Consumer preferences can serve as an effective basis for determining key product attributes necessary for market success, allowing firms to optimally allocate time and resources toward the development of these critical attributes. However, identification of consumer preferences can be challenging, particularly for technology-push products that are still early on in the technology diffusion S-curve, which need an additional push to appeal to the early majority. This paper presents a method for revealing preferences from actual market data and technical specifications. The approach is explored using three machine learning methods: Artificial Neural Networks, Random Forest decision trees, and Gradient Boosted regression applied on the residential photovoltaic panel industry in California, USA. Residential solar photovoltaic installation data over a period of 5 years from 2007–2011 obtained from the California Solar Initiative is analyzed, and 3 critical attributes are extracted from a pool of 34 technical attributes obtained from panel specification sheets. The work shows that machine learning methods, when used carefully, can be an inexpensive and effective method of revealing consumer preferences and guiding design priorities.

This content is only available via PDF.
You do not currently have access to this content.