Decision Theory - What's the point?

Early in my management consulting career in ICL I developed an interest in decision theory. This was partly as a result of meeting several academics from different business schools who had different angles on the traditional decision tree approach. There was Prof Larry Phillips at the LSE with multi-attribute utility decomposition (MAUD), Dr Colin Eden at the University of Bath with his cognitive processing engine (COPE), Dr John Friend from the Tavistock Institute with his strategic choice method, and several others. As is the case with academics (in my experience, at least) each was deeply sceptical of the others and attempts at reconciliation were doomed to failure.

However, I took the view that the best way of finding out the strengths and weaknesses of these approaches was to try them out in practice. Because of who I am, the softer, more judgmental approaches had a greater appeal and I was able to secure several small projects learning about them and putting them to the test. These experiments came to the attention of my management "grandfather" and he came to me one day to tell me that he had secured the agreement of a major ICL customer to use my skills to help them make a complex decision. To say I was scared was an understatement - I certainly didn't feel that my knowledge had reached that level - but, as tends to happen to me, he said "Don't worry. I'm sure you can be of significant help to them - I explained that it would be experimental, so I'm only charging them £x" - he quoted a number in five figures.

So I set off for London to see the client. I spent several weeks interviewing people, seeking to understand what the decision was all about and, over time, developed a large and complex model embodying both objective and subjective variables. We ran the model a number of times and began to understand its behaviour. Eventually we felt we had something we could rely on. We ran it for real and came up with a result which seemed intuitively in the right ball park. We presented it to the decision maker. "Very interesting," he said, and asked a couple of good questions about how the model was constructed. He then suggested a couple of changes in weighting, and we ran the model again. "Still not quite right," he said, and suggested a couple more changes. We implemented them. "Yes, that feels right," he said, and we went with the result. What I realised, in retrospect, was that my job had been to create a convincing justification of a decision that had already been made.

I wonder how often that happens.