Data Collection Method

Figure 5.1: Alternative models o f preference (green and Srinivasan, 1978).

5.6 Data Collection Method

Following the selection of the preference model, the next step is to determine the way in which the prototypes or mock ups are presented to respondents for evaluation. Data collection methods in conjoint analysis have mainly involved variations of two basic methods: the two-factor procedure (Johnson, 1974), and the full-profile approach (Green and Rao, 1971).

5.6.1 The Two-Factor Method

The two-factor procedure is also referred to as the trade-off analysis or two-at-a-time procedure. As the name implies it considers factors (attributes) on a two-at-a-time basis.

The respondent is asked to rank the various combinations of each pair of factor levels from most preferred to least preferred. Therefore, it reveals preferences for prototypes only that are partially described by 2<K attributes (Mohn, 1989).

In practice, both factors that are considered are combined together in a matrix with the levels of a factor allocated either horizontally or vertically on the matrix. Ranked preferences are indicated in the cells for all combinations within the matrix. Although the

two-factor procedure is simple to apply, reduces information overload on the part of the respondent and also lends itself easily to mail questionnaire form, since no special props are needed, it is not without its limitations:

First, the approach has been generally criticised for its being limited to raking data (Wittink et ah, 1994). Notwithstanding, its original introduced by Johnson (1994) for ranking data, Aust (1996) points out that it can be easily generalised for nearly other evaluation scale that is common in conjoint analysis.

Second, in decomposing the overall set of factors to two- at-a-time combinations, there is some sacrifice in realism. Moreover, respondents are usually unclear as to what should be assumed about the remaining t-2 factors that are not being considered in a specific evaluation task. Consequently, when the attributes of a product or service are correlated (e.g., for technological reasons), what the rank order in a particular table corresponds to is not clear (Green and Srinivasan, 1978).

Third, given an unwieldy number of conjoint tasks to be evaluated, respondent fatigue may introduce non-sampling errors into the study. For instance, given a hypothetical scenario of say, six factors with each at four levels, the respondent could be asked to fill out 15 tables, each consisting of 24 cells. While partially balanced incomplete block designs (Green, 1974) can be used to reduce the number of two-way tables, the total number of required judgments is still quite large.

Fourth, there is the tendency for respondents to either forget where they are in the table or to adopt patemised types of responses, such as always attending to variations in one factor before considering the other (Johnson, 1976).

Lastly, Green and Srinivasan (1978) argue that the two-factor method appears to be most suited to verbal descriptions of factor combinations, rather than pictorial or other kinds of iconic presentations. For example, a study of package designs in which colour logo, size and shape can be simultaneously varied and portrayed graphically would not lend itself well to this approach.

5.6.2 The Full-Profile Method

The full-profile method is also known as the concept evaluation task or the multiple factor evaluation method. This method utilises the complete set of factors or attributes considered. The main argument in favour of the full-profile approach is that it gives a more realistic description of stimuli by defining the levels of each of the factors and possibly taking into account the potential environmental correlations between factors in real stimuli.

This strength also appears to be its downfall in the sense that it makes the evaluation task tedious for the respondent by having to consider several factors at one time. Because of this overload problem, the full-profile procedure is generally confined to, at most, five or six factors in any specific sort (Green and Srinivasan, 1978). Based on these two considerations. Green and Srinivasan (1978) suggested that “in contexts where the environmental correlation between factors is large and the number of factors on the stimulus card is small (but greater than two), the full-profile approach is likely to be better in terms of predictive validity. However, if the environmental correlation between the factors is small and the number of factors on the stimulus card is large, the two-factor-at-a- time approach is likely to be better”. It is however noteworthy that when some industrial studies have involved up to 25 factors, each at from two to six levels, the analyst is forced to incorporate “bridging-type” factors. Here, the idea is to prepare several card decks in

which the full set of factors is first split into subsets of five or six factors each. Each card deck is then composed of factor combinations that involve up to five factors only. In each case one or two factors are common across decks so that they provide a basis for linking part-worth functions across the various subsets of factors. Table 5.2 compares the strengths and weaknesses of the two approaches to data collection.

Strengths: Two-Factor Method Strengths: Full-Profile Method

• Simplicity in application

• Reduces information overload

• No special props are needed hence lends its self to self-completion and mail questionnaire forms

• Relatively larger number of factors can be accommodated

• Utilises the complete set of factors hence portrays more realism

• Flexibility in accommodating both rank order and rating scales

• Most suited for graphical and pictorial descriptions of factors.

Respondent fatigue as a result of large number of evaluation tasks

Table 5.2: strengths and weaknesses of the two-factor and full-profile method.

Thus, it is clear that a full-profile approach would be best suited for this study based on the following considerations:

• The environmental correlation between factors is large i.e. information features appear in tandem in a graphical format on the user interface.

• The number of factors on the stimulus card would be greater than two and less than six.

5.6.3 Hybrid Conjoint Analysis

Another data collection method in conjoint analysis includes the hybrid conjoint analysis.

This method combines a direct (compositional) part of the study in which the respondents have to give direct judgements about the importance of individual attributes and an indirect (decompositional) part of the study that represents the actual conjoint interview with selected combination of attributes (Green et al., 1981; Green, 1984; Schubert, 1991).

5.6.4 Adaptive Conjoint analysis

The adaptive conjoint analysis (ACA) method is viewed as a modem form of the two- factor method (Johnson, 1987; Schubert, 1991). This method can provide a detailed analysis of the individual benefit stmcture of each respondent because the questions asked are adapted to the previous answers in a computer-aided data collection process (Green and Srinivasan, 1990).

According to Safizadeh (1989) the increasing popularity of the full-profile approach is due to the fact that multiple-factor evaluations have less problems resulting from overvaluation of single major attributes taken out of their context as compared to the other approaches. However, Wittink et al. (1994) have shown that adaptive conjoint is gaining ground.

The number of attributes that could be integrated into a conjoint design increases with the development of ACA or hybrid procedure. It therefore makes sense to pay attention to studies where these attributes are the object of interest. For instance, Orme et al. (1997) in their study concluded that an affirmative or negative presentation of items may influence the evaluation of alternatives. This proves the loss aversion effect from description decision theory (Kahneman and Tversky 1979, 1991).