1 Financing the Internet: Four Sustainable Business Models September 21, 1998 Xavier DREZE * François-Xavier HUSSHERR ** * Xavier Drèze is Assistant Professor of Marketing at the University of Southern California and visiting Professor at the Institut d Administration et de Gestion, Université Catholique de Louvain. ** François-Xavier HUSSHERR is a Ph.D. student at the Ecole Nationale Supérieure des Télécommunications de Paris and a researcher at the CNET (France Telecom).
2 2 Introduction The Internet has long been seen as a free country with no access restrictions. Roaming the land was free. The costs of owning real estate (ether estate) were insignificant compared to similar real estate in the physical world. Indeed, the Internet started as a means of sharing information between leading academic institutions. The project was funded by national and federal agencies. This panorama has now changed. The main Internet traffic is now of commercial or recreational nature. The institutions that financed the birth of the Internet have backed down and now raise the issue of who will be willing to bear its running costs. The costs of operating the Internet can be divided in two categories. (1) Hardware costs of developing and maintaining the Internet backbone, and giving access to anybody wishing to use it. (2) Content costs of publishing information on the Internet. These include both the costs of gathering and packaging the information and the cost of the hardware needed by the publishing agency. The question that we are trying to answer in this paper is: How can we successfully finance the Internet? It is our contention that, on the one hand, the hardware costs of running the Internet will be paid directly by individuals on a subscription basis in the same fashion that they are paying for the telephone system. The content, on the other hand, will be financed through one of three business models: the sales model, the media model, and the synergy model. These business models will coexist, giving rise to a selfsegmentation of both customers and providers.
3 3 Figure 1: Four Sustainable Business models COST INTERNET ACCESS REVENUE SUBSCRIPTION SALES WEB SITE COSTS MEDIA SYNERGY These four business models are only pertinent to sites that have a commercial raison d être and that can justify their existence for one economic reason or another. We do not consider the ever-growing number of web sites that are supported by private individuals simply for the sake of having a forum in which they can express their opinion to an uncaring world (e.g., Personal Web Pages). These sites probably constitute a majority of the sites in existence, but less than 1% of the actual Internet traffic. When describing each of these four business models, we divide our discussion in three sections. We start by looking at the costs brought forth by the models. We, then, consider typical revenue streams for the model. Eventually, we discuss additional issues that must be considered when implementing the model. The subscription model The Internet is not a free network. The information downloaded by Internet users travels across leased lines before being displayed on your computer screen and someone has to pay for the use of bandwidth. Like the regular phone system, end users have to
4 4 pay to access the network of networks. The subscription model has emerged as one of the most obvious one to bill Internet surfers. 1. Covering hardware costs The reason why consumers have long believed that the Internet was free comes from the history of the Internet. The Internet was indeed free for 25 years! The first network, ARPANET, was financed by the Defense Advanced Research Project Agency (DARPA). This network was so successful that, during the 70 s, the National Science Foundation (NSF) decided to support an additional small 56 Kbps network for institutions without access to the ARPANET 1. The new NSF network began to function in 1981 and was called CSNET. In 1985, the NSF decided to finance an additional network (NSFNET) based on T1 lines (1.5 Mbps). In 1991, the first two networks, ARPANET and CSNET, ceased their activity. One year later, NSF upgraded the speed of the NSFNET backbone to 45 Mbps (T3). Meanwhile the traffic grew exponentially and the commercial use of the then new World Wide Web became so important that the NSF announced on April 30 th, 1995, that it would no longer support commercial traffic. Private individuals or companies desirous to use the Internet backbone would have to pay access fees. People who were used to surf freely through their university network were now facing fees for private access. To offer Internet connection, an ISP has to buy one modem for every 8-12 new subscribers. It also has to set up different servers to support applications such as , news, web pages and domain name resolution and a LAN to connect the modem banks, 1 Access to ARPANET was limited due to its links with the Department of Defense.
5 5 POP servers and router(s). Routers are especially expensive and are price between $2,800 and $100,000 per unit. Transport costs consists in leased lines. MCI T1 lines are priced around $3,200 per month and $3.87/mile/mo. Transport costs also include the interconnection costs. Some interconnections occur at private peering points between two ISPs, and some occur at public interconnection points called NAPs (or MAEs). However, individual peering agreement must be established between any two ISP. The specific settlement arrangements between these ISPs are individually negotiated and kept confidential. In general when the two peering ISPs differ greatly in size, the smaller usually pays the larger one to interconnect. The larger one is often called backbone. Table 1: ISP costs breakdown Sales, Marketing, G&A Percentage Costs for a large ISP* 28% 17,533,000 Capital Equipment (hardware and software of the network) 11% 6,888,000 Operations (billing, maintenance, and operations) 11% 6,888,000 Customer service (support staff for customers) 26% 16,281,000 Transport (bandwidth from backbone companies) 24% 15,029,000 (source: McKnight, Lee and Brett Leida.) *ISP with end user customers paying a monthly bill of $20 Surprisingly, customer service can represent a significant portion of an ISP s costs (See Table 1). It is furnished by representatives who provide technical support over tollfree phone calls between the ISP and the customer. While telephone companies mostly
6 6 deal with billing related questions, ISPs deal with more technical questions, such as my modem won t connect or I forgot my password. These questions usually need to be handled by more qualified personnel. The more advanced the technology involved in the connection to the ISP (ISDN, T1), the more technical and expensive the support. For instance, when one of the authors decided to upgrade his Internet connection from a 56K modem to a 1.5MB Cable Modem connection, his ISP first sent a team of 4 people for 3 hours to install the modem. This was followed the next day by a 2-hour visit by a team of 2 technicians. This represents a total of 16 man-hours, a far cry from a typical telephone installation. Operation costs correspond to the routine tasks necessary to keep the ISP functioning and fall into three principal sections (Network operation and maintenance, facilities, billing and collections). The high percentage of sales and marketing costs mirrors the hard competition this market has to face. In the US, the number of ISP is estimated to exceed 3,000. From a customer stand point, most of these ISPs are highly substitutable. Customers are known to readily switch ISPs if the level of service falls below expectations, or if a competitor runs an attractive special offer. This may explain why so many Internet access providers are loosing money. The number of competitor is so high that companies need to merge to own a large enough customer base to allow them to amortize fixed costs. 2. Revenue streams The market for backbone services is very competitive both in Europe and in the United States. Very few players are able to show any profits. The market has already
7 7 experienced a first wave of consolidation such as the takeover of UUNet by Worldcom in 1996, BBN by GTE the same year, and more recently, Oleane the second French backbone by France Telecom in March This merger wave is all the more easy that the turnover of the original backbone companies (See Table 2) represents less than one percent of the turnover of telecommunication operators such as AT&T ($50 billion in 1997), MCI ($19.7 billion in 1997) or Sprint ($14.8 billion in 1997). These operators, fearing that the transmission of both voice and data over the Internet will put them out of business, are eager to buy their way into the data market. Table 2: Backbone Revenues Company 1997 Turnover UUNET $540 Sprint $295 BBN/GTE $265 Transpac $874 Oleane $8 UUNET France $3 (Source: in millions of dollars) At the retail level, the market has also experienced a series of consolidation. In the US, telecommunication operators are less present in this sector although AT&T, MCI and Sprint are among the largest players. The market is in fact completely dominated by AOL and its 11 million subscribers and by players such as MSN, Prodigy. Individual access price varies greatly. In the US, Internet users are accustomed to pay a monthly flat rate in the vicinity of $20. This yields a turnover for the industry in the vicinity of 5 billion dollars (See Table 3). In most of the European countries, surfers face a two-part tariff with a monthly subscription plus a usage charge for the time they
8 8 spent online. Australian surfers face a flat fee for access throughout the country and a steep per usage fee when trying to access sites outside Australia 2. Internet business services also vary significantly. A 64Kbit/s access usually is about $1,000 a month whereas a 1.5 Mbit/s access is close to $3,000. Table 3: ISP retail revenues Access Type 1997 Dedicated Access (T1 or better) $407 Dial-up access - Business Dial-up - Consumer Dial-up $1,200 $3,600 TOTAL $5,207 (source: Sprint, in million of dollars) 3. Improving the odds The ISP market is very competitive. Most of the players are still loosing money. Hence, they are not operating a sustainable business model. The most common cause of losses is one of economies of scale. Prodigy estimates that the break-even point is around customers. Very few Internet companies are capable of reaching this figure. The only outstanding exception seems to be AOL which has positive results due both to its unique size of 11 millions customers, which enables the provider to amortize fix costs, and to its complementary advertising revenue accounting for more than 15% of turnover. The Subscription business plan will be sustainable in the long run under any of three conditions. First, the Internet access providers can supplement their revenue stream with revenues from complementary products (Internet advertising or e-commerce revenues). The singer David Bowie for instance recently launched an ISP that offers special music 2 Australia uses satellite up-links to communicate with the rest of the world.
9 9 related products in addition to the traditional ISP offerings. Second, one can expect operation costs to decrease. Hardware costs (11% of running expense) and transport costs (24% of running expenses) are decreasing at a steady rate. Break-even points naturally follow suit. Third, ISP might be able to increase monthly access charges. As the industry consolidates, the companies still in operations gain more power over their customers and can thus charge higher prices. The sales model For many companies, the first and foremost reason to have an Internet presence is to sell. They view the Internet as a new distribution channel that has low fixed costs and a wide trading area. This new channel is sometimes compared to the catalog industry. It does, however, have unique features. 1. Setup costs In the physical world, companies need to own a piece of real estate in order to interact with their customers. This requirement entails search costs, transformation costs as well as rental or mortgage fees. The suitability of the real estate as a business location will depend on such characteristics as location, price, supporting infrastructure,.... These characteristics make locations such as Rodeo drive or 5 th Avenue desirable, but so expensive that only the more profitable establishments can justify them. By contrast, the Internet s ether estate is available to everybody at the same low cost. Anybody can setup shop on the Internet for a few hundred dollars. One can rent space on an Internet Service Provider (ISP) for $30 a month. It costs only one hundred dollars to register a domain
10 10 name for two years. Once a web site is setup, all the owners have to do is to register their site with the various search engines in order to generate traffic. Adding one's site to search engines is free. The authors know of multiple companies that have started doing business on the Internet with capitals of less than two thousand dollars. These companies have Internet store fronts that are similar to the storefronts of companies that have capitals of millions of dollars, or have hundred of employees. These low cost outfits represents however a minority among commercial web sites. According to a study from the Association of National Advertisers (1998), the average cost of developing a web site is $228,000, while the running cost of maintaining the site is $150, Revenues from E-Commerce Electronic commerce is still in its infancy. Its scope is still largely influenced by the biased make up of the Internet population. Internet users are still very much a technology oriented group. Thus, it comes as no surprise that close to 50% of the items purchased on the Internet are computer related (See Table 4). One can expect the repartition of purchases across the various categories to change over time as the demographics and interests of the Internet population change to come closer to the population as a whole. Projections for 1998, however, still put computer-related products at the top of the E-Commerce charts. Table 4: Online retailing revenues Site Estimated 1998 Sales Cisco 4,100 Dell 1,016 Amazon 160
11 11 Microsoft Expedia 150 Preview Travel 150 (Source: Iconocast, April 1998, in $ millions) 3. Issues in Internet Retailing While the ISP operators are still in the red and hope for a better future, online catalogs are for the most part profitable and hope that the future will remain bright. There are however some issues for concerns. Cluttered Environment By setting up shop on the Internet, a company gains instant access to a worldwide market. It will be reachable by millions of consumers around the world regardless of whether they are located in Manhattan (New York, USA) or in Jumet (Belgium). By the same token, consumers who login to the Internet gain instant access to worldwide shopping. They do not need a Passport or to suffer jetlag in order to shop in Singapore one minute and in Paris the next. This ubiquitous access to ether estate makes the web a cluttered environment reminiscent of an Arabian souk. There is currently over a million sites around the world. Alta Vista has a database of 125 million web pages. Some Internet shopping malls provide direct access to more than 1,500 stores. How is an individual supposed to find what he/she needs? How is one supposed to differentiate flyby-night operators from legitimate long-term outfits? Likely signals of trustworthiness will be Brand Equity and Advertising Spending. The online retailer needs to convince consumers that he is legitimate and is there to stay. The easiest way to convince consumers is by already having a relationship with them. Retailers who have been in
12 12 business in the real world for an extended period of time are at a clear advantage. Consumers will associate the online operations with the physical operations and transfer the brand equity from one to another. Another way to indicate a long-term position is to invest heavily in advertising. Since such investment can only be recouped over an extended period of time, consumers associate high advertising spending with legitimate businesses. Increase in competition Economic theory tells us that the more intense the competition, the slimmer the profits. In case of perfect competition, profits even vanish to zero. Companies generate just enough revenues to cover their expenses (raw materials, salaries, capital expenses,...) with no extra benefits. The companies whose costs are a bigger burden than their competitors eventually disappear. If we look at the Internet from this perspective we will have to note a few things. First, there is virtually no transportation cost for surfers. For the most part, the cost of time is the only transportation cost. One can zap between multiple web sites in a matter of seconds. Second, most of the products sold by Internet retailers are undifferentiated. The copy of Michael Crighton s Airframe sold by is exactly the same as the copy sold by Third, the Internet suffers from an acute free rider problem. Indeed, given the negligible search costs, consumers can easily gather information from multiple sources when deciding what to buy. Once that decision is made, they can use price as sole determinant of where to buy. There already exists
13 13 shopping agents to which on can ask to find the best price on Pearl Jam's latest CD. It will not be long until this bargain shopping ability will be extended to other goods. This threat of perfect competition might be a bonanza for consumers; it is not for retailers! A great deal of the effort expanded by marketers is actually aimed at differentiating products and avoiding pure competition. We see two scenarios as possibly evolving. First, for some product categories, retailers will be able to differentiate themselves. Second, for other product categories, differentiation will not be possible and only manufacturers will sell their wares on the Internet. How can Internet retailers differentiate themselves? Retailers differentiate themselves by adding value to the transaction they have with consumers. This added value can be provided under many forms such as lower prices, better services, longer warranties,.... There are three points in the interaction between retailers and consumers during which this added value can come into play: before, during, and after the sale is concluded. For obvious reasons linked to the free rider problem evoked earlier, before sales added value is a very limited differentiation factor. For instance, by creating a mailing list that informs potential customers of new products one can create a slight edge over a competitor. This might allow the retailer to capture some profit as long as consumers believe that the potential savings generated by a complete product search are not worth the time spent on that search. In order to generate sizeable profits however, retailers will have to come up with during and after sale added value. During sales added value can be generated by transaction enabling devices such as exclusivity agreements or by transaction facilitating devices such as wide ranges of payment methods, or large selections. A common advantage of online retailers is that
14 14 they often allow customers to save their profiles so that their static information (e.g., name, address ) does not need to be entered for each transaction. After sales added value can be generated by prompt delivery of goods (requiring large inventories, and/or a large number of warehouses) or by some device that minimizes the perceived risk in the mind of the consumer. Dell is a good example of a company that provides after sales value with prompt delivery and good technical support. It is at this level that reputation of the seller, or money back warranties will come into play. One last possible way for retailers to differentiate themselves is through loyalty programs. Following the arguments of Klemperer and Png (1986) or Drèze and Hoch (1998), such programs increase the switching cost imposed on consumers, and might very well be one of the best ways to create transaction costs (assuming one uses a convex reward structure as proposed by Klemperer (1995)). For product categories where retailers cannot differentiate themselves, an alternative viable option is for the manufacturers themselves to sell on the Internet. We will then have a classic dual channel of distribution strategy where manufacturers sell both directly on the Internet, and through resellers in the physical world. By controlling the prices on the Internet, manufacturers can insure that the prices they charge do not unduly compete with the resellers. Trustworthiness of WWW One of the greatest problems currently facing commerce on the Internet is one of trust. Trust is not really an issue for the other three business models. However, in order
15 15 for business to start using the Internet as a significant sales avenue, both businesses and consumers must have confidence in the Internet as a secure medium for transaction. Unfortunately, the current track record of the Internet is not the best in terms of perceived security. There isn t a week going by without some reports of a new security loophole found in the latest version of a significant piece of Internet software. Media are filled with articles on the weaknesses of ActiveX, bugs in the security routines of Netscape Enterprise, or new and improved algorithms that can break encryption keys in a matter of hours. People are warned not to send credit card or social security information over the Internet for fear of being ripped-off. Far from us to deny these problems. It is true that sending a credit card number over the Internet is as dangerous as shouting one s PIN number to an ATM machine. It is true also that this is only a problem if (1) somebody is listening, (2) that person has malicious intents and (3) that person has the means of its intents. Before condemning the Internet for its unsuitability for commerce, one can try to compare its level of security to the one achieved by more traditional transactions. We will preface this comparative exercise by mentioning that one of the authors has had his cellular phone cloned twice, one of his phone cards used fraudulently, and has lost two credit cards. Each of the phone incidents generated bills in the vicinity of a thousand dollars (of which he did not pay a dime). None of credit cards lost resulted in fraudulent charges. We will also mention that when he moved from one part of the country to another, this author asked a friend to drive his car (for roughly 2,000 miles) and gave him one of his credit card to pay for travel expenses. The friend had no problem paying for gas, food, and lodging with this credit card that did not belong to him.
16 16 With this preamble, let us look at the case of a consumer, Peter, buying a CD player from Best Buy. Why is this consumer not worried about the transaction? First, he can look at the merchandise before buying it. He knows exactly what the product will be! Second, Best Buy is a national retailer. Each location cost millions to open, and is there to last. There is little doubt in Peter s mind that the store is a legitimate business, and will still be there next week if need be. Even if the store were to close, the warranties, return policies or other agreements would be honored by other Best Buy locations. Third, Best Buy has a thirty days money back guarantied policy. And forth, Peter receives a receipt with Best Buy s logo, the purchase information, and the credit card authorization number on them. This receipt should be all that is needed to resolve later disputes. Let us now assume that Peter, on his way back from Best Buy, stops at his bank and deposits a check at the bank s ATM. How trustworthy is this transaction? Here again, Peter receives a receipt from the bank to which he can refer in case of disagreement. The ATM is located on the side of the bank s building. It is therefore safe to assume that the ATM is legitimate. Later that day, Peter is watching television, and sees an infomercial for Ginsu Knifes. After seeing how the knifes cut though metal cans, he decides that he should get a set, and calls the toll-free number displayed on the screen. This transaction is much less safe than the preceding two. Peter does not actually know whom he is calling, where the Ginsu Corporation (if that is really their name) is located. Peter does not either receive any proof of transaction. All he might get is a transaction number. What makes Peter rest easy is, first, that he has seen the commercial dozens of time before. The Ginsu
17 17 Corporation has been around for a while (see previous argument about signaling). Second, if the knifes do not arrive, he can always cancel his credit card payment for nondelivery of goods. And third, even if the knifes were not performing up to expectation, $49.95 is not a big loss. This makes that although most calls are made using cordless phones or cellular phones, which are the easiest thing to tap (as Newt Gingritch will attest to), billions of dollars are spend every year on product bought from Infomercials or the Home Shopping Network. For dinner, Peter and his family go to a restaurant. Peter pays with his credit card. When the waiter comes back, after five minutes, with the receipt, Peter does not consider for one minute that the waiter had ample time to go to the next door surf shop and buy a surf board using his credit card. After dinner, Peter receives a call from a telemarketer who tries to sell him supplemental life insurance. Peter, being already well-insured declines the offer. Nevertheless, every year American consumers spend billions on telemarketed Figure 2: Trustworthiness of the Internet Banks National Chains Small Independent Businesses E-Commerce Infomercials Cold Call Telemarketing products. So, where do Internet based transactions stand compared with the ones we have just described? Probably at a level that is similar to television Infomercials. Better than
18 18 cold call telemarketing, but not as high as transaction that produce a receipt. Given the ubiquitous position achieved by telephone based shopping (catalogues, telemarketing, infomercials,...), we are confident that as consumers get over their fears, as they have gotten over their fears of ATMs and check-out scanners, the Internet will grow to become a significant source of revenue (with the caveats mentioned above concerning the competitive environment). This spread of commercial transactions will no doubt be facilitated by the progress to be made in terms of encryption, digital signature, and other security enhancing techniques. E-Commerce compared to Catalogs The dilemma faced by retailers when deciding whether or not to go online is similar to the one faced by catalog retailers. The choice of going online versus bricks and mortar will depend on whether the goods sold need to be touched or tried on before purchase, or are perishable, and whether or not the potential customer can wait for the product to be delivered. On the surface, online stores operate in the same way as mailed catalogs. Consumers peruse them at their leisure, they can order at anytime of day or night, and they receive the merchandize after a week or two. A major difference with catalogs however is where the information transfer originates. With catalogs, retailers use mailing list that they build over time or purchase to seek out prospects. They then mail the catalogs to each selected prospect. The catalog acts both as a tool that consumers can use to select an item to purchase and as highly targeted advertisement that the producer places in the consumers mailboxes and coffee tables. The catalog thus is both an ad and
19 19 a means of purchasing. On the Internet, the online catalog serves only one of the two functions. It is only a purchase vehicle. The relationship is initiated by the prospect not the retailer. This limits the power of the catalog, but also greatly reduces its cost. The largest costs incurred by catalog merchants are database, printing, and mailing costs. These costs are eliminated in the online catalog model. This allows retailers to reach a much broader audience at a fraction of the cost. The media model As pointed out in the former chapter, the costs of sophisticated web sites have skyrocketed during the last two years. Search engine such as Yahoo, Excite or Lycos reflect particularly well this new trend. These companies are steadily offering new features transforming themselves in Internet service aggregators or portal services. They keep on adding new channels (weather forecast, sports, games...) and proposing new services such as free or customized web pages of news. The more Internet services you append, the more money you have to spend. 1. The costs of producing Internet services The costs linked to the media models can be divided in two broad categories: the development costs and the sales costs. High development costs are needed to ensure that the site used to sell advertising is on the leading edge of the industry and can thus attract a large segment of the population. The sales and marketing costs are even higher due the competitiveness of the industry and the relatively small size of the advertising budgets allocated to this medium.
20 20 Table 5: Search Engine Income Statements 1997 annual financials YAHOO EXCITE INFOSEEK LYCOS Revenue Cost of goods Sold Gross profit Operating expenses Sales and Marketing Products development General and administrative Others (acquisition ) Total of operating expenses Operating Revenue (28.5) (29.7) (18.6) (8.7) Net loss (22.9) (30.2) (24.6) (6.6) (source: SEC filings) Development costs Acquisition costs are remarkably heavy. As the Internet grows at a dramatically rapid pace, Internet companies have not enough time to develop internally all the services they intend to offer. So, as a result of their race against time, they try to acquire companies. Yahoo bought Four 11 and its fre solution Rocketmail for $90 million in October It was imitated by MSN which acquires a few weeks later Hotmail, the largest American fre company for $400 million. Excite took over Matchlogic to improve its customized services and Lycos merged with Tripod to append online communities services to its existing offer. Internal costs are also really high. One must especially considered the costs related to the provision of search services. Yahoo is paying several dozens of people to surf all day long and classify the 5,000 new sites that are submitted daily by companies,