Introduction

There are two kinds of decisions: intuitive and analytic. Intuitive decisions are not supported by data or formal documentation. Many decisions made in organizations around the world are intuitive. The decision maker does research, gathering documentation, and soliciting advice from peers and subordinates. Armed with this information, the decision maker adds a significant amount of their bias, and simply makes a decision.

These intuitive decisions are exposed to numerous potential holes that make them especially weak and prone to errors. For instance, intuitive decisions:

• Are based on “gut” feelings and lack robust logic needed to convince others and defend the decision made
• Are not easily taught, transferred, or repeatable for others to use
• Provide little or no means to ensure group consensus and decision making

An analytic decision making approach is significantly different in that the model is founded on logic by exploiting mathematics, key to an effective and deployable decision making model. The analytic approach has the following characteristics:

• Simple to construct
• Adaptable, from individuals to groups to enterprise-wide application
• Consistent with human intuition and thinking
• Encourage compromise and consensus
• Repeatable approach

As organizations need to move from individual to group to enterprise decision making, analytic decisions provide the only means to make unbiased, repeatable decisions that support a broad user community with varied agendas.

This paper defines and describes the use of Analytic Hierarchy Processing (AHP) as a robust, effective means to address the complexity facing the modern analytic space. A follow-up paper will provide a real-world example of its application.

The AHP Method

Today’s organization needs to blend intangible and tangible criteria and cover a broad range of data and information. Moreover, each datum and criteria must be represented at different scales of impact and implication. Analytic applications epitomize the need to embrace this broad landscape of criteria and data. This is particularly true when attempting to identify and prioritize user requirements for analytic initiatives. Figure 1 illustrates this challenge.

As demonstrated, analytic teams must not only identify and prioritize competing requirements from users communities that span the organization, but also attempt to map the right techniques and technologies that address those requirements.

Figure 1. Mapping Requirements and Technologies

The Analytic Hierarchy Process was developed by Dr. Thomas L. Saaty in the early 1970’s. This process has been in use for more than 30 years and found extremely useful when applied properly. AHP is a decision-making method based on a problem space divided into hierarchies. Refer to figure 2. This hierarchical model allows an organization incorporate a broad range of criteria and information that may influence the decision process into what Saaty refers to as a “stratified dominance structure for representing the spread of influence”.

Figure 2. Problem Space

Many decision problems involve multiple criteria and qualitative factors which are not agreeable to alternative decision analysis techniques. The AHP is a decision analysis technique employed to address qualitative, multi-factor decision problems in which the decision maker must make a choice among several alternatives.

There are three broad objectives for each AHP model implemented, including:

1. State the Objective: The overall goal or problem being addressed is stated at the very top of the hierarchy. For example, Which prototype is better?, or Which technology best addresses the highest priority requirements?
2. Define the Criteria: These are the factors or criteria that contribute toward reaching the stated goal. It is represented by the middle layers of the hierarchy.
3. Identify the Alternatives: Alternative solutions to address our stated goal are considered individually. All alternatives within the scope of the study are represented as the very last layer of the hierarchy.

Cornerstone to an AHP model is the hierarchy. Following are five steps to defining, building, and implementing the structure.

1. Structure the problem as a hierarchy of criteria or a system of criteria dependencies that resembles a decision tree
2. Elicit judgments or qualitative assertions for each criteria
3. Convert those judgments into quantitative values
4. Use those values to calculate the priority and relative importance of the criteria
5. Synthesize an overall outcome

The AHP Survey

The decision tree of the AHP model has criteria at each level of the tree. Every criteria of every level will be compared with other criteria of that particular level. This paired criteria survey is illustrated in Figure 3. The participant taking the survey must evaluate the criteria pair by answering two types of questions: comparison and intensity.

Figure 3. Paired Criteria Comparison

While the paired criteria questions are qualitative, the response is reduced to a quantitative value. Each of the intensity levels are associated with a numeric value ranging from 0 to 9. These values are then used in the final calculations. The final values ensure that a nonbiased, exhaustive evaluation is conducted, one that not only establishes the priority of a particular criteria, but also the relative importance of that criteria with other criteria at that same hierarchy level. Figure 4 demonstrates a response to a particular pair of criteria.

Figure 4. Criteria Paired Response

Conclusion or Summary

AHP modeling is the culmination of significant research in a practical approach to group decision making with significant benefits for analytic initiatives, including:

• Provides a repeatable process for identifying and prioritizing analytic requirements
• Provides a structure of interdependent criteria
• Establishes the relative importance of criteria as a means to logically choose among alternative solutions
• Provides a method for incorporating the input from enterprise-wide user communities and agendas
• Minimizes bias introduced by vendors or sponsors
• Quantifies priority and rank of requirements for planning

The 2^nd part of this series will detail a real-world application of AHP to identify and prioritize enterprise analytics for a leading organization.