Experimental designs are required in widely used techniques in marketing research, especially for preference-based conjoint analysis and discrete-choice studies. Ideally, marketing researchers prefer orthogonal designs because this technique could give uncorrelated parameter estimates. However, orthogonal design is not available for every situation. Instead, efficient design based on computerized design algorithm is always available. This paper presents the method of efficient design for estimating brand models having attribute and availability cross effects. The paper gives a framework for implementing designs that is efficient enough to estimate model with N brands, each brand have K attributes, and brand attribute has specific levels. The paper also illustrates an example in food consumption study.
Australian and New Zealand Marketing Academy Conference (anzmac), 2011