Detail versus relevance: getting the balance right

Submitted by Owen Plowman on 23 December, 2007 - 17:39

In my earlier post on making decisions, I noted that the use of models is a useful way of applying automation, consistency, and reusability to the decision-making process. When somebody makes decisions based on intuition, they’re really using an informal mental model that can’t be easily explained or shared with others (and that is, quite possibly, incomplete and therefore unreliable). Expressing the situation in a more formal model can be very difficult, but raises the quality of decisions a great deal.

We’ve very familiar with decision-making models at Actenum. One of the steps in the deployment of an Actenum application is to model the various aspects of the target operational situation in a way that supports our automated decision support technology. For example, the model underlying Actenum MPS has multiple layers:

  • Component, used to express the hierarchies of equipment to be maintained;
  • Flow, which represents the network and throughput of the process chain components;
  • RCM, which captures the Reliability Centered Maintenance aspects of the operation.

In devising the decision-making model, the fundamental issue that has to be addressed is the level of sophistication that’s appropriate. When modeling a complex real world situation, there will always have to be concessions in fidelity: not all aspects of the situation can or should be incorporated into the model. But what is included and what is omitted may have dramatic impact on the speed of the process, and the reliability of the resulting decision alternatives. The trick is to get the balance right.

This sounds obvious, doesn’t it? We find, however, that there’s often a desire in organizations to place too much emphasis on fidelity, which leads to clutter and irrelevancy. And we’re not alone: the phenomenon is discussed in the SPE paper that I referred to in my earlier post (“Decision-Making in the Oil and Gas Industry: From Blissful Ignorance to Uncertainty-Induced Confusion”, J. Eric Bickel and Reidar B. Bratvold, SPE Paper 109610-MS; find it at OnePetro). The authors state, “... companies tend to build too much detail into their decision-making models from the start and focus too much energy on specific cases or particular inputs ....”

One way around this, as the authors suggest, is to consider the distinction between cogency and verisimilitude. Cogency is having the property of appealing to reason, or being compelling. Verisimilitude means being true to life. Decision-making models need to have cogency, not verisimilitude. The models should, therefore, represent the aspects of the operational situation that are most relevant for making decisions.

Our applications provide insights to decision makers regarding critical operational issues. Modelling the details of these issues is easy. Building in incisiveness is hard work.