{"id":16705,"date":"2021-04-23T10:21:47","date_gmt":"2021-04-23T08:21:47","guid":{"rendered":"https:\/\/www.munich-business-school.de\/insights\/?p=16705"},"modified":"2023-07-01T14:35:00","modified_gmt":"2023-07-01T12:35:00","slug":"impact-of-recommender-systems-on-consumers-purchase-intention","status":"publish","type":"post","link":"https:\/\/www.munich-business-school.de\/insights\/en\/2021\/impact-of-recommender-systems-on-consumers-purchase-intention\/","title":{"rendered":"Impact of Recommender Systems on Consumers\u2019 Purchase Intention"},"content":{"rendered":"\n

Due to the flood of information and the resulting inability of consumers to make decisions, recommender systems are increasingly becoming a key tool in e-commerce. But how do these systems influence consumers‘ purchase intentions? A new study conducted as part of a Master’s thesis at Munich Business School examines this question using the example of Taiwan.<\/strong><\/p>\n\n\n\n

\n\n\n\n
\n

\u201cBeing digitally connected substantially changes the way companies compete and succeed.“<\/p>\nJung, Kraft, 2017<\/cite><\/blockquote>\n\n\n\n

With advancements in information technology, e-commerce plays an important role in our daily lives and provides consumers with a great deal of convenience. However, consumers tend to become overwhelmed by the ever-growing amount of product-related content they receive, and suffer from information overload, which may lead to poor decision making. To cope with this information overload, recommender systems are increasingly becoming a key tool in e-commerce. The main feature of recommender systems is their ability to analyze user behavior, especially which item a particular user might be interested in, and further provide users with item suggestions according to their preferences and behaviors (Lu, Wu, Mao, Wang, & Zhang, 2015, p. 12). To generate recommendations, recommender systems follow the process that consists of three phases. In the first phase, the system collects extensive user information to gain a deeper understanding of the target users and lay a solid foundation for future phases. This input can be either explicit feedback, such as product ratings, or implicit feedback, such as purchase history. In the second phase, the recommender system filters and utilizes the user data collected in the first phase by adopting a machine learning algorithm. Lastly, the system predicts and presents item recommendations to the users (Isinkaye, Folajimi, & Ojokoh, 2015, pp. 263-264).<\/p>\n\n\n\n

Recommender systems – a powerful tool for decision making and influencing purchase intention<\/h2>\n\n\n\n

Based on the data filtering and rating estimation methods, recommender systems are usually divided into the following categories: content-based, collaborative and hybrid filtering systems. In the content-based filtering technique, recommendation is generated by matching up the contents of items and the users\u2019 profiles. The content of items is represented as a set of descriptions which areextracted from the features of items. The user profile consists of the information about the interests, preferences and needs of the user and is constructed by analyzing the features of items that the user has purchased or viewed in the past (Lops, Gemmis, & Semeraro, 2011, p. 75). Collaborative filtering considers the opinions of other users as important factor and proposes recommendations to a user based on the analysis of other users who have similar preferences (Sivapalan, Sadeghian, Rahnama, & Madni, 2014, p. 163). This approach requires historical data and past ratings of users and works by building a database made up of users\u2019 preferences for items and searching the users who have more in common with the target user (Cheng, Wang, 2014, p. 290). Hybrid filtering was proposed by combining two or more recommendation techniques to achieve higher performance and to avoid the limitations of the individual techniques it combines (Isinkaye, Folajimi, & Ojokoh, 2015, p. 269). Most commonly, the hybrid filtering approach combines the strengths of the content-based and collaborative filtering approaches to optimize recommender systems (Bagherifard, Rahmani, Nilashi, & Rafe, 2017, p. 1777).<\/p>\n\n\n\n

This study – conducted as part of a final thesis at Munich Business School<\/a> – sets out to examine the factors related that affect consumers\u2019 purchase intention toward the products suggested by recommender systems in online shopping and gain a holistic view of recommender systems from the consumers‘ perspective. To this end, our research model integrates the unified theory of acceptance and use of technology (UTAUT) model proposed by Venkatesh, Morris, Davis, & Davis (2003) with the internet users\u2019 information privacy concerns (IUIPC) model proposed by Malhotra, Kim, & Agarwal (2004). The UTAUT model helps to understand users\u2019 perceptions and behavioral intentions towards recommender systems, while the IUIPC model explores users\u2019 privacy concerns regarding recommender systems. Furthermore, to better understand what makes a recommendation useful to consumers, this research model includes the three factors that influence performance expectancy, i.e., recommendation accuracy, novelty, and diversity.<\/p>\n\n\n

\n
\"Research
Research Model (illustrated by the authors)<\/figcaption><\/figure>\n<\/div>\n\n\n

For the purpose of investigating the determinants of Taiwanese online shoppers‘ intention to purchase recommended products, the quantitative approach was chosen to develop the survey. The online survey was distributed via social media platforms and messaging applications, and a total of 276 complete and valid responses were collected.<\/p>\n\n\n\n

How effective recommender systems should be designed and developed<\/h2>\n\n\n\n

The findings indicate that consumers‘ trust in recommender systems and performance expectancy of recommender systems are important factors influencing consumers\u2019 purchase intention and that recommendation accuracy, novelty, and diversity indirectly affect purchase intention by shaping consumers\u2019 perceptions about the performance expectancy of recommender systems. Thus, this study makes several suggestions for practices to facilitate the design and development of effective recommender systems by service providers.<\/p>\n\n\n\n

Building consumer trust<\/h3>\n\n\n\n