In general web personalization has the capability to achieve two important business goals: increase advertising revenue and increase sales revenue. The realization of these two business goals is directly related to item sampling and item selection. Where Item Sampling takes the form of a user’s clicks on personalized recommendations, Item selection involves the user choosing one of the personalized recommendations as the final choice.
In this paper by Ying Ho and Bodoff (2013) a scientific gap is identified which limits the understanding of how web personalization can be used to increase advertising and/or sales revenues. In order to fill this gap the researchers try to develop a theoretical model of these behaviours, item sampling and item selection, and their attitudinal antecedents. Additionally, this model is based on the integration of two theories: the Elaboration Likelihood Model and the Consumer Search Theory.
Elaboration Likelihood Model
The Elaboration Likelihood Model, ELM for short, models the effects of a user’s elaboration of individual persuasive items on his or her overall attitude. The model thus shows how a user’s elaboration, defined as the extent to which a person carefully thinks about an argument (Petty and Cacioppo 1986a), of individual recommendations influence a user’s attitude towards the personalization agent, in turn influencing the users decisions to select a personalized recommendation (Petty and Cacioppo 1986a, 1986b; Tam and Ho 2005).
In lay man terms, with the Elaboration Likelihood Model one is able to define the degree to which a user will cognitively process a given recommendation. This is quite limiting as it does not allow for the investigation of the number of recommendations the user investigates, something important if the business goal is to maximize user clicks. Hence the ELM model is combined with CST, Consumer Search Theory.
Consumer Search Theory
Where ELM is limited to individual recommendations the CST model models the umber of a items a user inspects on his way to the completion of a search task. Thus, in CST the item source is seen as a distribution where each act of inspecting an item is viewed as sampling an item from the original distribution. Where ELM thus focusses on an individual recommendation CST is able to capture the entire array of inspected elements prior to the completion of a search task. This breadth of sampling of CST and depth of processing from ELM allows for the exploration of influence attitude formation, which in turn affects item selection.
In order to develop a theoretical model of item sampling and item selection, the researchers employ both a lab and a field study. In the lab study a personalized online bookstore is developed and used in combination with a thought-listening technique in order to capture participants’ depth of processing. Despite being a widely used method in order to capture ELM and depth of processing thought-listening technique is an obtrusive measure and hence a field study was used to increase external validity.
During the field study the researchers collaborated with a large music provider in the Asia Pacific region in order to develop a personalized music website. During this 6 month period participants visiting this music website either received good or bad recommendations and where hence tracked on their behaviour.
|Table 1. Summary of Findings|
|H1: Depth of processing has a positive effect on the persistence of attitude that user form toward the personalization agent||Supported||–|
|H2: Confidence in one’s attitude toward the personalization agent has a negative effect on subsequent breadth of sampling from the personalization agent.||Supported||Supported|
|H3: Cumulative breadth of sampling from the personalization agent positively influences confidence in one’s attitude toward the personalization agent||Supported||Supported|
|H4: Item variance negatively moderates the positive effect of cumulative sampling breadth on attitude confidence||Not Supported||Not Supported|
|H5: Attitude persistence moderates the effect of cumulative breadth of sampling from the personalization agent on attitude confidence.||Supported||Supported|
|H6: Attitude confidence moderates the relationship between attitude valence toward a personalization agent and actual selection from the agent||Supported||Not Supported|
As visible in Table 1. one can see that for hypothesis 1-5 there are consistent results for both the lab and field study. Whereas hypothesis 6 was significant in the lab study such a significance could not be observed in the field study. This difference in observation however, could be explained due to the large difference in available alternatives and hence a possible higher required attitude confidence for significance to be visible.
Apart from acknowledging several hypotheses the researchers also discovered two significant predictors of item section they had not hypothesized about. First, the NFC (Need for Cognition), the extent to which individuals are willing towards cognitive activities, exerts a negative effect on the person’s actual selection of a personalized item in the lab study. Second, attitude confidence, the degree of certainty a person has in a choice, has a positive effect on item selection in both the lab and field studies.
Strengths and Weaknesses
First of all, in the lab study the researchers made use of a though listing method, this method although widely used might have altered participants behaviour and thus jeopardize the result of hypothesis 1. In order to compensate for this result, the researchers used a second, field experiment, to increase external validity. Apart from using a thought listing method the researchers also made sole use of experience goods in their study. Arguably they did make use of these goods because clicks are meaningful in such a setting, however one may argue this limits generalizability as the developed model might be inconsistent for other types of goods.
Ho, S. Y., & Bodoff, D. (2014). The effects of Web personalization on user attitude and behavior: An integration of the elaboration likelihood model and consumer search theory. MIS quarterly, 38(2).
Lavie, T., Sela, M., Oppenheim, I., Inbar, O., and Meyer, J. 2010. “User Attitudes Towards News Content Personalization,” International Journal of Human-Computer Studies (68:8), pp. 483-495.
Petty, R. E., and Cacioppo, J. T. 1986a. “The Elaboration Likelihood Model of Persuasion,” Advances in Experimental Social Psychology (19), pp. 124-205.
Petty, R. E., and Cacioppo, J. T. 1986b. Communication and Persuasion: Central and Peripheral Routes to Attitude Change, New York: Springer.
Tam, K. Y., and Ho, S. Y. 2005. “Web Personalization as a Persuasion Strategy: An Elaboration Likelihood Model Perspective,” Information Systems Research (16:3), pp. 271-293.