Assurance Protocols and Small Web Retailers
Distributed Systems Technologies Group
High Performance Systems and Service Division
National Institute of Standards and Technology
100 Bureau Drive Stop 8951
Gaithersburg MD 20899-8951
Many Web areas are in an early technological period of rapid evolution and intense competitive selection. Nowhere is this truer than with electronic commerce. While much is being considered today for business-to-business transactions, the Web also represents a marvelous opportunity for small retail establishments. However, the needs of small establishments differ from those of larger firms. In particular, customer assurance is very important. Several assurance protocols are examined for their utility to small retail sellers on the Web. One new possibility involves using bankcard records to build assurance ratings.
Assurance; customer; e-commerce; implementation; seller; verifier; World Wide Web.
The Internet and its architecture have evolved in a rapid and sometimes chaotic manner. It has been stated, with perhaps some hyperbole, that "The Internet is the largest engineering undertaking ever, and it is evolving without a grand design blueprint.". The scale of the Internet is definitely sprawling, but whether a grand design is possible, or even desirable, is an open question. In many respects the Internet has characteristics more in common with national highway systems, aviation commerce or urban developments than with conventional, closed-participation software projects, whatever their scope. One thing that emerges with urban development is a set of architectures specific and local to their circumstances. Certainly, common design elements are often successfully shared among similar applications. Nonetheless, who has ever seen two identical airports or cities? The sources of design divergence are simply too rich, the advantages of shared, common designs not obvious. The economics of one Web application may not work for another that appears to be nearly identical.
Many Web areas are in an early developmental stage of rapid evolution and intense competitive selection very reminiscent of early computer language developments decades ago. Nowhere is this truer than in electronic commerce. The pace in e-commerce is swift. And, while most e-commerce services being considered today are for business-to-business transactions, the Web also represents a marvelous opportunity for small retail establishments. Internet connections make possible sales far removed from a shop's physical location, which can be quite rural. All that is required is adequate telephone service. However, some needs of small establishments differ from those of larger firms. Small firms are usually unknown entities to potential customers, so customer assurance is of paramount importance. Several assurance protocols will be examined for their utility to small retail sellers on the Web.
2. ASSURANCE PROTOCOLS
Purchasing via the World Wide Web can trigger considerable customer anxiety. This angst is not unfounded . An electronic virtual world, like smoke and mirrors, is easily manipulated. J. Schoenfeld, CEO of Net Effect, says only 5.75 percent percent of Web site visitors even attempt to purchase something and that 67 percent of online purchases are aborted. Her observations come from a five-month study covering two dozen network enterprises. In the Web marketplace, an unfamiliar vendor suffers significant disadvantage. Customer reluctance persists even when the network is known safe. The true issue for a would-be buyer is whether the seller is bona fide. For example, are premises of the establishment dusty and ramshackle or modern in appearance? Network customers cannot make physical inspections of a business to allay their doubts.
The issue of assurance becomes especially pronounced when a potential supplier appears highly attractive over the network but remains a completely unknown entity. Smaller businesses often fall into this category, which is unfortunate. Small businesses can be especially vigorous and competitive. They are numerous--in the United States, firms with fewer than 500 employees employ about 53 percent of the commercial work force. These businesses contribute 47 percent of all US sales and 51 percent of the US GDP (figures from U.S. Departments of Commerce and Labor via www.nsbu.org). If customers can be made more confident, Web commerce represents an ideal place for small firms to thrive.
|ASSURANCE PROTOCOL||SUBJECTIVE SCORE||REL. COST/TRANSACTION (EST.)|
|Simple Acknowledgment||Confirmed (+1)||1|
|BizRate/query directly after sale||Confident (+3)||10|
|WebTrust™||More Confident (+4)||10 (heavy use) to 100+ (light use)|
Table 1. Marks and Relative Cost for Four Assurance Protocols
|1||Customer C to Seller S||Inquiry via Web (includes payment)|
|2||Seller S to customer C||Sales completion—merges into step 4|
|3||Seller S to verifier V||Assurance provider notified|
|4||Verifier V to customer C||Inquiry on sales transaction|
|5||Seller S to customer C
|Inform on details of shipment|
|6||Verifier V to customer C
|Query on quality of experience and purchase(s)|
Table 2. Typical Use of BizRate.comâ
2.1 Assuring the Uneasy
An assurance protocol is designed to dispel wariness toward a new party. Social instruments such as letters of introduction play this role in everyday life. A Web-based assurance protocol should function analogously to build trust. Discussion begins with several assurance protocols, including BizRate.comâ and WebTrust ™ [4,12]. Emphasis is upon levels of assurance versus cost per transaction: A large business-to-business transaction can tolerate more cost overhead than can a Web retail sale for $19.
2.2 No Confirmation
Suppose customer C wishes to interact with a reliable seller supplier S somewhere in the virtual world of the Web. C is unsure whether S is such a supplier. Exchanges can be made among C, S and a third party V who may, in varying degrees, vouch for the authenticity or reliability of S. The most unsatisfying of C’s Web-purchase transactions occurs when the customer completes a Web purchase and there is no further response from S until the merchandise appears (or worse, does not appear). Ill ease descends upon C and lingers until S delivers the product. Even then, C has residual smarting about the purchase. Such behavior by S is gauche but is easily found on the Web. Table 1 depicts this circumstance in its first row, where a subjective grade or mark of –1 has been assigned to "uneasy." The cost to seller S (third column, row 1) is zero, since no assurance has been given to C. In actuality, there is another cost for this negligence¾ customers who find a more communicative site will readily abandon S.
2.3 Simple Acknowledgment
A more polite seller acknowledges an order with e-mail shortly after the automated Web session. This small but effective gesture indicates something is actually happening- it is as assuring as a failure to confirm is discouraging (the respective marks being +1 and -1 by the grading scheme¾ cf. Table 1). Confirmation cost is higher than doing nothing, but it is still slight.
Table 2 gives a scenario with the BizRate assurance protocol . This scenario begins with the customer C being unfamiliar with BizRate.com® URL. (The other case follows in section 2.6.) In this case the purchase is initiated without prior assurance (Table 2, rows 1 and 2). However, in closing out the transaction, the assurance provider (or a proxy) questions the customer (step 4) and mentions a later, follow-up session. This raises customer confidence considerably more than a simple confirmation (+3 in Table 1). The next working day or so, the seller may further contact C (step 5, Table 2) with details such as shipping schedules, carrier and invoice number. This again improves confidence. Later still (step 6) the verifier V queries on the overall quality and satisfaction of the transaction. While not pertinent to the sale (goods have arrived by now), this action again increases the customer’s confidence in S. Repeat sales are more likely. Customer responses go automatically into V’s vendor database. Digests of responses are also incorporated into a Web public rating list maintained by BizRate.
|1||Verifier V to Seller S (earlier)||Certificate to seller; ~ 4 month lifetime|
|2||Customer C to Seller S||Initial inquiry via Web|
|3||Seller to customer||Web response (incl. certificate)|
|4||Customer to seller||Purchase details (incl. payment)|
|5||Seller to customer||Automatic confirmation|
Table 3. WebTrust ™
|1||Customer C to Verifier V||Inquiry on seller’s standing; open rankings; free.|
|2||Step 1, Table 2 above||Balance of transactions as per Table 2|
Table 4. Prior-use of BizRate.comâ
|1||Customer C to Verifier
Alt: C to S directly
|Background check on seller; use closed proprietary files on bank card payouts; has general metrics (below)|
|2||Verifier V to Customer
Alt: S to C
|Rating of seller S based
upon credit-card volume, logged complaints/volume, other business
Alt: send certificate on most recent BCI rating
|3||Customer C to Seller S||Order via Web; includes payment|
|4||Seller S to Customer C||Sales confirmation|
Table 5. Using Bank Card Information (BCI)
2.5 WebTrust ™
Table 3 depicts another available protocol, WebTrust™. Proposed by the American Institute of Certified Public Accountants (CICP in Canada), this mechanism is structured very similar to common accounting practices. An accountant (part of the verifier V here) audits supplier S and finding satisfactory conformance to principles of good business practice, issues an electronic certificate valid for three or four months (step 1). Customer C subsequently sees this certificate during an inquiry (step 2) and gets further details (step 3). The WebTrust™ seal of assurance is managed by VeriSign, Inc. Assured by reading the certificate, C makes an order (step 4) and receives confirmation (step 5). The process seems fairly effective, although CPAs have voiced some reluctance about management (and knock-off copying) of the approval seal. They also worry about the legal liability attending their endorsement.
Quarterly AICPA audits required for retaining the seal add considerable expense. A large volume seller will incur a far lower cost per sale for assurance overhead than will smaller retail merchants. With heavy use, WebTrust™ might be similar in cost to BizRate (both are rated 10 in Table 1, third column). Light use of WebTrust™ will kite the per-sale overhead very high (100+). The good thing is that assurance comes before conclusion of the sale session—this might help stem the loss of those two-thirds of Web sales that are aborted. A real weakness is the manual audit step and its concomitant high cost.
2.6 BizRate Revisited
An alternate use of BizRate has customers going first to its public rating Web pages. Table 4 shows the beginning of this. An early check with V gives customer C an assurance similar to WebTrust™. The cost of this assurance should be relatively low, since its collection methods can be highly automated. Another advantage is its incremental nature. It avoids a massive quarterly audit. However, with few sales and likelihood of many customers skipping the BizRate rating form, the smaller seller may suffer a significant disadvantage. At present (September 1999), BizRate has only 53 sellers in its music category for the whole USA, and not all the rankings are based upon customer responses (the others are "staff reviewed"). Imagine if the list got to be 100 to 200 times that long with some entries having only a few tally points. Clearly, scalability of the mechanism can be a challenge when the number of sellers becomes quite large, yet a good fraction of these merchants do only a tiny amount of Web business.
2.7 Umbrella Organizations
E-commerce organization may diverge from conventional retailing. Nothing illustrates this better than Vstore.com, an umbrella store supporting merchants—manufacturers, distributors, retailers- of all sizes . Each participant must (1) sell products that ship via UPS or FedExâ and (2) provide electronic-format descriptions of their product line. Vstore.com hosts all transactions and customer services—it runs a business’s Web pages, collects payments, orders shipments. Participating stores and Vstore.com receive commissions on sales. Customer assurance comes from dealing through Vstore.com, a "digital intermediary" .
Assurance via an umbrella company is attractive in many ways. For one thing, the host company is a known entity. Lack of seller name recognition poses much less a problem, since the participant seller S works under the mantle of the virtual store's implied and express guaranties. For instance, buyers at Vstore.com are given a Secure Shopper Guarantee--any unauthorized bankcard charges are covered up to the buyer’s limit of liability, which is $50 in the USA. Verisign provides Vstore.com security via 128-bit SSL (Secure Sockets Layer) encryption. In a sense, the verifier V and the seller S are combined in Vstore; seller S is actually a store affiliate paid commissions on sales by the umbrella organization. Seller commissions range from 5 to 25 percent, depending upon the merchandize.
3. A NEW APPROACH, BCI
Lessons from the foregoing examples can be summarized as:
Item (1) follows from the discussion on behavior of would-be Web shoppers. Items (2) and (4) define a mechanism cheap to run. Item (4) also mentions a source of statistical unreliability—the use of voluntary responses. Merchants with few customer sales will have even fewer tallied evaluations. The statistics for assurance may be shaky. Item (3) specifies good scalability. As discussed earlier, a tabulation of customer comments is fine for large retailers with name recognition and a large volume of sales. It is unclear how well this works when the merchant list comprises several thousand names, many of which are unfamiliar. (BBBOnline® uses a database for its listings; see http://www.bbbonline.com.)
Table 5 sketches a new assurance approach, BCI (BankCard Information), that addresses items (1) through (4) above for small, competitive but unknown vendors. Customer C first invokes an automated background checker from a browser. This background is based upon information that is implicit in any bankcard-based sales seller S has made in the past. The check is performed by a credit card organization V upon seller S’s records without S knowing who has made the request. The service V reports back upon the seller’s viability and standing among the credit card user community. Indicators might include the seller’s level of activity (transactions/day), complaints/sale and other similar metrics. There are no forms for customers and participation by sellers is voluntary.
The approach inverts the usual credit association, where business S checks upon buyer C. Here, customer C checks (electronically) upon the seller’s business standing via credit-card payment records. Since these might come from many different banks, the payments could be logged locally and aggregated by the verifier V on a periodic basis. Daily updates of the ratings seem unnecessary—a weekly update cuts the overall effort and allows the load to be staggered. Special updates can be made when a rating changes drastically. Many confidential details can be masked while still providing the necessary indications for assurance. Bypassing a manual audit and assessment, the approach should be less costly than WebTrust™. It would automatically qualify every Web-based merchant, since bankcards are one of their principal methods of payment. Merchants will also appreciate that their local established bankcard trade contributes to their Web assurance standing. That is, a rural seller with a sterling local reputation will have this reflected immediately in his Web assurance rating, even before he has sold anything through e-commerce channels. This "auto-start" reputation is attractive.
Customers would not necessarily have to make their assurance request to verifier V. The rating could be reissued weekly or monthly, much like WebTrust™'s seal of assurance in Table 3, step 1 (using the same security mechanisms) and posted at the seller's Web site. This would disperse assurance inquiries away from the verifier, V and thereby promote better scalability. Another simplification would be for the verifier V to issue assurance seals in only a few types, like investment ratings. An excellent merchant would be rated A or AA, a lesser one B, and so on. This would allay sellers' questions about outsiders getting digests of their firms' business. The seller would pay the verifier for the rating service. Provided the banks and card issuers handle the needed archival files efficiently, cost should be less than WebTrust™. BCIwould be highly automated.
4. A RELATED INITIATIVE
BCI might qualify as a future topic for the Financial Agent Secure Transaction (FAST) project . This recently initiated effort addresses how unknown entities such as small shops or Web customers might authenticate and validate crucial information in transactions over the Internet. Each FAST entity has links to its own financial institution--a bank would be typical in the USA. FAST assumes that pairs of transacting entities have no authentication means common between them. They will rely instead upon their financial institutions, as proxies, to provide secure, anonymous transactions and to guarantee payment and shipment obligations. Entities first negotiate with each other over the Internet. Each then authorizes its respective financial institution to exchange, authenticate or assure the more sensitive elements of the transaction with the counterpart institution. Particular emphasis will be given to building FAST from existing, economically efficient technology that can support needed levels of anonymity, privacy and payment method options.
Illustrating where the future might lead, a set of possible FAST transaction topics is considerably broader than the simple ratings proposed for BCI. For example, FAST might authenticate citizenship, marital status, education, membership and licensing standings; it might provide assurance on processing performance and merchandise returns. On an even broader front, the National Center for Manufacturing Sciences (NCMS) has expressed an interest in linking FAST-like authentication and assurance with Internet-based manufacturing functions . Items of concern to NCMS include an environment that securely transacts international finances or that ties payments to manufacturing achievements. Continuing beyond this, a NIST colleague suggests checks on legal matters would serve well in some lines of business—this includes indictments, lawsuits brought by and against the entity in question, criminal records, civil judgments and related settlements. This partial enumeration of desired yet sensitive information—be it associated with BCI, FAST, NCMS or other parties—signals a need to mention privacy .
5. QUESTIONS OF PRIVACY
Implicitly or explicitly, privacy always remains an element in the design of commercial systems. While it is beyond the scope here to explore privacy with any thoroughness, a few remarks will highlight typical problems. Privacy will be discussed along dimensions that are technical (security), policy (trust) or economic (cost and profit) in flavor . On the technical side, there is always a tradeoff between security and cost. Even excellent security protocols will have to survive a business case analysis . PKI (public key infrastructure) is probably too complex for most small Web retailers. The SET (secure electronic transaction) protocol is also somewhat expensive: Smaller US merchants evidently find it cheaper to absorb losses incurred with lesser levels of protection than to use SET.
5.1 Trust and Profit
Banks in the USA traditionally have fulfilled many community trust responsibilities. The quaint names for some establishments, such as "Farmers and Mechanics Bank and Trust," attest to this fact. Today, however, banking is experiencing a period of rapid transition. Many banks have become large aggregates of regional, national or international sweep. Recent US federal legislation allows banks to sell insurance and stocks. Even before this development, privacy lapses within banks served to illustrate pressures on modern businesses . This pressure, along with the enlarged scope of banking enterprises, will generate further temptations to blur trust functions with other profit opportunities . For this reason, institutional policy commitments on privacy are as important as any technical approaches.
Assurance methods are perhaps a metaphor for network commerce today. Early assurance efforts have attempted to merge older practices (acknowledgments, customer surveys, and audits) into Web technology. Several of these assurance protocols have been evaluated for their cost and effectiveness for small retail merchants. The evaluation shows that one might want a protocol with (1) assurance given early, (2) a fully automated data collection scheme, (3) system scalability, and (4) no crucial human intervention. . Requirements (1) through (4) suggest a new approach, BCI, an assurance method a new implementation based upon bankcard information, . BCI may offer attractions for small firms. . However, e-commerce is new and unpredictable. An emerging architecture like the umbrella store or FAST could change the scene completely. Searches for efficient e-commerce architectures will continue for some time as the various technical elements adjust to market forces.
7. ACKNOWLEDGMENTS, REMARKS
Thanks go to Alan Mink of the DST Group for comments and suggestions.
Contributions of the National Institute of Standards and Technology are not subject to U.S. copyright. Mention of commercial products or services is only for illustrative purposes and should not be taken as an endorsement, express or implied.
 Ba, S., Whinston, A. B. and Zhang, H. Small business in the digital economy: Digital company of the future. Proceedings of the Conference on Understanding the Digital Economy: Data, Tools and Research (Washington, D.C. 1999), MIT Press [to appear]. http://www.digitaleconomy.gov
 Barncik, S. OCC’s Chief Counsel warns banks to shield data privacy. American Banker (October 7, 1999) 2.
 Benda, M. Internet architecture: Its evolution from an industry perspective. IEEE Internet Computing 2, 2(March-April 1998), 32-35
 BizRate.comâ reference text. http://www.BizRate.com.
 Dugas, C. Banks sell your secrets: Account numbers, names, addresses are all fair game. USA Today (October 21, 1999), 1B.
 FAST project of the Financial Services Technology Consortium. http://www.fstc.org
 Ghosh, A.K. E-Commerce Security. John Wiley & Sons, Inc., New York NY, 1998.
 Gray, G. L., and Debreceny, R. S. Electronic commerce assurance services and accounting information systems: A review of research opportunities. Advances in Accounting Information Systems 6, (1998), 185-206.
 Penenberg, A. The end of privacy. Forbes 164,13( November 29, 1999), 182-189. BizRate.comâ reference text. http://www.BizRate.com.
 Talley, M. Memorandum to FAST Project on initiatives of the National Center for Manufacturing Sciences, October 21, 1999.
 Vstore.com organization. http://www.vstore.com
 WebTrust™ description text. http://www.aicpa.org/webtrust/into.htm.
Biographical Sketch: Gordon LYON manages the Distributed Systems Technologies Group at the National Institute of Standards and Technology, Gaithersburg, MD. He joined NIST in 1972 upon completing a Ph.D. in computer science from The University of Michigan. Prior to NIST, Lyon worked in advanced applications prototyping at the General Motors Research Laboratories and in human performance at The University of Michigan Medical School. Primarily interested in non-numerical programming techniques, Lyon has contributions in the areas of syntactic pattern recognition, scatter storage, programming languages, programming environments, and performance measurement tools. He is a member of ACM, IEEE Computer Society, and SIAM.
SAC 2000 March 19-21 Como, Italy