Reducing the demand uncertainties at the fuzzy-front-end of developing new online services

Addressing the demand uncertainties at the fuzzy-front-end of developing new online services, this paper tests the roles of numerous cluster-based methodologies in improving the predictive accuracy of consumer opinions. The results with an online service revealed that both crisp and non-crisp clustering methodologies improve the predictive accuracy and hence reduce the demand uncertainties at the fuzzy-front-end of the new product development process. They also showed that non-crisp clustering increases the accuracy more than does crisp clustering. Implications of the findings for our understanding of the earlier stages of the new product development process and for making informed R&D policies are discussed.