Decision trees are used when complex branching occurs in a structured decision process. Trees are also useful when it is essential to keep a string of decisions in a particular sequence. Although the decision tree derives its name from natural trees, decision trees are most often drawn on their side, with the root of the tree on the left side of the paper; from there, the tree branches out to the right. This orientation allows the analyst to write on the branches to describe conditions and actions.
Unlike the decision tree used in management science, the analyst’s tree does not contain probabilities and outcomes. In systems analysis, trees are used mainly for identifying and organizing conditions and actions in a completely structured decision process.
It is useful to distinguish between conditions and actions when drawing decision trees. This distinction is especially relevant when conditions and actions take place over a period of time and their sequence is important. For this purpose, use a square node to indicate an action and a circle to represent a condition. Using notation makes the decision tree more readable, as does numbering the circles and squares sequentially. Think of a circle as signifying IF, whereas the square means THEN.
When decision tables were discussed in an earlier section, a point-of-sale example was used to determine the purchase approval actions for a department store. Conditions included the amount of the sale (under $50) and whether the customer paid by check or credit card. The four actions possible were to: complete the sale after verifying the signature; complete the sale with no signature needed; call the supervisor for approval; or communicate electronically with the bank for credit card authorization. Figure below illustrates how this example can be drawn as a decision tree. In drawing the tree:
- Identify all conditions and actions and their order and timing (if they are critical).
- Begin building the tree from left to right, making sure you list all possible alternatives before moving to the right.
This simple tree is symmetrical, and the four actions at the end are unique. A tree does not need to be symmetrical. Most decision trees have conditions that have a different number of branches. Also, identical actions may appear more than once.
The decision tree has three main advantages over a decision table. First, it takes advantage of the sequential structure of decision tree branches so that the order of checking conditions and executing actions is immediately noticeable. Second, conditions and actions of decision trees are found on some branches but not on others, which contrasts with decision tables, in which they are all part of the same table. Those conditions and actions that are critical are connected directly to other conditions and actions, whereas those conditions that do not matter are absent. In other words, the tree does not have to be symmetrical. Third, compared with decision tables, decision trees are more readily understood by others in the organization. Consequently, they are more appropriate as a communication tool.
Choosing a Structured Decision Analysis Technique
We have examined the three techniques for analysis of structured decisions: structured English, decision tables, and decision trees. Although they need not be used exclusively, it is customary to choose one analysis technique for a decision rather than employing all three. The following guidelines provide you with a way to choose one of the three techniques for a particular case:
- Use structured English when
- There are many repetitious actions, OR
- Communication to end users is important.
- Use decision tables when
- Complex combinations of conditions, actions, and rules are found, OR
- You require a method that effectively avoids impossible situations, redundancies, and contradictions.
- Use decision trees when
- The sequence of conditions and actions is critical, OR
- When not every condition is relevant to every action (the branches are different).