ANALYTIC HIERARCHY PROCESS
AHP (Analytic Hierarchy Process) is a compositional approach to the measurement of preferences, with hierarchical arrangement of features and levels.
FEATURES IN BRIEF
- Permits convenient hierarchy definition with an unlimited number of levels
- Determines individual utility values, even with very small samples
- Offers multiple aggregation options
- Runs consistency checks based on various criteria at case level
AHP was originally an instrument to support decision-making, in which a ranking of alternative options is determined on the basis of items that are structured in a hierarchy.
To measure preferences, attributes and their levels are arranged in a hierarchy. The simplest hierarchical structure consists of the individual best concept on the top step, the attributes on the second step and the levels of each attribute on the third step (see illustration). In principle, this structure corresponds to the structure of a conjoint design.
But in contrast to the holistic rating of concepts in conjoint approaches, in AHP the individual levels of an attribute as well as the attributes themselves are presented as alternatives in paired comparisons. Thus, in the example, 5 times 6 paired comparisons for the attribute levels as well as 10 paired comparisons for the attributes are to be done by each respondent. For each one, the respondents must indicate how much they prefer one attribute level against the other or how far one attribute is more important than the other.
This eliminates the need to create a survey design as in conjoint, which determines which overall concepts are presented. Although this makes the survey less realistic, the individual tasks are easier for the respondent to complete.
The results of the AHP are also partial utility values of the characteristics and their importance, which are directly available for each respondent. Thus, as with conjoint, simulations of scenarios are possible.
If there are many characteristics or many values of a characteristic, the number of possible pairwise comparisons increases considerably. However, for the method implemented in the AHP tool of the ADABOX, it is sufficient to present a selection of these pair comparisons to determine part-worth values and importance.