5.5. Conjoint analysis (III, IV)

Conjoint analysis, similarly to the ranking done in this study, was applied to the whole product, not to its various single features. The subjects compared whole product variants, but the computer software applied in the conjoint analysis also yielded the relative significance of the different features and the utility of the different feature levels. The software – based on the theory of Statistical Design of Experiments (SDE) – helps both in formulating representative product concept and in analysing the results as far as the importance of the selected attributes and attribute levels is concerned. The inter-rater consistency of the two substudies utilising conjoint analysis (papers III and IV) showed that the similarity was quite good. The following factors improved the reliability of conjoint analysis: (1) thorough planning, (2) selection, number and type of attributes and their levels with as many numerical and continuous variables as possible, (3) illustrative communication about products and concepts, (4) full-profile approach, giving a complete description of the "product stimuli" across all attributes, and (5) giving priority to attributes known to be independent.

The ergonomic feasibility of the basic conjoint method was supported in the papers III and IV, which showed it to (1) comprise ergonomic simulation of the subjects’ daily activities, (2) be a hybrid method, including characteristics of both performance and preference criteria, (3) make it possible to use and evaluate each product concept separately, and (4) utilise computer-aided conjoint analysis.

Conjoint analysis seemed to be a suitable method for eliciting the subjects’ opinions. In the chair experiment (paper IV), all the chairs were tested and compared in pairs by the subjects. This made the judgement easy for the elderly. In the microwave oven experiment (paper III), not all of the alternatives were simulated so concretely as the four main ones, but the leader of the experiment presented others with the help of the four workstations available. Without the help of simulation, the comparison cards would have been difficult to fill in by the elderly, since some subjects already had minor difficulties with the ones used.

When comparing products, the average consumer can evaluate a maximum of five to six features simultaneously. If there are more features, the respondents tend to concentrate on the features they find most important and ignore the others. For the elderly, four features were found to be especially appropriate. The levels in the experimental stimuli should be as balanced as possible, since the importance of an attribute tends to increase as the number of levels associated with the attribute increases. The range of levels should be believable, actionable and communicable.

The EEE₅ procedure can be described as follows:

1. Experts collect data on the product and its users.

2. Experts as a team formulate the features (total number of attributes preferably 4, max. 5) of the product concepts and the feature levels.

3. Conjoint cards are formulated according the features and their levels with the conjoint software (SDE).

4. The product pairs of each conjoint card are compared and rated by end-users supported by real prototypes or sketches, which, together with instructions and cards, guarantee unambiguous communication.

5. Based on the ratings, the conjoint software calculates the relative significance of the features and the utility of the feature levels.

6. The best combination of the product concept can be derived, based on maximal ergonomic utility among a considerable amount of possible concepts.