level are focused on transactions, not decisions (and the potential for making decisions based on transactional information is often ignored)
2. Managers don’t know all the information and knowledge that is available to help make a decision (with the potential for poor decision making as a result)
3. Managers aren’t held accountable for how they make their decisions (and as a result, organisations don't know what information is being used to make those decisions)

When I read this, I was reminded of the contrast between standard business decision making and and scientific experimental design. Scientific experiments generally follow a fairly well-established process, such as:

1. Reading available literature
2. Deciding on an experimental design
3. Capturing data
4. Writing up a report which justifies the conclusions with empirical data and with cross-reference to past experiments

The validity of an experiment is judged according to how well these processes stand up to scrutiny. Yet business decision making ignores all of these steps. All too often decisions are made without reference to past decisions (experiments) or empirical data (transactions).

Why? Normally, for just one simple reason: time.

Experiments are slow and deliberate processes. But business decisions often need to be made quickly. The tenet of "an imperfect decision being made quickly is better than no decision being made at all" is often invoked. But can it be justified?

In recent years we have repeatedly seen that businesses that exercise poor control over decision-making expose themselves to catastrophic failure (e.g. SocGen). Maybe it's time to revisit our approach towards decision-making? Ensuring that all decisions are explicitly justified in terms of either empirical data or a past decision would be a good start.