Friday, 22 May 2009

Fast and Frugal Decision Trees

Many processes involve decision makings. Often making these decisions involves searching for, considering and making judgments on a wide range and large volume of data. Having found all the information, and being assured of the relevance of every item it, somehow making the decision doesn’t seem any easier. In fact the sheer volume of information may just be hiding the wood from the trees.

One helpful approach is to stand back from all the information and try to identify which few elements actually “feel” to be key to making the decision. This may involve tapping into your intuition or, if you prefer, your “gut feelings”. Next express these elements as closed questions – is the answer either “Yes” or “No”. Then add the outcome associated with the “Yes” answer and with the “No” – typically this will be the action that the decision maker needs to take next. Assemble the questions into a decision tree.


In some instances, reaching a decision may involve answering several questions, trying to find an overall “Yes” or “No”. If time is a priority – perhaps you have to review many cases per day and need to spend as little time on each as possible – try to rank or sequence the questions to increase the chances of getting to the answer high up the decision tree. Review the questions and amend them – aim to draft the questions so that should you achieve the desired answer at any point you can stop the whole process. What you are aiming to do is to reach a decision as early as possible and not have to pose every question to get to the answer. In the example decision tree below, developed using bCisive, achieving a “Yes” answer at any point achieves the goal of the decision tree and stops the process.

This is all explained in an entertaining and enlightening fashion by Gerd Gigerenzer in his book, “Gut Feelings: Short Cuts to Better Decision Making”. If you are involved in decision making or are preparing procedures or decision trees this book will be of great help in developing your analysis of the problem and the design of the procedure.


Dane said...

I am intrigued by this idea of using a visual tool to make cleaner decisions! Would enjoy seeing your map, but clicking on the image reveals only a slightly larger, blurry pic. Could you maybe reply with a link where a larger version is posted?


Steve Rothwell said...

Thanks for the feedback. I will put up a larger image. I have deliberately blurred some of the text due to commercial restrictions though I hope it will still serve to get the main idea across. Drop me a direct email and I may be able to share directly.

Tim van Gelder said...

Steve, very interesting to see bCisive deployed in this way. As you're aware the standard bCisive decision map has quite a different structure to it, presenting a number of options, perhaps in a hierarchy, and then raising the various qualitative arguments for/against options. Your decision trees are quite a different paradigm (though no less valid). The standard bCisive decision map is more suited to complex, one-off decisions where there is plenty of time to deliberate, whereas your decision trees are more suited to repeatable decisions on topics that have been reasonably well defined in advance, and where a quick way to a good answer is wanted. Different horses for different courses!

Steve Rothwell said...

Thanks Tim - I think you've neatly summarised what I was struggling to express. The decision trees are exactly for the purpose of supporting people in making repeatable decisions - quickly and with minimal risk. There are especially useful when developing processes and procedures to identify both sequence and significance. They also become useful guidelines for training and ongoing decision support.