 |
|
 |
 |
 |
|
|
Authors: Abstract Electronic commerce (e-commerce) provides an important chance for established large companies to gain more customers as well as for new small companies to have a good start and rapid growth. Persuading customers to return to an e-commerce site is even more important than attracting them in the first place. Establishing trust between customers and companies through web interface is not as easy as through human-buyer-human-seller interaction. In the past few years, a number of experiments have been conducted in order to determine the factors that influence customers' trust in online businesses. The goal of our experiment was to establish which features that appear on commercial websites are trust-inducing. Our study focused on three independent variables: customer service, testimonial (self and third party) and security features representation (graphics and text). We designed the homepages of eight web sites with different combinations of treatments of the three independent variables. Each of the 52 subjects reviewed all the eight homepages and gave each of them a relative rank of trustworthiness. After the experiment subjects answered additional e-trust related questions. The results of our experiment and the survey show that all three features that we tested are important in establishing strategic trust among on-line customers.Introduction Background The fast growth and development of Internet and WWW shows the possibility of a new but important medium of business. Electronic commerce (e-commerce) opens the door for a large variety of businesses including information-technology related ones and old traditional ones. It is an important chance for established large companies to gain more customers as well as for new small companies to have a good start and fast grow.Through e-commerce, businesses have the opportunity to sell products and services 24 hours a day and to reduce costs associated with personnel and retail space. Also, online businesses have more chances of attracting new customers and having worldwide exposure. Designing web sites that efficiently present products and easily attract customers is one of the challenges of e-commerce. But selling on the internet is not just creating a web site. Business-to-customer commerce involves interactions and transactions between a company and its consumers. Similar to traditional businesses, it is of great importance for e-commerce not only to attract customers but also to make them come back. Traditional commerce uses catalogs, door-to-door sales people, mail orders and storefronts for their products and services. Unlike traditional stores, online stores use web-based technologies to sell products. A transaction in this case is done between a human and a computer (web) interface. Although information about different products and vendors is more accessible and orders and payments can be done electronically, customers still have to deal with the buying process. They still have to collect and interpret information, make buying decisions and enter purchase and payment information. Relevant theories and previous research To establish trust between customers and companies through web interface is not as easy as through human-buyer-human-seller interaction.The risks associated with e-commerce can be broadly classified into the following categories: In the past few years, a number of experiments have been conducted in order to determine the factors that influence customers’ trust in online businesses. Some researchers indicate that several factors (comprehensive information, shared value and communication) can effectively increase trust, which in turn can increase customer loyalty. [2] According to Egger’s research [3], trustworthiness often depends upon the strength of a brand name, i.e. its reputation, whether the vendor has a presence off-line or only online. In addition, to increase consumers’ trust in online vendors, consumers should also be informed about the vendor’s privacy policy, e.g. why vendors might require unusual personal details and what happens to confidential information after the transaction. Consumers' judgment of trustworthiness can also be aided by the involvement of truly independent parties. Moreover, a valid legal framework that supports the transaction may also be a determinant of consumers' trust.
Another e-commerce study was conducted by Cheskin Research and Studio Archetype [6]. Four hundred and sixty three Web users and a wide range of experts in the worlds of e-commerce, Website development and academia contributed to this study. The study determines the elements that communicate trust in e-commerce sites, based on consumer and experts feedback. The 6 fundamental forms to communicate trust are: brand, navigation, fulfillment, presentation, up-to-date technology, seals of approval. Trustworthiness is defined as experience over time (build, confirm and maintain trust). A second study done by Cheskin Research [7] extends the learning from the initial Trust Study [6] by investigating which are the most recognizable and trusted symbols of online security and by identifying the most trusted websites in the Americas. Shneiderman establishes in [1] that trust is a positive expectation about the future based on past performance and truthful guarantees and he gives some guidelines on how to establish trust in e-commerce. The following are some important ones:
Current products Examples of organizations that provide certificates of trust and seals of approval for e-commerce web sites:
Websites The list includes websites with high and low trustworthiness reputation among online customers:
Experiments Introduction and hypothesis Our experiment is designed to determine what features are important in inducing strategic trust among online customers, which factors will most likely make a B2C e-commerce website trustable and persuasive to the extent that customers will purchase from the website. We hypothesize that in addition to standard usability issues, these three issues will influence trust: extensive customer support, external testimonials and graphical representation of security and privacy policies.According to our hypothesis the most trusted page is the one containing external testimonial, detailed customer support and contact information, and graphic seals. Variables We chose three common strategic trust inducing features from real websites that were not thoroughly studied in previous experiments, but are considered to be important for an e-commerce site's trustworthiness. The final version (we changed one independent variable according to the pilot study results, see section on pilot study results) of independent and dependent variables are as follows:Independent variables: Customer support (2 treatments):Minimal contact information: only email address is provided. Maximal contact information: Email address, telephone numbers, real store addresses are provided.
Text: the testimonial is represented as text. 
We designed 2 x 2 x 2 = 8 web pages (one of the pages is shown above) with different combinations of treatments of the three independent variables, as shown in the table below:
Dependent variables: Relative rank of trustworthiness: each subject estimated trustworthiness of each of the eight web pages on a scale from 1 (least trustworthy) to 9 (most trustworthy). The pages were presented to the subjects in a random sequence, but the subjects could re-visit each page. Subjects were timed and observed during the evaluation.Pilot study results Our original hypothesis stated that customer support, external testimonials and short forms requesting only necessary billing and shipping information are important in inducing trust and was changed based on the results of this study.We conducted a pilot study on 4 subjects. This small group represented our target test audience, which is college students or graduate students who are comfortable with computers and have been using Internet before the experiment. All subjects were timed and observed when they viewed and ranked the pages. Based on the results of the pilot study, we estimated the amount of time the experiment will take, broken down into stages: 1. Read the instructions, ask questions, sign Consent Form: ~1 minute 2. View eight web pages, rank them: ~6 minutes. 3. Fill out the questionnaire: ~3 minutes Based on feedback from our subjects and suggestions from Dr. Eric M. Uslaner, we made the following changes and adjustments to our experiment: We changed the independent variable of Customer Information Filling Form to Security Features Representation, because Customer Information Filling Form was not a factor in web page's trustworthiness. All subjects said they do not care about the filling form because they can always give fake personal information, except for the necessary shipping and billing information. On the other hand, some subjects showed great interest in the security seals on the web pages that we grabbed from the real web sites. They suggested graphic security seals convey trustworthiness of a website. In the questionnaire, we removed the items about Customer Information Filling Form, added questions about the style of security features representation, and shortened the questionnaire to make it more related to trust inducing features customers might care about. We decided to warn subjects that our pages are intentionally similar, because the subjects of the pilot study could hardly see the difference, and thought they were viewing the same page in the beginning. We rephrased the wording of testimonial and customer support. We added privacy policy statement on all the pages. We removed the horizontal scrolling on each page. Subjects We chose subjects who are comfortable with computers and have been using Internet before the experiment. 6 undergraduate and 46 graduate students from the University of Maryland participated in our study. Our subjects consisted of 34 males and 18 females between the ages of 20 to 37. The majority of the subjects are computer science majors. There were also physics, economics, art history, library science, electrical engineering, mathematics, mechanical engineering, French and biochemistry majors.We conducted a within subjects experiment. No training was required. Prior to the experiment, our subjects were given instructions about the purpose of the experiment, their tasks, and the procedure of experiment. They were given chance to have their questions answered to satisfaction. All subjects signed the consent forms. Materials The following materials were used to conduct the experiment:
Procedures and problems The subjects had no special training to participate in our experiment. The subjects were given the instructions and consent forms at the beginning of the experiment. They had their questions about the experiment answered to their satisfaction, and signed the consent forms. Then they were represented with the first web page of the experimental interface that is divided into two frames. The left frame contains links to the eight web pages. Initially the right frame contains the instructions. The order of the eight web pages linked to the eight buttons is randomly initialized when the index page is loaded. When the buttons are clicked, the web page corresponding to the button is shown in the right frame. The subjects clicked the eight buttons to browse the eight web pages, and gave each page a relative trust rank on a scale from 1 (least trustable) to 9 (most trustable) by clicking the radio button in the left frame. They could click the eight buttons in any order they wanted, and could jump back and forward to make changes of ranks. The experiment software recorded the time subjects took to finish ranking, which was from the moment the index page was loaded to the moment the subjects clicked the submit button to indicate they have finished ranking. On submission, the ranks were recorded by the experiment software. Then a short questionnaire asking about subjects' background and opinion about e-commerce web site trustable features showed up. The subjects answered all the questions, clicked the submit button. The software recorded their answers to the questionnaire and brought up an acknowledgement page. The subjects were observed and asked some questions by the observer about their answers to the questionnaire at the end of experiment.Problems Encountered We encountered the following problem while conducting our experiment:When we ran the first 26 subjects, we suspected that some of them did not take the experiment seriously, based on the short time they spent (around 2 minutes) ranking the pages. We recruited 26 more subjects and observed them carefully. Our observations showed that some of them did the evaluation very quickly but still seriously, so we did not discard any subjects.
Results Our program collected the results of evaluations and the answers to the post-experiment questionnaires into two tables in a MSSQL database. The tables were exported into MS Excel for statistical analysis. (See appendix). We also used VassarStats at http://vassun.vassar.edu/~lowry/VassarStats.htm for 2x2x2 Analysis of Variance. ANOVA Summary
was also a Customer Support x Testimonial interaction, Testimonial x Graphics interaction and a significant interaction between customer support, testimonials and graphical representation. We ran a set of paired t-tests on maximal vs. minimal customer support for pages with external testimonials, as well as for the self testimonials that showed a statistically significant difference in degree of trustworthiness both for the pages containing only textual references to security sites and for the graphically enhanced pages.
Similarly t-tests confirmed our hypothesis that external testimonials matter in establishing customers' trust and that trustworthiness of the graphically enhanced pages is greater than the trustworthiness of the text only security references. The means of trustworthiness ranks and standard deviations for all pages are displayed below:
Questionnaire results Out of 52 subjects 11 buy online more than 10 times a year, 23 buy 3 to 10 times, 15 - 1 to 3 times and 3 students never bought anything online. Only two students are online at least 2-3 times a week, the rest is online every day. 20 of 52 people spend more than 3 hours online; 29 spend 1 to 3 hours and only 3 students spend less than an hour.
Some values were very close to critical, for example merchandise type (t =2.59).
Discussion The goal of our experiment was to establish which features that appear on commercial websites are trust inducing. Previous studies determined that elements like brand, navigation, fulfillment, presentation, up-to-date technology, seals of approval communicate trust in e-commerce websites.Our study focused on customer service, testimonial (self and third party) and security features representation (graphics and text). Three-way analysis shows that there is interaction between three tested features. Interaction between testimonials and their graphical representation and testimonials and customer support is significant. The analysis of customer service feature showed that extensive customer support is very important. The pages that have extended customer service scored on average 2 points higher than the pages that contained only the email address of the online store. This outcome is supported by the post-experimental questionnaire results which showed that information like contact address and number and clearly stated return policy are a major concern for the customers. These two features scored highest among all decisional factors for online purchases. The analysis of self and third party testimonial showed that the presence of an external testimonial also contributes to building up trust. Although this proved to be statistically significant in the experiment, in the questionnaire it was not considered to be that significant. 16 out of 52 subjects never heard of the third parties presented in our pages. So there may be other reasons that contribute to this significant difference. According to subjects' feedback one of the reasons may be the amount of information customers see on the front page of the website - the longer text is visually more convincing. The analysis of security features representation showed that graphic seals of approval and graphic security seals induce more trust as opposed to the same information just textually represented. The reason for that may be that visual memory triggers trust without analysis of content. The questionnaire the subjects filled after the experiment helped towards a better understanding of the problem. Our expectation was that the mentioning of a real store would add to the credibility of the website. The subjects' feedback show that they do not consider the online store to be an extension of the real store, but rather its convenient replacement. Therefore, details about the real store proved to be not important. As opposed to previous studies and our expectations, merchandise type was not a concern for our subjects. The reason for this may be that most of our subjects were computer science students and the websites were selling computers, and they might have assumed that the merchandise type meant brand. Conclusions Impact for practitioners Our experiment shows that all three features that we tested are important in establishing strategic trust among on-line customers. Website designers should include all customer service information available, provide phone numbers for technical support, clearly state the return policy, and provide address for merchandize return. Graphic seals should be included as well as external testimonials, if available.Suggestions for future research Our study can be considered as a preliminary investigation into a very interesting topic of on-line trust. Future research should take into consideration the following issues that we came across when running our experiment:1. We were very careful about the visual impact of colors, size of tables, but we did not take into consideration the length of text. 2. The source of the external testimonial has to be meaningful to the subjects. 3. The questionnaire should be more specific about the kind of information requested. 4. Because the dependent variable of the experiment is subjective, the number of subjects should be greater, and they should belong to different population groups. |
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
