The role and perception of the Web in its various usage contexts is rapidly changing – from an early focus on “Web-only” interaction with customers, information seekers, and other users, to the Web becoming one central component in a multi-channel information and communication strategy. In fact, multi-channel retailers increased their online market share from 52% in 1999 to 67% in 2001 – in contrast to Internet-only retailers, who lost market share respectively [Silverstein, et al., 2002]. Incumbent companies with a traditional store network seem to dominate the online market currently. With the increasing online competition, measuring success has become crucial for both Web-only and multi-channel retailers.

Web site owners have the opportunity to collect, analyze and use an increasing amount of online consumer information. On the Internet users transmit personal information, either actively by sending customer data (e.g. a shipping address for books), or passively, by leaving traces that are registered with the server side (in the so-called Web server log). In a multi-channel context, Web sites can also collect information about online consumers’ use of offline channels. Despite the increasing flow of consumer data, Web sites often lack the ability to utilize the information for measuring e-business success. Multi-channel retailers in particular lack a measurement system to analyze online success in a complex multi-channel information, communication and distribution strategy.

While yielding benefits to the companies, the analysis and use of consumer data increases privacy concerns significantly, which has become a primary impediment for successful e-commerce. Online shoppers claim they would buy considerably more if they were less concerned about their online privacy [Cyber Dialogue, 2001 Department for Trade and Industry, 2001 Forrester, 2001]. Privacy legislation and industry-driven initiatives aim at alleviating some of these concerns. As a consequence, a Web site that aims at analyzing and using online consumer data must include privacy requirements in its analysis practices. Moreover, it must efficiently communicate these privacy standards to its users in order to increase consumer trust.

Contribution

The thesis’ specific contributions are the following:

Measuring antecedents of trust in multi-channel retailing (Chapter 2)
The success of multi-channel retailing and the importance of privacy is discussed from a consumer perspective. We present a structural equation model of consumer trust in a multi-channel retailer. Trust is a well-known predictor of willingness to buy.
A significant influence of perceived store reputation and perceived store size on trust in an e-shop has been identified, which supports our hypothesis that cross-channel effects exist between a retailer’s physical store network and its e-shop. We found that consumers’ perceived privacy had the strongest influence on trust. The results suggest to further integrate distribution channels and to improve the communication of privacy online.
Design and testing of a Web analysis framework (Chapter 3)
Our research on consumer perceptions in multi-channel retailing motivates to further investigate the notion of success measurement on the Internet. We propose an analysis framework consisting of 82 analyses for measuring the online success of Web sites. New conversion success metrics and customer segmentation approaches have been introduced. A particular emphasis has been placed on metrics and analytics for multi-channel retailers. The framework has been tested on Web data from a large multi-channel retailer and an information site.
Prototypical development of a privacy-preserving Web analysis service (Chapter 4)
The analysis of Web data requires that privacy restrictions must be adhered to. The impact of privacy requirements on our analysis framework is discussed. Legal restrictions and requirements specified in the Platform of Privacy Preferences (P3P) are presented. We propose a privacy-preserving Web analysis service that calculates the set of 82 business analyses and indicates when an analysis is not compliant with privacy requirements or when data is not available. A syntactical extension of P3P is proposed.
Extension of user privacy requirements (Chapter 5)
An important application that uses results from the described Web analysis service are personalization systems. These systems become more efficient with an increasing amount of user information. Thus, the impact of privacy concerns is particularly high for personalization applications. An overview of consumer privacy concerns and their particular impact on personalization systems is provided, which is summarized in a meta-study of 30 privacy surveys. Approaches to privacy-preserving personalization have been discussed.
Development of a privacy communication design (Chapter 6)
A company must not only respect privacy requirements in its Web analysis and usage purposes but it must also effectively communicate these privacy practices to its site visitors. A new user interface design approach is proposed, in which the privacy practices of a Web site are explicated in a contextualized manner, and users’ benefits in providing personal data clearly explained. A user experiment has been conducted that compared two versions of a personalized store. Subjects who interacted with our new interface design were significantly more willing to share personal data with the Web site. They rated its privacy practices and the perceived benefit significantly higher and made considerably more purchases.

The thesis concludes with a summary and an outlook on further research in Chapter 7.

A sketch of the thesis structure is captured in Figure 1-1:

Figure 1-1: Thesis structure

Methodology of the thesis

This thesis chooses exploratory and confirmatory research approaches that aim at balancing advantages and disadvantages of both theory-building and theory-testing methodologies.

Chapter 2 takes a confirmatory approach to data analysis. Hypotheses are developed and tested on data from 1048 online consumers. Multi-causal relationships have been observed using LISREL 8 [Jöreskog and Sörbom, 2003]. Chapter 3 uses an exploratory research approach. Techniques from data and Web mining are applied on Web user and usage data. The data sample includes customer information from 13,653 customers, 92,467 sessions from a multi-channel retailer’s Web logs and external information. Confirmatory elements have been integrated into the Web mining approach in Section 3.7, where background knowledge is used as guidance for the mining process. Preconceptions about the data are tested against a reference set of 27,647 user sessions from a non-commercial site. Chapter 4 develops a prototype, which integrates the exploratory analysis techniques of Chapter 3. Chapter 5 is based on a comparative literature review. Chapter 6 concentrates on an experimental approach. A between-subjects design has been chosen to explore the impact of privacy and personalization communication on users’ data disclosure behavior.