Applied Ergonomics 31 (2000) 609 } 619

Modeling the customer in electronic commerce

Martin G. Helander *, Halimahtun M. Khalid
School of Mechanical and Production Engineering, Nanyang Technological University, Singapore 639708 Institute of Design and Ergonomics Application, Universiti Malaysia Sarawak, 94300 Kota Samarahan, Sarawak, Malaysia Received 31 December 1999; received in revised form 22 June 2000; accepted 1 July 2000

Abstract This paper reviews interface design of web pages for e-commerce. Di!erent tasks in e-commerce are contrasted. A systems model is used to illustrate the information #ow between three subsystems in e-commerce: store environment, customer, and web technology. A customer makes several decisions: to enter the store, to navigate, to purchase, to pay, and to keep the merchandize. This arti"cial environment must be designed so that it can support customer decision-making. To retain customers it must be pleasing and fun, and create a task with natural #ow. Customers have di!erent needs, competence and motivation, which a!ect decision-making. It may therefore be important to customize the design of the e-store environment. Future ergonomics research will have to investigate perceptual aspects, such as presentation of merchandize, and cognitive issues, such as product search and navigation, as well as decision making while considering various economic parameters. Five theories on e-commerce research are presented. 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Web customer model; Electronic commerce; Usability; Design principles

1. Introduction Shopping on the Internet is a fairly recent phenomenon. Although some e-stores were created in the 1980s, the great majority are only a few years old. There are several problems in designing an e-commerce facility. The "rst problem is that the computerized medium makes it possible to display products in many new ways. There can be links organized by product, by brand, and by functionality. There can be written lists of products for sale, and there can be store environments set up as a virtual reality simulation of aisles, shelves and products. There can be product illustrations in color and multimedia, and streaming video with consumer testing. There can be personalized stores, which utilize `cookiesa to display only items of interest to individual customers. In analogy with Herbert Simon's concept `Sciences of the Arti"ciala this is the `Marketplace of the Arti"ciala (Johnson et al., 1999). These stores are symbolic and they are unconstrained by the physical prescriptions of a real store environment. Anything is possible. This, then, makes it di$cult to design the ideal store.
* Corresponding author. Fax: #65-791-1859. E-mail addresses: mahel@ntu.edu.sg (M.G. Helander), mkmahtun@idea.unimas.my (H.M. Khalid).

The second problem is that there are many forms for buying/selling, and the types of economic transactions that will prevail in the future are yet unclear. Presently, the following types of transactions dominate: 1. Fixed price. Conventional e-stores with goods from one retailer, 2. Auction houses. Customer may bid on items, such as airplane tickets (e.g. www.priceline.com), or on antiques and other artifacts (e.g. www.eBay.com), 3. Infomediaries. These pool the merchandize from several stores or manufacturers under one umbrella, for example, Chemdex (www.Chemdex.com). 4. Hybrid stores. NexTag (www.NexTag.com) sells computers and electronics items to customers who submit a bid. It is also an infomediary, with a supply chain consisting of other well-established e-stores. The economic transaction is di!erent in di!erent store environments. An auction or a bid will imply repeat visits with strong customer motivation. The price is essential, but strategy in bidding may be equally important. The future e-store will o!er many services that are made possible only because the environment is programmable and interactive. For example, Home Depot, an American chain of stores for home improvement

0003-6870/00/$ - see front matter 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 0 0 3 - 6 8 7 0 ( 0 0 ) 0 0 0 3 5 - 1

610

M.G. Helander, H.M. Khalid / Applied Ergonomics 31 (2000) 609 } 619

o!ers special services to small contractors. A contractor may log in and specify a job. Home Depot will suggest what he needs to buy, how to schedule the job, and "nally deliver the materials to the construction site `just-intimea. Home Depot will also supply experts to solve special problems. The company has become a service provider. At the same time it acquires a wealth of information about customers, and can, therefore, deepen the relationships. The suppliers to Home Depot are also linked to the web site, and they are fully informed of all customer transactions. Therefore, they can deliver required materials to Home-Depot just-in-time. The advantage for Home Depot is that the company can carry less stock (The Economist, 1999a). The largest e-auction house is www.eBay.com . At the time of writing this article there were about 4 million items for sale and 2 billion web pages were accessed per month. eBay has incorporated many essential design features on their web pages. Items for sale are well documented with "gures and verbal descriptions, which are supplied by the seller. Buyers seem to expect this and avoid buying items that are not well documented. eBay has established a trusting environment. After each purchase, buyers are asked to rate sellers and sellers are asked to rate buyers. Past ratings can be read by prospective buyers and sellers, and they form the basis for trustful interactions. E-auctions are not only for antiques and second-hand goods. They have also become a mechanism for price setting. In conventional price-setting auctions, the items were limited mostly to Agricultural and Commodity Business and Financial Securities. E-auctions can, however, solicit a wide range of bids from many people around the world. Thereby, prices can be established even for unusual goods, such as radio bandwidth licenses and rail operating franchises (The Economist, 1999b). An infomediary acts as a common node for many suppliers and many customers. Chemdex, for example, is a company that markets 1.4 million chemicals and reagents from 2200 suppliers. The buyers are mostly chemical laboratories, which in the past used individual product catalogs. In Chemdex all information is pulled together on one web site. Presently, there are 30,000 transactions per day } a 10-fold increase in 12 months. There are several reasons why Infomediaries are successful including: 1. Internet shifts the power from the seller to the buyer. Buyers want one stop shopping with accurate trustworthy information. The traditional manufacturer/ supplier can, however, not o!er disinterested advice. This opens up opportunities for a third party * the infomediary. Chemdex does not manufacture any drugs, and there is no con#ict of interest with the companies that are represented by Chemdex.

2. Internet reduces transaction costs. An electronic banking transaction is 1 cent, compared to 27 cents for an automatic teller machine (ATM). Processing an airline ticket costs $1 compared to $8 through a travel agent. 3. Greater speed, range and accessibility of information. Infomediaries can distribute information to customers around the world.

2. Design of web environments For human}computer interaction (HCI) design, as well as Web design, there are many general guidelines (e.g. Helander et al., 1997; Nielsen, 2000). Although many of them apply, e-commerce is a very speci"c domain. Many motivational and attitudinal factors are speci"c to ecommerce and a!ect web page design. The main goal is to attract customers, support customer decision-making, and encourage customers to return. Given the economic signi"cance of e-commerce, and the economic rewards of creating good web sites, researchers in Human Factors and HCI will have their hands full for several years to come. Below, we "rst describe user characteristics * who are the buyers? This information is followed by statistics and economic trends of e-commerce, which validate our interest. A systems model for research is then presented. Existing research is described and summarized. Di!erent types of e-commerce are outlined, and we speculate how they di!er from a design perspective. Finally, we discuss theories for research that may drive the theoretical research interest for a few years to come.

3. Who are the buyers? The Internet is user-driven (Tilson et al., 1998). The interaction is initiated by, and terminated by, the user. Throughout the transactions, the user is only one click away from leaving. Unlike physical stores, there is no sales person and no social pressure to stay. It is easy to "nd substitute stores and compare prices. In fact, intelligent search engines are available for price comparison, and some infomediaries (e.g. NexTag.com) incorporate price comparison in their web site. Because of this situation vendors are forced to take users seriously, and it is imperative to understand who the customers are. Bellman et al. (1999) reported a study on Web customer demographics, attitudes toward on-line shopping and predictors of on-line behavior. About 10,000 web users completed a questionnaire during the time period December 1997}June 1998. Ninety-two percent of the respondents were from North America, and the rest from Europe and Asia.

M.G. Helander, H.M. Khalid / Applied Ergonomics 31 (2000) 609 } 619

611

Respondents typically spent more time on the web at home than at work. Twenty-one percent spent over 20 h browsing per week from home. At home the Internet was used regularly to read news (19.1%), for entertainment (10.8%), "nding software (6.7%), banking (6.4%), and personal "nance and investment (3.8%). Only 2.9% used the Internet for product information, and 3.1% said they used the Internet for shopping. Forty-three percent had never bought anything on-line. The annual purchase value was about US$ 200 in the USA, US$ 240 in Europe, and US$ 160 in Asia. This amount was not a!ected by personal income. Multiple regression was used to analyze which variables could predict on-line buying. The following variables predicted 45% of the buying: 1. 2. 3. 4. 5. Looked at product information on the Web. Months of experience with on-line activities. Number of daily e-mails. Work on-line at work every week. Read news on-line at home.

a world market has inspired several start-ups in Western Africa. The success of the Internet in India is promising, and demonstrates what a so-called poor, albeit democratic, society can achieve.

4. User attitudes * not only security! It is common that users have strong attitudes about e-commerce security. Tilson et al. (1998) asked users to rate 40 listed items in terms of their e!ect on the decision to purchase. The top seven items are listed in Table 1. Although several of the items dealt with usability, the top two were related to security. Participants were then asked to list the three most important deterrents for them to buy. Across the sample the most important were, in order: (1) credit card security, (2) easy return/exchange methods, (3) detailed descriptions of items, (4) price, (5) secure personal information, (6) pictures of merchandize, and (7) simple to search. According to Kehoe et al. (1998), about 40% of users are concerned about security, although only 1.9% actually had a bad experience in e-shopping. About 50% stated that the use of encryption was an important factor in deciding to submit personal information. These factors did not, however, in#uence buying behavior. Jarvenpaa and Todd (1997) interviewed 220 e-shoppers with respect to their attitudes to e-shopping. Three factors were statistically signi"cant: product perception, shopping experience and risk. The perception of a store is related to product variety, product quality, and price. Customers were impressed with the variety of stores and range of unique items that could be found. Shopping opportunities for brand names were however limited. Low pricing was important. However, pricing was less important than variety. Although they found on-line browsing of products enjoyable, it was di$cult to navigate to speci"c items. Goal-directed shopping was not supported. The shopping experience saves time and effort. Some of the respondents also thought it should be compatible with consumers' lifestyle, and it should be fun! In e-shopping the customer takes two main risks. There is a performance risk that the item will not

To summarize the "ndings, buyers had a `wireda lifestyle and they typically had used the Internet for several years. Another important aspect was `time starvation.a E-buyers have little time, and mail orders from catalogs are perceived as too slow. Demographics (such as income and education) do not seem to in#uence the buying habits. Once people are on-line they buy for the same amount. Demographics do, however, a!ect who uses the Internet. A follow-up study, one year later with the same population veri"ed that `wireda lifestyle and `time starvationa were the two most important predictors of online shopping (Lohse et al., 2000). A year later the median purchase had increased from US$50 to US$86, and the average number of purchases from 4.3 to 7.4. By using these data and extrapolating the trends across all buyers in the USA, the authors predicted the on-line shopping volume of US$ 46 billion for the year 2000. The number is in agreement with predictions by Forrester Research (Lohse et al., 2000). The trend for some developing countries is quite similar. Wilson (1999) distributed a questionnaire on-line in India, China and Chile. He reported that 33% of respondents had bought items on the net during the last 3 months. One may speculate whether the Internet will adversely a!ect income distribution to the poor parts of the world. The sub-Saharan African continent, with a few exceptions (e.g. South Africa), does not participate. This is not the fault of the Internet, which is open and democratic. It is rather the fault of governments who disallow free access to information and of economics. On-line or telephone costs can be prohibitive. Once trade politics and information access are liberalized there are opportunities for e-commerce in all countries. In fact the potential of

Table 1 Assessments of items that a!ect the decision to purchase electronically, from a list of 40 (a scale of 1}7 was used) 1. 2. 3. 4. 5. 6. 7. Credit card security Feedback that the order was received The ability to edit purchase list Easy methods for exchange/return Knowing where to "nd the item Trial purchase without being a member Personal information kept private 6.94 6.75 6.63 6.56 6.50 6.38 6.31

612

M.G. Helander, H.M. Khalid / Applied Ergonomics 31 (2000) 609 } 619

function as expected: this risk is similar for catalog and telephone shopping. There is also a personal risk due to credit card fraud, which was mentioned by 30% of the respondents. Based on the results and a thorough literature review, Jarvenpaa and Todd (1997) gave several recommendations for developers of e-stores: 1. O!er discounts, except for unique products. 2. Focus on products that have low delivery cost. 3. Benchmark the e-store against traditional retail stores and catalog stores. 4. Emphasize brand name products and product quality. 5. Reduce shopping e!ort by providing search techniques. 6. Respond promptly to questions. 7. Provide no-cost/no-hassle return policy. 8. Provide rich product descriptions including images and words. 9. Emphasize security measures. 10. Provide customer testimonial. In a follow-up, Crisp et al. (1997) analyzed the attitudes of e-shoppers using the same sample as Jarvenpaa and Todd (1997). Attitudes to e-commerce were positively correlated with attitudes toward computers (r"0.26, p(0.01) and prior web experience (r"0.32, p(0.01). One interesting "nding was that education was negatively related to e!ort (r"!0.15, p(0.05). This may be interpreted as the greater the education level, the higher the personal time cost, and as a result less time is spent searching. Kirakowski et al. (1999) presented attitudes of 250 web users concerning factors that make a web site interesting and useful. The most important items were: (1) attractiveness (this web site is attractive), (2) control (easy to go from one part to another), (3) e$ciency (easy to "nd items), (4) helpfulness (all items are clearly labeled), and (5) learnability (all material is easy to recall). According to a survey of users in the USA, over half of 10,000 respondents had purchased items on-line (Kehoe et al., 1998). The web was used for personal shopping mainly due to: convenience (65%), availability of vendor information (60%), no pressure from sales people (55%), and saving time (53%). Keeney (1999) used value propositions to quantify customer priorities with respect to design features in an Internet store. From interviews of 100 persons he developed value propositions which were used to derive fundamental objectives related to Internet commerce. These were: product quality, cost, time to receive product, convenience, time spent, privacy, shopping enjoyment, safety, and environmental impact. A means-ends objectives network was then developed to illustrate how the fundamental objectives are related. This methodology can be used to represent customer priorities and evaluate trade-o!s in design of an e-store.

5. Systems model for design of e-commerce The systems model in Fig. 1 serves as an overview that can be used to conceptualize research and design issues in presenting a productive and useful store environment. There are three subsystems: store environment, customer, and web technology. The arrows in the "gure show the direction of information #ow. Note that the system is dynamic with feedback. Once a decision has been taken, the customer supplies input manually through a keyboard or vocally through voice recognition. The display will then change, a new image is presented, and a new decision must be made. Thus e-commerce may be conceptualized as a series of decisions potentially leading to a purchase and payment. We will "rst explain the model below, and then present relevant research. We will then discuss the "ndings and suggest some directions for future research. In Fig. 1, the customer is of central interest. This is a natural focus, since the interaction takes place at the user initiative (Vora, 1999). The user has a stronger role in e-commerce than in many other technologies. The customer part of the system consists of the three phases in human information processing: perception (of the details in the store), decision making and control (action). These processes are modulated by several customer characteristics including individual needs, attitudes, competence and motivation, which explain why some customers are willing to purchase and others are not. Learning, habituation and addiction may motivate buyers to revisit certain e-stores time after time. These personality aspects have not yet been well researched. Recent models in psychology may be used as a theoretical foundation (Rabin and O'Donoghue, 1999). Mandel and Johnson (1999) experimented with the e!ect of priming on customers' behavior. They investigated di!erent patterns of the `wall papera in the e-shop. They found that green wallpaper in which dollars were embedded increased the sensitivity to price. Similarly, a #ame-like background increased the sensitivity to safety issues. The results were recon"rmed in a replication of the experiment with two types of subjects. Menon and Kahn (1998) also demonstrated the e!ects of priming. They showed that Internet shopping is in#uenced by characteristics of products encountered early in a shopping experience. Speci"cally, highly novel products lead to less exploration and fewer purchases of other novel products during the remainder of the shopping experience. Kim and Moon (1998), in a well-controlled study, investigated the e!ects of di!erent web site layout factors on customers' trust in e-banking. Several design features were related to emotive factors including symmetry, awkwardness, and elegance. Johnson et al. (1999) showed that there is a strong learning e!ect of visiting an e-store. Twenty visits to

M.G. Helander, H.M. Khalid / Applied Ergonomics 31 (2000) 609 } 619

613

Fig. 1. A systems model for human factors research in e-commerce. There are three subsystems: web environment, customer, and web technology. The issues noted in the boxes represent independent design variables. Dependent variables are referred to in the bottom of the "gure under headings of measurement of outcomes and satisfaction.

a facility will halve the search time (from about 400}200 s). They noted that just as knowledge of the layout of a supermarket creates customer loyalty, so does knowledge of an e-commerce facility. Such knowledge will then act as a barrier to switching to another store. Customers' willingness to purchase is a!ected by the design of the store environment. Important design features include: product information, navigation facilities, ease of purchasing, promotions, and feedback information. These design features constitute the so-called `a!ordancesa (Norman, 1990) and customers' decisions are generally facilitated by good a!ordances. However, experience is helpful. As an example, an experienced customer will perceive the store environment di!erently from an inexperienced customer. He/she may be looking for feedback about the seller, and associations with the TrustE organizations. Once the customer has identi"ed these design features, he/she will search for the merchandize. A store with great performance a!ordances invites purchase decisions and thereby commits the customer. Thus a customer may decide to visit a section of a store by following a well designed link, which o!er high-performance a!ordance. Following the decision to visit, a customer may decide to navigate in the store to "nd the item in question. Errors in navigation are common, and

they discourage customers. Store design variables, such as number of clicks to product, must be considered. A store that lacks performance a!ordances will discourage sales. As Fig. 1 illustrates, there are many measures of outcome or dependent variables that can be used in research * the decision to purchase is but one. A successful purchase comes about as a consequence of a series of "ve decisions: the decision to visit, the decision to navigate, the decision to buy, the decision to pay, and the decision to keep. Each of these should be supported by information that can facilitate e$cient decision-making. The decision to purchase depends, among other things, on how the product is presented. Poor photographs or images (with low-performance a!ordance) discourage. The decision to purchase is, however, not "nal. There is yet the decision to pay. Gunnerson (1999) reported that 27% of customers abandon the e-commerce transaction because they did not want to take the time to "ll in the on-line form. A decision to not purchase is potentially a!ected by a variety of factors. The price may be too high. It may be the wrong merchandize. Customers may not "nd what they look for. Customers may not be aware of what is in the store. The store interface design is poor. It is not

614

M.G. Helander, H.M. Khalid / Applied Ergonomics 31 (2000) 609 } 619

a trusting environment. As mentioned, many potential customers are discouraged by poor interface design, including poor interactivity, long waiting times, and navigation problems. Faced with low sales volume, it is di$cult for a storeowner to understand why items are not selling and what can be done to improve sales. Much research is still needed to distinguish important design features that facilitate sales in an e-store. The system in Fig. 1 suggests a strategy for research by operationalizing dependent and independent variables that a!ect e-commerce. Dependent variables include: number of purchases, number of store visits, time to "nd merchandize, time to pay, satisfaction with store, satisfaction with purchase, and number of errors in navigation. Independent variables primarily relate to design features * the store environment and the web technology. The customer modulating variables can be treated as either independent variables or blocking variables. Some of these variables may be di$cult to measure, since customers may not want to submit personal information. Measures of (dis)satisfaction are fairly easy to obtain through interviews or questionnaires. As indicated in Fig. 1, these measures depend on comparisons of customer needs and attitudes with the measures of outcome. For example, a person with well-articulated needs for a new product will be disappointed if the o!ered item does not meet expectations. A person with a positive attitude to a particular store environment (i.e. a prestigious store) may accept di$culties in navigation, whereas a person with neutral or negative attitude will leave the store. Web technology, the remaining subsystem, refers to technological features both in browsers and in the store environment. For example, a store environment may be designed to be used by visually impaired customers (Rogers and Rajkumar, 1999). The web technology may then incorporate synthetic speech (auditory display) or braille display/input. Likewise, there may be special search functions or agents built into the browser to present price comparisons. While web technology is important, it is not the focus of this paper.

environments (Helander et al., 1997). Examples include: E Simplicity: Do not compromise usability to gain function. E Support: User should be in control with proactive assistance. E Visibility: Make objects visible and intuitive to control. E Reversible action: Make actions reversible. E Feedback: Provide visible (or auditory) feedback on all user actions. E Accessibility: Make all objects accessible at all times. E Personalization: Allow the user to customize the interface. Many studies have dealt with usability of web pages. Since the purpose of this paper is e-commerce, a rather speci"c application, we will only make brief reference to a few sources. For general information, an excellent textbook is: Forsythe et al. (1997). Several interesting papers have been published. Lynch and Palmiter (1999) presented a systems model of web users. Larson and Czerwinski (1998) analyzed web users' cognitive load and the implications for web page design. Bellman et al. (1999), Marcus et al. (1999), and Kirakowski et al. (1999) addressed user-centered web site design, while Tiller and Green (1999) described a work procedure that is likely to result in a better web page. A well-designed web site combines many design elements: Good layout, e!ective graphics, exciting content and intuitive navigation (Fuccella and Pizzalota, 1997). Nielsen (2000) provides abundant and excellent examples of good and poor web page design. The IBM Ease of Use Group studied four e-commerce stores: two clothing stores and two computer stores (Tilson et al., 1998). Some of the common design problems were: 1. Poor help on how to add an item to the shopping list. 2. Poor feedback on which items remained on the shopping list. 3. Little clarity about what stage users needed to register/log in. 4. Di$culty to buy a second item. This is analogous to paying, then walking to the storefront, "nding the new department, and paying again. 5. There should have been more pictures of clothes. 6. Drop down menus were di$cult to use. 7. For one site, seven clicks were required from the home page to the product. Four clicks were su$cient for another site * much better. 8. There were numerous, tiresome security messages. 9. Long lists of merchandise took a long time to scroll down. 10. Clicking on the wish list produced an empty page with the words `log ina. Users preferred to browse

6. Ease of use The success of e-commerce is highly related to ease of use. A study by Hurst (1999) showed that 39% of shoppers in a user test failed in their buying attempts because sites were too di$cult. This rate of failure was predicted to lead to a loss of sales of US$ 6 billion during the Christmas season 1999. In addition 56% of the search attempts failed, which corresponds to approximately $8 billion lost. Many general principles for design of human-computer interfaces apply to the design of e-commerce

M.G. Helander, H.M. Khalid / Applied Ergonomics 31 (2000) 609 } 619 Table 2 Comparisons between retail stores, paper catalogs and on-line stores (source: Spiller and Lohse, 1998) Retail store Sales clerk service Store promotions Store window displays Store atmosphere Aisle products Store layout Number of #oors Number of outlets Checkout cashier See and touch of the merchandise Paper catalog Printed product descriptions, sales clerk on the phone Special o!ers, sales catalog Front/back covers, copy quality, product arrangement Graphics/print quality, product arrangement First 2}4 spreads, middle spread Page arrangement Catalog organization Frequency of mailing Phone/mail order Limited to image quality and description Online store

615

Product descriptions, search functions, gift services On-line games and lotteries Home page Interface and graphics quality, product arrangement Featured products on each hierarchical level of the catalog Screen depth, browse and search functions, image map, indices Hierarchical levels of the catalog Number of web links to a catalog Shopping basket or order form Limited to image quality and description, future potential for sound and video

the store before logging in and establishing an account. Ideally, one should adapt the design of web sites to suit individual customers. There are two ways of doing this: covert and overt. Covert adaptation is when the customer is not being told. One common technique that is employed by, for example, www.fashionmall.com is to track the visitor's trip through the site to determine what areas they visited most frequently. Such monitoring is recorded using `cookiesa, which are small text elements placed on the customer's hard drive by the visited site. The next time the person visits the web site, it can be customized to "t the customer's past behavior (an example is given at www.magnet.com). Overt adaptation is through the use of surveys and questionnaires. Customers are asked about their preferences in web page design. Another overt technique is to let the individual users customize a site using web tool, such as CGI scripts. Thomson Investor Services (www.thomsoninvest.net) uses this technique to let clients customize their investment web site to "t their investment priorities.

7. Design of the Internet store Below, we will present research relating to two design aspects: How to present the merchandize and the store environment, and how to design advertisements. 7.1. Design of Internet stores and presentation of merchandize To design Internet stores it is useful to understand how the task in e-commerce is di!erent from regular retail stores and from mail order. Spiller and Lohse (1998)

noted that there are several analogies between a regular (physical) store, mail order, and e-stores (see Table 2). The comparisons emphasize the need for real world metaphors in the e-store. Many functions that seem selfevident in a real store, such as aisle products (which sell quicker), have no correspondence in the e-store, so new concepts are necessary for fast promotion. Such concepts are intimately associated with ease-of-use. Lohse and Spiller (1998) classi"ed the store environment of 28 very large e-stores using 32 independent variables. The following independent variables were classi"ed and measured: description of merchandize, service o!ered, promotion of items, convenience and interface characteristics. These variables were used to explain two dependent variables: number of store visits and number of sales. The greatest number of store visits was in stores with a FAQ section (R" 0.451). One possible explanation is that stores with many e-mail questions felt obliged to implement a FAQ facility. However, the availability of a FAQ section did not a!ect sales. Sales were mainly a!ected by a good product presentation. A basic product list consists of a scrolling menu. It would be much better to provide "gures of products. A list with buttons to navigate to product descriptions improved sales the most (R" 0.570). The researchers recognized several methodological weaknesses in their study: Correlation analysis does not infer causation, and due to the small sample size the results must be interpreted with caution. None-the-less, the outcomes of the study are interesting. 7.2. Design of advertisements It is well known that size of advertisement is important for paper based ads * the larger the better. However, this is not so with web-based ads. Hoque and Lohse (1999) found that online ads with extensive graphics have

616

M.G. Helander, H.M. Khalid / Applied Ergonomics 31 (2000) 609 } 619

Table 3 How users "nd web sites, and factors which drive repeated visits Users found web sites through (%) Factors driving repeat visits (%) Search engines E-mail messages Links from sites Word of mouth Magazine ads TV commercials Periodic articles Vendor catalogues Newspaper ads Banner ads Radio ads Mail ads 57 38 35 28 25 14 11 11 9 7 2 2 High-quality content Ease of use Quick download Updated frequently Coupons and incentives Favorite brands Cutting-edge technology Games Purchasing Capability Customizable content Chat and BBS Other 75 66 58 54 14 13 12 12 11 10 10 6

long download times and are not preferred. In yellow pages the customers select anywhere in the listing and a local business is preferred over non-local. Not so with web pages, where the top listing is preferred. This may have the e!ect that the customer unwillingly selects a website from some other country. Customers may "nd an e-store following an advertisement such as a banner. But they will not return to the same store unless usability is good. In comparing the e!ectiveness of advertising and usability, Hurst (1999) concluded that one dollar spent on advertising yielded US$5 in incremental revenues. Usability is more important. One dollar invested to create a good customer experience yielded $60 in incremental revenues. Banners, the most common form of on-line marketing, according to a report by Forrester Research, are not e!ective (McLatchie, 1999). Table 3 shows how readers "nd a web site and the factors that make them return to the same web site. The results for banners are in agreement with an experimental study published by Benway and Lane (1998). They performed two experiments and demonstrated that web users do not pay attention to banners. Norman (1999) commented that `banner blindnessa may be expected, since web users are goal-driven. Since banners are usually not task related, users have been conditioned to not look at them. 8. Current theories Several theories from engineering psychology and HCI apply to design of e-commerce. However, we have not found many published studies. Below, we will resort to the description of relevant theories, most of which have been applied in management research. 8.1. Theory of yow To create an enjoyable experience in e-commerce, several researchers, particularly in management and

business schools, have made reference to Csikzentmihalyi's (1975, 1990) Theory of Flow and Work#ow (see Nel et al., 1999; Johnson et al., 1999). Csikzentmihalyi (1975, p. 72) described the #ow state accordingly: `players shift into a common mode of experience when they become absorbed by their activity. This mode is characterized by a narrowing of the focus of awareness so that irrelevant perceptions are "ltered out; by loss of self-consciousness, by a responsiveness to clear goals and unambiguous feedback, and by a sense of control over the environment * it is this common #ow experience that people adduce as the main reason for performing an activitya. Trevino and Webster (1992) provided the following de"nition: `#ow represents the extent to which (a) the individual perceives a sense of control over the interactions with technology, (b) the individual perceives that his or her attention is focussed on the interaction, (c) the individual's curiosity is aroused during the interaction, and (d) the individual "nds the interaction interestinga. The goal is to design an interface which will improve user attention and make the user feel in total control, and free from distractions from other non-related tasks or distractions due to poor usability. Nel et al. (1999) did an experimental evaluation of 20 e-commerce sites. These were classi"ed in two ways: domestic/international and informational/transactional. Thirty-three students were used as experimental subjects. Domestic, informational facilities had better #ow than domestic, transactional sites, followed by global, informational and global transactional sites. The authors concluded that the greater #ow of informational sites is due to the fact that they require less e!ort than transactional sites. It is also easier to construct a site for a domestic community than an international community. 8.2. Cost of search Johnson et al. (1999) noted that the concept of `cost of search for merchandizea can be useful in e-commerce research. This concept has been well documented over the years in management research. The purpose is to reduce the search cost (time), thereby increasing customer welfare. The longer the search, the better the information and the lower the price of merchandize (or the greater the quality). The present authors note the similarities with concept of `speed-accuracy trade-o!a in visual inspection and manual labor (e.g. Wickens and Hollands, 2000). To reduce the cost of search, customers may take advantage of search agents, such as Anderson's Consulting Bargain Finder, which can "nd the least expensive site (Lynch and Ariely, 1999). Ha K ubl and Trifts (2000) examined two such aids, one provided a tool similar to elimination by aspect (eliminates alternatives that violate

M.G. Helander, H.M. Khalid / Applied Ergonomics 31 (2000) 609 } 619

617

one or several customer criteria, such as price, quality, color, etc.), and the other was a tool for calculating additive di!erences. Elimination by aspect signi"cantly improved the quality of decisions. NexTag (www.NexTag.com), an infomediary which sells computers, has solved the problem by incorporating the search mechanism into their web site. List prices and auction prices for all items on sale are readily available. In addition, graphs with the price development over the last 12 months are supplied. 8.3. Theory of seductive interface design One important challenge in theory and application is the design of seductive and fun interfaces. The main goal is to please the user rather than maximize transactions and productivity. The dual advances in home computing and the World Wide Web emphasize this aspect. The research in this area is just about starting. Kahneman (1999) brought psychological theory to `hedonic pleasuresa. The research on hedonic values and seductive interfaces will be a welcome contrast to concentration of safety and productivity, which have dominated human factors and ergonomics: from pain and performance to pleasure! Jordan (2000) recently arranged a conference on `Pleasure-Based Human Factors Designa. He identi"ed four conceptually distinct types of pleasure that a person can experience when using a product. Physiopleasure, has to do with the body and the senses, sociopleasures with inter-personal and social relationships, psycho-pleasure with pleasures to do with the mind (cognitive and emotional), and ideo-pleasure with pleasure to do with values. 8.4. Theory of economic trade-ow Theories in economic psychology can be used to predict customer needs. For example, the prospect theory introduced by Kahneman and Tversky (1979) could be developed to explain customer behavior in e-auctions by evaluating subjective losses and gains (see Helander and Du, 1999). 8.5. Theory of Internet addiction The success of Internet auctions is not only a matter of service provided. There is also an element of compulsive behavior and addiction or conditioning. Many users develop addiction to web sites with a compulsion to check: the news, "nancial information, and e-mail. Recent models in psychology may be used as a theoretical foundation to explain why and how users get addicted to the internet (Rabin and O'Donoghue, 1999).

9. Discussion 9.1. Attitudes There seems to be two dominant attitudes among e-shoppers: The fear of credit card fraud and the appreciation of usability. Credit card fraud is most likely not any greater in e-commerce than in regular commerce. Yet customers are suspicious of the new medium. There are analogies from experiences in the past. When robots were "rst introduced in factories 20 years ago there were concerns about a sociological impact on the shop #oor * that the personality mix between men and machines would become too intimate. When VDTs were "rst introduced there were great concerns about the radiation hazard. These are no longer threats, and the `scienti"ca debate for these if not for other concerns has ceased (Helander, 1995). As customers become more experienced with e-mail and the Internet, and as the technology improves, we predict that customers will feel safer. There will also be many supporting technologies including encryption and digital wallets. The other issue, the importance of usability, holds promise for our profession of ergonomics. In the future, usability professionals should be in great demand. Every single e-store or e-business needs usability design. Unlike Microsoft and MSWord, where global usability problems can be solved at the headquarters, e-stores have individual programming. Each store is unique, and usability problems are unique. Many developers of web pages have been mistakenly guided by the challenges of technology rather than by user needs. A common mistake is the use of dynamic graphics. They look spectacular, but they are actually distracting, and the download time is often prohibitive (Dellaert and Kahn, 1999). The golden rule is to keep the total "le size of a screen to less than 20 kB, or corresponding to a down load time of less than 8 s (Vora, 1999). The much publisized e-commerce has prompted companies to pursue e-business (Smith et al., 2000). In terms of economic value, e-business is about 15 times larger than e-commerce. In e-business, which is the future for supply chain management, employees working for di!erent companies exchange information in buying and selling. There is no personal economic threat, and there is no concern for credit card fraud. As more companies join e-business full force, there will also be a bene"cial impact on e-commerce. The loop closes * from customers to suppliers to manufacturing and sales. 9.2. Theories The research studies and theories quoted in this paper are mostly from management and business schools, which seem to be well ahead in research on human

618

M.G. Helander, H.M. Khalid / Applied Ergonomics 31 (2000) 609 } 619

factors of e-commerce. There are several theoretical developments. The Theory of Flow discussed earlier has the goal to design an interface which will improve user attention and make the user feel in total control. The main goal is to construct a goal-driven task that is free from distractions, such as distractions from other non-related tasks or distractions due to poor usability. There are similarities with the theory of `situation awarenessa (SA) (Endsley, 1995). A common application in SA has been design of cockpit displays. The information delivery in the cockpit must be such that the pilot may anticipate what will happen next, and react quickly. Situation awareness has been controversial, since it is di$cult to de"ne good measures. It is likely the Theory of Flow will "nd it equally di$cult to de"ne operational measures. If so, we may have to resort to well established methods in design, such as usability assessment, including task analysis and cognitive walkthrough (Mueller et al., 1997). There are other important challenges. One is the design of seductive and fun interfaces (Nielsen, 1996). The research in this area is just starting. Monk (2000) suggested that a common element associated with enjoying the use of programs is the concept of &engagement' * the feeling of being within the &world' of the system. It may seem that Theory of Flow or Situation Awareness would "t into this sphere of interest. 9.3. Unobtrusive research Customer behavior on the web o!ers interesting opportunities for unobtrusive research. Presently, eventrecording software is used in marketing. This can be setup to record user behavior as an e-commerce user enters a store. Reaction times (RT) or dwell times as well as errors may be recorded. These types of data can be extremely powerful, given that there are opportunities to record the behavior of many users. Event recording can be used unobstrusively, and very realistic reaction time (RT) experiments can be designed. The design of web pages may be altered and web users' RTs evaluated. It would also be possible to evaluate customers' search behavior and quantify the trade-o! function. The ethical considerations for such unobtrusive research will still need to be examined of course. 9.4. Do-it-yourself design Another emerging "eld of research is Web-based doit-yourself product design. The customer may design a product on the web. The product is then manufactured according to the speci"cations and shipped to the customer. Khalid (1999) proposed a hierarchical approach for uncovering customer needs in such design activity. As mass customization becomes more important in the third millennium, design of e-commerce web sites will

have to meet and match the requirements of the customer, the manufacturer and the supplier in a dynamic global customize}supply}manufacture chain. This provides many opportunities for interdisciplinary research from the related "elds of psychology, communications, management, economics, engineering, ergonomics and human factors.

10. Conclusion Many executives, professionals, and general users are already weary of the IT revolution. They question if there have been any productivity or other bene"cial e!ects of the massive onslaught of software, and they are cautious in adopting new software and systems (Landauer, 1995). E-commerce however, should be di!erent. It is not technology-driven, but user-driven. There will be many opportunities for human factors and ergonomics specialists. The chances are that our science will be in great demand.

References
Bellman, S., Lohse, G.L., Johnson, E.J., 1999. Predictors of on-line buying: "ndings from the Wharton virtual test market. Comm. ACM. 42 (12), 32}38. Benway, J.P., Lane, D.M., 1998. Banner blindness: web searchers often miss `obviousa links. Internetworking 1(3) (Internet Technical Group, Human Factors and Ergonomics Society, Santa Monica, CA), pp. 1}4. Csikzentmihalyi, M., 1975. Beyond Boredom and Anxiety. Jossey-Bass, San Fransisco, CA. Csikzentmihalyi, M., 1990. Flow: The Psychology of Optimal Experience. Harper Collins, New York. Crisp, C.B., Jarvenpaa, S.L., Todd, P.A., 1997. Individual di!erences and internet shopping attitudes and intentions. Working Paper, Graduate School of Business, University of Texas at Austin. http://ccwf.cc.utexas.edu/&crisp/Indiv}Shop.htm. Dellaert, B.G.C., Kahn, B.E., 1999. How tolerable is delay? Consumer's evaluations of internet web sites after waiting. J. Interactive Marketing 13, 41}54. Endsley, M.R., 1995. Toward a theory of situation awareness in dynamic systems. Human Factors 37, 32}42. The Economist, 1999a. The net imperative.The Economist, June 26, pp. 1}44. The Economist, 1999b. The heyday of the auction. The Economist, July 24. Fuccella, J., Pizzalota, J., 1997. Separating content from visuals in web site design. Internetworking 2(1) (Internet Technical Group, Human Factors and Ergonomics Society, Santa Monica CA, USA). Forsythe, C., Grose, E., Ratner, J., 1997. Human Factors and Web Development. Lawrence Erlbaum Associates, Mahwah, NJ. Gunnerson, G., 1999. Behind the scenes. PC Mag. 18 (22), 12. Ha K ubl, G., Trifts, V., 2000. Consumer decision making in online shopping environments: the e!ect of interactive decision aids. Marketing Sci. 19, 14}31. Helander, M.G., 1995. A Guide to the Ergonomics of Manufacturing. Taylor & Francis, London, UK. Helander, M.G., Landauer, T.K., Prabhu, P.V., 1997. Handbook of Human}Computer Interaction, 2nd Edition. Elsevier Science, Amsterdam.

M.G. Helander, H.M. Khalid / Applied Ergonomics 31 (2000) 609 } 619 Helander, M.G., Du, X., 1999. From Kano to Kahneman. A comparison of models to predict customer needs. Proceedings of the International Conference on TQM and Human Factors, CSHTO, Sweden, pp. 322}329. Hoque, A.Y., Lohse, G.L., 1999. An information search cost perspective for designing interfaces for electronic commerce. J. Marketing Res. 36, 387}394. Hurst, M., 1999. Holiday E-Commerce '99 Good Reports. Order from: update@crativegood.com. Jarvenpaa, S.L., Todd, P.A., 1997. Consumer reactions to electronic shopping on the world wide web. Int. J. Electron. Commerce 1 (2), 59}88. Johnson, E.J., Lohse, G.L., Mandel, N., 1999. Designing marketplaces of the arti"cial: four approaches to understanding consumer behavior in electronic environments. Working Paper Colombia School of Business, Colombia University, New York. Jordan, P.W., 2000. The four pleasures * a framework for pleasures in design, In: Jordan, P.W. (Ed.), Proceedings of Conference on Pleasure Based Human Factors Design. Taylor & Francis, London. Kahneman, D., 1999. Objective happiness. In: Kahneman, D., Diener, E., Schwarz, N. (Eds.), Well-Being: Foundations of Hedonic Psychology. Russell Sage Foundation Press, New York. Kahneman, D., Tversky, A., 1979. Prospect theory: an analysis of decision under risk. Econometrica 47, 263}291. Kehoe, C., Pitkow, J., Rogers, J., 1998. GVU's ninth WWW user survey report. O$ce of Technology Licensing, Georgia Tech Research Corporation, Atlanta, GA. Keeney, R.L., 1999. The value of the internet to the customer. Manage. Sci. 45, 533}542. Khalid, H.M., 1999. Uncovering customer needs for Web-based DIY product design. Proceedings of the International Conference on TQM and Human Factors, CSHTO, Sweden, pp. 343}348. Kim, J., Moon, J., 1998. Designing towards emotional usability in customer interfaces-trustworthiness of cyber-banking system interfaces. Interacting Comput. 10, 1}29. Kirakowski, J., Claridge, N., Whitehand, R., 1999. Human Centered Measures of Success in Web Site Design. Proceedings of Human Factors and the Web, National Institute ofg Science and Technology, Washington, DC. Larson, K., Czerwinski, M., 1998. Web page design: implications of memory, structure and scent for information retrieval. Proceedings of CHI, ACM, New York, pp. 25}32. Lohse, G.L., Bellman, S., Johnson, E.J., 2000. Consumer buying behavior on the Internet: "ndings from panel data. J. Interactive Marketing 14, 15}29. Lohse, G.L., Spiller, P., 1998. Quantifying the e!ect of user interface design features on cyberstore tra$c and sales. Proceedings of CHI, ACM, New York, pp. 211}218. Lynch, G., Palmiter, S., 1999. The standard web user model. Proceedings of Human Factors and the Web, National Institute of Science and Technology, Washington, DC. Lynch, J., Ariely, D., 1999. Interactive home shopping: e!ects of search cost for price and quality information on consumer price sensitivity. Working Paper, Duke University, Raleigh, NC.

619

McLatchie, H., 1999. `E-business essentials 08-23-1999, Clip } The Webzine for eBusiness Computing, http://www.clipwebzine.com/. Mandel, N., Johnson, E., 1999. Constructing preferences online: can web pages change what you want? Working Paper, University of Pennsylvania, Philadelphia, PA. Marcus, A., Armitage, J., Frank, V., 1999. Globalization of user-interfaces for the web. Proceedings of human factors and the web. National Institute of Standards and Technology, Washington, DC. Menon, S., Kahn, B.E., 1998. Cross-category e!ects of stimulation in the shopping experience: an application to internet shopping. Working Paper, Department of Marketing, The Wharton School, University of Pennsylvania. Monk, A., 2000. Computers and fun. In: Jordan, P.W. (Ed.), Proceedings of Conference on Pleasure Based Human Factors Design, Philips Design, Groningen, The Netherlands. Mueller, M.J., Haslwanter, J., Dayton, T., 1997. Participatory practices in the software cycle. In: Helander, M.G., Landauer, T.K., Prabhu, P.V. (Eds.), Handbook of Human}Computer Interaction, 2nd Edition. Elsevier Science, Amsterdam. Nel, D., van Niekerk, R., Berhon, J.-P., Davies, T., 1999. Going with the #ow: web sites and customer involvement. Internet Res: Electron. Networking Appl. Policy 9, 109}116. Nielsen, J., 1996. Seductive Interfaces, www.useit.com/papers/. Nielsen, J., 2000. Designing Web Usability. New Riders, Indianapolis. Norman, D.A., 1990. The Design of Everyday Things. Doubleday/Currency, New York. Norman, D.A., 1999. Commentary: banner blindness, human cognition and web design. Internetworking 2(1) (Newsletter of the Internet Technical Group. The Human Factors and Ergonomics Society, Santa Monica, CA), pp. 1}4. Rabin, M., O'Donoghue, T., 1999. Addiction and Self Control. In: Elster, J. (Ed.), Addiction: Entries and Exits. Russsel Sage Foundation, New York. Rogers, M., Rajkumar, T.M., 1999. Developing electronic commerce web sites for the visually impaired. Inform. Systems Manage. Winter, 15}25. Smith, M.D., Bailey, J., Brynjolfsson, E., 2000. Understanding digital markets. Working Paper, Sloan School of Management, MIT, Boston. Spiller, P., Lohse, G.L., 1998. A classi"cation of internet stores. Int. J. Electron. Commerce 2 (2) 29}56. Tiller, W.E., Green, P., 1999. How to make your web site fast and usable. Proceedings of Human Factors and the Web, National Institute of Standards and Technology, Washington, DC. Tilson, R., Dong, J., Martin, S., Kieche, E., 1998. Factors and principles a!ecting the usability of four e-commerce sites. Proceedings of Human Factors and the Web, http://www.research.microsoft.com/ users/marycz/hfweb98/tilson/index.htm. Trevino, L.K., Webster, J., 1992. Flow in computer-mediated communication. Comm. Res. 19, 539}573. Vora, P., 1999. Personal Communication, Kuala Lumpur, August 3}5. Wickens, C.D., Hollands, J.G., 2000. Engineering Psychology and Human Performance, 3rd Edition. Prentice-Hall, Englewood Cli!s, NJ. Wilson, T., 1999. Not a global village after all? InternetWeek, 12/06/99, p. 13.