In the digital age of consumption, the internet has been a platform that has greatly enhanced the choices available to consumers. Despite the internet’s potential to enhance consumer utility, this vast increase in choice can potentially overwhelm customers and lead to worse choices (Lurie 2004). A growing trend has been the use of recommendation systems by online retailers and content providers. These systems seek to predict the rating or preference that potential consumers would give an item and suggest products or content based off those. In the world of e-retail, Amazon’s recommendation system is widely recognised as the industry leader, which helps consumers through item-to-item collaborative filtering, suggesting items based on previous purchases or views by other consumers with potentially similar interests or preferences.
The decision quality of electronic recommendation systems also serves as an indicator of service quality (Zeithaml et al, 2002), however, little academic research has been done in this field. The 2011 paper by Askoy et al ‘Decision Quality Measures in Recommendation Agents Research‘ takes a closer look at the decision quality of such systems. The paper examines empirically examines the relationship of different measures that have been used to date in recommendation agent literature. In the paper, they distinguish between preference-dependent measures (PDM), which require knowledge of the different decision maker preferences as well as the attribute values of alternatives that are available; preference independent measures (PIM), that do not require any knowledge of individual decision maker preferences but do require product attribute values, and subjective measures (SM), which depend neither on the knowledge of consumer preferences nor attribute values (Askoy et al 2011: 111). Through a simulated experiment, participants were assigned to conditions simulated by a recommendation agent for a database of 32 cellphones, which were each defined by different attributes (price, weight, talk time and stand-by time). This allowed the experiment to simulate a wide range of recommendation agents currently offered by online retailers to their customers.
Results which compared PDM and PIM results verified that there are considerable gains from knowing consumer preferences that are gathered by recommendation agents. Furthermore, PIM measure prove to a way of assessing decision quality when decisions are unknown. These results suggest that setting up mechanisms to compare product attribute values may be highly beneficial to online retailers even if recommendation agents are not in use.
Electronic recommendation systems help customers sort through the ever-increasing array of choices and alternatives and make better and more informed choices. Similar to real salespeople, electronic agents can serve dual roles of being both providers as well as collectors of information. The paper serves as an initial step in assessing the decision quality as a metric for gauging the impact of marketing activities and to help marketing practitioners have a better understanding of the effectiveness of such electronic recommendation agents. This is certainly an area that will be a growing area of interest as companies aim to better their own recommendation systems to increase sales and increase utility and satisfaction of customers.
Askoy, L., Coil, B., Lurie, N. (2011) “Decision Quality Measures in Recommendation Agents Research”, Journal of Interactive Marketing, vol. 25, pp 110-122
Lurie, Nicholas H. (2004), “Decision Making in Information-Rich Environments: The Role of Information Structure,” Journal of Consumer Research, 30, March, 473–86.
Zeithaml, Valarie A., Parsu Parasuraman, and Arvind Malhotra (2002), “Service Quality Delivery through Web Sites: A Critical Review of Extant Knowledge,” Journal of the Academy of Marketing Science, 30, 4, 362–75.