Virtual Reality Stores for 1-to-1 E-commerce

Transcription

1 Virtual Reality Stores for 1-to-1 E-commerce Luca Chittaro and Roberto Ranon Department of Mathematics and Computer Science University of Udine via delle Scienze 206, Udine, ITALY ABSTRACT Virtual Reality (VR) interfaces to e-commerce sites have recently begun to appear on the Internet, promising to make the e-shopping experience more natural, attractive, and fun for buyers. Unfortunately, usability of these interfaces is a major and still unexplored issue. In this paper, we present the approach we are following to guarantee and increase the usability of VR stores, also providing two practical examples of the solutions we are studying. In particular, the last part of the paper is devoted to discuss in detail how our VR store can be extended to 1-to-1 e-commerce applications. Keywords E-Commerce, Virtual Reality, 3D, Adaptive Interfaces. INTRODUCTION One of the challenges of e-commerce is the design of web sites which effectively present products and are easy and pleasant to use for buyers. Studies of the usability of current e-commerce sites are beginning to appear (e.g., Dong, Martin, and Kieke, 1998), reporting major design problems. These result in buyers failing to find what they are looking for, or abandoning the purchase even if they have found the product. Moreover, as pointed out by an analysis of e-commerce open issues from the HCI point of view (Steiger, Stolze, and Good, 1998), e-commerce sites should satisfy not only the rational needs of buyers (such as ease of use in product finding), but also their emotional and social needs. Although almost all e-commerce sites use traditional 2D user interfaces, some sites are currently making attempts at providing users with 3D interfaces at or Cybertown Shopping Mall at allowing them to explore a VR representation of the store. A VR store has some relevant advantages, if properly implemented: (i) it is closer to the real-world shopping experience, and thus more familiar to the buyer, (ii) it supports buyer s natural shopping actions such as walking, and looking around the store, (iii) it can satisfy emotional needs of buyers, by providing a more immersive, interactive, and visually attractive experience, (iv) it can satisfy social needs of buyers, by allowing them to meet and interact with people (e.g., other buyers or salespeople). Although the idea is promising and can lead to more natural, attractive, and fun e-commerce sites, usability guidelines for designing e-commerce sites (e.g, Serco Ltd., 1999) do not deal with VR stores, and usability issues of VR stores have not yet been investigated in the literature. Our research focuses both on how to design usable VR stores, and how to obtain a personalized user interaction with them. In the following, we first sketch our approach to VR store usability (also providing two practical examples), then we discuss some of the possible ways our VR store approach can be made adaptive to the customer profile for 1-to-1 e-commerce.

2 VR STORE DESIGN BASED ON BOTH MIMICKING AND AUGMENTING REAL STORES In our approach, we tackle the problem of designing usable VR stores from two synergistic directions: (i) designing usable environments by following selected guidelines used in real-world stores (e.g., shop layout, signs, product positions, ), and (ii) taking advantage of the virtual nature of the store by providing users with empowerments (unavailable in real-world stores) to augment their navigation capabilities. To proceed in the first direction, a proper set of valid guidelines has to be identified. To achieve this goal, we are performing four main activities: (i) acquisition of the guidelines used in real-world stores (extensive experience is available on the topic, and can be acquired through specialized literature and expert interviews), (ii) identification of the VR counterparts of real-world guidelines (some might be directly translated, some need adaptation, and some can be impossible to translate, e.g. those concerning scents), (iii) identification of the proper implementation choices to adequately apply the guideline in VR, (iv) evaluation of effectiveness on users (a guideline which has been proven to work in real stores does not necessarily work in virtual ones). As part of the first activity, we are interviewing a local manager of a leading European department stores company, eliciting the merchandising guidelines they follow, i.e. how they present and group products in their stores to motivate purchase. An effective product presentation has indeed to consider many choices ranging from space allocation to product positioning and visual presentation. The first example in the following focuses on the experimentation of one of the guidelines we have elicited. To proceed in the second direction, one could simply take traditional VR navigation aids (such as the electronic map or teleporting) and include them in the VR store. Although useful, this is not the optimal solution, because it does not take into account merchant needs. For example, when it comes to product finding, one merchant priority (both in traditional and electronic commerce) is to achieve the best compromise between two (often conflicting) goals: (i) allow the customer to find the desired products quickly and easily, and (ii) make the customer take a look also at other products while on her/his way to the desired ones. The second goal is essential for merchants to increase their sales. It is indeed wellknown that a substantial part of purchases are not planned in advance before entering a store, but occur as an impulsive response to seeing the product (the reader might be familiar with the experience of going to a supermarket and buying more things than planned). Moreover, merchants are also typically interested in increasing the likelihood that specific products are seen by the customer according to their merchandising strategies (e.g., swimsuits when summer is approaching, Christmas gifts in December, products being heavily advertised in the media, special offers, ). A navigation aid for e-commerce sites should thus take into account the above mentioned merchandising considerations. The VR solution we are proposing is shown in the second example. EXAMPLE 1: THE MASSIFICATION GUIDELINE The term massification was used in the interviews to indicate the display of a mass of several identical instances of the same product, and will be used with this meaning through this paper. When the number of displayed items for a product in a department store is below a given minimum (which varies with the product), it is empirically known that there will be negative effects on sales of that product. This seems to happen for reasons such as: (i) customers are more motivated to buy if they have the impression of a rich store (e.g., imagine how unattractive could be a department store where just one item of any product is on display), (ii) displaying many identical items of a product makes the product more visible (on the contrary, when too few are displayed, managers say the product "disappears"), (iii) buyers prefer to pick up one item from a group of identical items (e.g., contrast the feeling of picking up a bottle of milk in a supermarket with just one bottle left on the shelve versus a lot of bottles available).

3 Figure 1 A non- massified VR product. Figure 2 A massified product visualization. The considered guideline prescribes thus a massified display of products: several instances of the same product have to be displayed together. As any guideline, this has an exception. More specifically, it does not apply to shops (such as boutiques or painting galleries) which sell exclusive, very expensive products, where the buyer needs to feel that (s)he is buying an unique piece. While traditional 2D e-commerce sites typically show only one instance of a product in their menu-based organization, it is interesting to see that the new emerging 3D sites are not making a different choice. This is probably due to considerations such that there is no difference in choosing one instead of another identical item, or worries about decreasing rendering speed. The first kind of considerations is purely rational and does not take into account emotional aspects. Effects of implementation choices on rendering speed should be instead taken into proper account: we tried different approaches and found that texture mapping on simple 3D shapes provided the best compromise between speed and quality of visualization. The texture mapped product visualization we adopted can be seen for a sample product in Figure 1 (before applying the guideline) and Figure 2 (after applying the guideline). We have determined that the effectiveness of the guideline transfers to VR stores with a between-groups experiment. Two groups of users shopped in two versions of the store: a control version where no product was visualized as massified, and a version where some of the previous products became massified. The experiment produced statistically significant results showing that the shopping behavior of buyers was affected by the change: sales of the same product increased when it was shown in massified format. EXAMPLE 2: WALKING PRODUCTS In (Chittaro and Coppola, 2000), we proposed a navigation aid based on 3D animated representations of products that move through the store and go to the place where the corresponding type of products is. In particular, the animated products we implemented walk towards their destinations and will thus be called walking products (WPs) hereinafter. A buyer in a VR store sees a number of WPs wandering around (as shown in Figure 3): if (s)he is looking for a specific type of products, s(he) has just to follow any WP of that type and will be quickly and easily lead to the desired destination. The specific path followed by the WP to accompany the customer to his/her destination is chosen by taking also into account the

4 Figure 3 Four Walking Products (Chittaro and Coppola, 2000). merchandising strategy of the store. If the buyer wants to stop along the way and take a look at other products, s(he) can count on the fact that WPs will be always available to lead her/him to the original destination. A proper balance among a number of WP design aspects has to be found to effectively give the impression of a live environment inhabited by WPs, while achieving the goals of a navigation aid for e-commerce sites. Detailed design choices are discussed in (Chittaro and Coppola, 2000). The navigation aid we proposed has several advantages over those adopted in current VR stores, i.e., signs, maps, and teleporting. Signs do not give a detailed enough indication of product location; maps impose on the buyer a translation effort from their exocentric perspective to her/his egocentric one; teleporting (i.e., being taken immediately to product location, e.g. by choosing from a menu) brakes the continuity of the shopping experience and limits the possibility of looking at other products. On the contrary, WPs: (i) support product finding in a easy, natural way: one has just to follow the WP as (s)he would do with a friend or a salesperson in a real shop; (ii) increase the number of products seen by the customer, especially those which are relevant for current merchandising strategies, (iii) convey the feeling of a living place and contribute to satisfy the need for interactive experiences typical of emotional buyers. We have evaluated the effectiveness of WPs for a product finding task in a between-groups experiment with one two-level independent variable (presence/absence of WPs in the store). The experiment produced statistically significant results showing that the time spent by buyers to find a product when WPs are present is less than 40% the time spent when WPs are absent. Moreover, user questionnaires confirmed that buyers found WPs fun and easy to use. A more thorough presentation of the WP approach is provided in (Chittaro and Coppola, 2000). In a following subsection, we will instead extend the WP idea to the context of 1-to-1 e-commerce applications. USING VR STORES AND WPs IN 1-TO-1 E-COMMERCE While in the previous Section we have focused on usability issues of VR stores from two different perspectives, in this Section we extend our attention to a further improvement, i.e. making the VR store adaptive. We propose and discuss some ways of automatically personalizing the VR store using the profile of the customer who is visiting it. Some notes on building the customer model In order to adequately support personalized interaction, a model of the buyer is needed. The two main issues to deal with are: (a) what should this buyer model contain? (b) how to build the buyer model? These two issues are not independent. For example, if one decides to build the buyer model by asking

5 the buyer to fill a form, only a limited amount of information can be acquired (buyers would not be able or would not like neither to fill large forms, nor to provide some personal details and preferences). Three possible approaches are: - Have the buyer fill a form that asks for typical information (buyer's gender, age, occupation, ), plus some specific information (interests on certain product categories, level of experience with certain products, ). A useful form is more easily designed for sites that sell specific products (e.g., cellular phones, records, ) than general sites (malls, supermarkets, ). - Exploit demographic profiles (available on the market) that give detailed and readily available information on the different categories of buyers, and can be used to make predictions about interests, preferences, and behavior of a category of customers. - In case the buyer is not a first time visitor of the site, automatically build (update) the model by considering data (which can include purchases made, number of visits, searches made, time spent looking at specific products, ) recorded on his/her past visits. However, since the three solutions can complement each other, a mixed approach allows one to obtain a more complete buyer model. For example, the approach proposed by (Ardissono and Goy, 1999) mixes the first two solutions: it asks the buyer to fill a form, and then assigns him/her a stereotyped profile by matching the form data with a data base containing an hierarchical taxonomy of stereotypes clustering the properties of homogeneous customer groups. An approach of the third type is adopted in (Jording and Michel, 1999): the system tries to tailor future presentations of products by monitoring past buyer choices in terms of kind of medium presentation chosen (e.g., text, graphics, video, ), downloading time (e.g., does the buyer interrupt downloading of pictures?), and content (e.g., what is the time spent by the buyer on a specific element of the presentation?). Personalizing the type of interface: VR vs. traditional While a VR interface can be appealing to emotional buyers or to buyers who are not expert in using computers, it must also be said that it is definitely not suited for some other categories of buyers. For example, the buyer could be very experienced in the use of computers and prefer to use an advanced search engine, or s(he) can be a rational buyer who prefers to see plain text tables of product features, or s(he) can be using a very slow computer which does not support graphics at a reasonable speed. We thus believe that a VR store should be provided in addition to more traditional interfaces to an e-commerce site. Said that, we concentrate in the following on the personalization of specific aspects of the VR store interface. Personalizing WPs The navigation aid we have previously presented (WPs) can be tailored to the buyer profile in several ways, such as the following. WP choice. The specific WPs shown in the virtual store can be chosen according to user profile to attract her towards the targets which are more likely to interest her, such as: (i) specific product categories, or (ii) new products not seen in previous visits, or (iii) special offers and discounts, or (iv) products which s(he) is likely to have seen recently advertised in her preferred magazines and TV shows. Dynamical WP set. The size and the members of the set of WPs can be changed over time. A first choice is about WPs presence or absence in the virtual store. This choice could be both left to the buyer and guided by previous interaction history (e.g., WPs can be present in the first visits, and eliminated after a given number of visits if the specific buyer does not seem to use them). When WPs are present, there are many additional possibilities for dynamically adapting the set of WPs. For example, WPs for the product categories that the buyer finds (or has learned to find from previous sessions) without help

6 are eliminated. This allows one to: (i) save store space and rendering resources which would be wasted by unnecessary WPs, (ii) introduce in the set new WPs which can attract the customer towards potentially promising purchases, (iii) create a continuously changing environment that keeps buyers attention alive (a static environment could easily become boring for the emotional buyer after some visits). Purchasable WPs. An idea that emerged during the evaluation of WPs was to give the buyer the possibility to purchase directly by clicking on WPs, without walking up to the corresponding shelve. In this case, the products displayed by some WPs can be chosen according to the products the buyer purchases more frequently. WPs can thus act as a "favorites" list of products which can quickly lead to product purchase. WPs behavior. Many features of WP behavior can be adapted to user profile. For example, there should be the possibility to vary WPs speed (e.g., slow for the first visits, then progressively faster; or slow for buyers that are classified as novice users of computers, and faster for other categories), and WPs paths (e.g., dynamically determining them according to the parts of the store that are more likely to interest the buyer). Personalizing Audio During the evaluation experiments of our VR store, many users explicitly complained about the absence of sounds and background music. As witnessed by the growing attention devoted to audio aspects in the videogame industry, sound and music can greatly enhance immersion. In the case of store background music, unlike real stores, we have the possibility to play music according to buyer's preferences (country, classical, rap, metal, or anything else). Moreover, WPs animation can be partly influenced by the type of music played to better relate visual and auditory stimuli. Background music can be interrupted by voice messages as in real stores (e.g., informing of a special offer), with the advantage that the chosen voice messages can be targeted to the specific user, both in the promoted product, in the type of voice, and in the choice of words (e.g., a teenager and a elder user would certainly prefer very different kinds of message style, voice, and emphasis). WPs themselves could directly talk with the customer to deliver the intended voice messages. Personalizing Store Organization and Look The adaptation of the VR store to the specific customer could be taken to the point of changing shop layout, organization, and look. For example, product position in shelves can be automatically rearranged according to different criteria, such as function, price, brand, chromatic similarities, or any other criteria that the buyer profile can predict as relevant for the current user. One can also consider to allow the buyer to directly adapt the shop according to her/his own taste. In this case, key issues are: (i) what parts of the shop should be customizable?, (ii) to what degree they should be customizable? Several options are indeed available, such as the possibility of freely changing the position of shelves in the shop, the possibility of choosing which product categories have to be shown, the possibility of choosing the position of products on given shelves (e.g., the "favorites" shelf), the possibility of changing colors and shapes of the shelves, etc.

7 CONCLUSIONS AND FUTURE WORK This paper has examined the current status of our VR store project, including a detailed discussion of what is going to be experimented in the short term to make the interface adaptive. A more long term goal is to explore the possibility of significantly extending the competence of WPs, enriching them with further user assistance functionalities besides that of leading buyers to specific parts of the shop. These extended animated characters would become shop assistants, capable for example of addressing simple buyer questions, or performing product presentations. Some issues in adding animated presentation characters to a Web interface are discussed in (Rist, André, and Muller, 1997). Another aspect we intend to investigate is the impact of allowing buyers to meet and interact in the VR store. From this point of view, it is interesting to note that adding this social dimension can conflict with personalization aspects, limiting the possibilities of 1-to-1 e-commerce. For example, if multiple users have to walk together and interact in the same store, the customization of the several features mentioned in this paper cannot target anymore the specific profile of a single buyer. Trying to find the best compromise which maximizes the match with the different user profiles can be a possible solution, but it would not be easy to implement, considering that the set of customers in the stores continuously changes. REFERENCES Ardissono L. and Goy A. Tailoring the interaction with users in electronic shops, in Proceedings of UM99: 7 th International Conference on User Modeling (Banff, Canada, June 1999), Springer Verlag, Chittaro L., and Coppola P. Animated Products as a Navigation Aid for E-commerce, in Proceedings of the CHI2000 Conference on Human Factors in Computing Systems, (The Hague, The Netherlands, April 2000), ACM Press. Jording T. and Michel S. Personalized Shopping in the Web by Monitoring the Customer, in Proceedings of the Active Web 1999 Conference (Stafford, UK, January 1999), Serco Ltd How to Design a customer friendly on-line store, Steiger S., Stolze M., and Good M. Beyond Internet Business-as-Usual. SIGCHI Bulletin 30, 4 (October 1998), Rist T., André E., and Muller J. Adding Animated Presentation Agents to the Interface, in Proceedings of the IUI97 Conference on Intelligent User Interfaces, (Orlando, FL, USA, January 1997), ACM Press, Tilson R., Dong J., Martin S., and Kieke E. Factors and Principles Affecting the Usability of Four E- commerce Sites, in Proceedings of the 4 th Conference on Human Factors and the Web (Basking Ridge NJ, June 1998),