Apple boasts that more than one billion songs have been purchased from its iTunes music service. That sounds like a great number—until you consider that an estimated ten million users of Internet-based peer-to-peer (p2p) networks are logged on at any one time to swap music.
How does Apple, which sells music titles for 99 cents each, compete with free music downloads on peer-to-peer networks? Do the two approaches to distributing digital content complement each other? What can the music industry, which aggressively fights p2p downloaders, learn from Apple's experience?
Those kinds of questions attracted the research attention of Ramon Casadesus-Masanell, a professor at Harvard Business School, and Andres Hervas-Drane, a PhD candidate in Economics at the Universitat Autónoma de Barcelona and a Visiting Fellow at Harvard University. Their working paper, "Peer-to-Peer File Sharing and the Market for Digital Information Goods," is among the first efforts to study the interactions of two entirely new and radical business models operating in the same market.
Sean Silverthorne: What attracted you to research this area and what were some of the major insights you discovered?
Ramon Casadesus-Masanell and Andres Hervas-Drane: The nature of competition is changing rapidly. Drivers such as globalization, deregulation, and technological change are opening opportunities for the development of new, original business models. Competition is progressively moving away from imitation and the development of incremental tradeoffs towards radical moves to create new business models, new forms of satisfying needs that drastically reduce costs and/or raise value perceived by customers.
One important enabler of new business models is the Internet. This is especially true in industries where the product can be delivered directly online such as software, travel agencies, or media. Indeed, the digitalization of content paired with widespread adoption of broadband Internet is driving a major shift toward digital distribution. ITunes, set up as a traditional client-server architecture with content offered at positive prices, and peer-to-peer file sharing networks, which do not seek profit maximization, use p2p network architectures, and offer content for free, constitute two new business models for the distribution of digital goods.
While the study of business models in isolation is of great interest, we believe that the analysis of interactions between business models can highlight important hidden insights. In this paper, for example, we show that contrary to the popular view, p2p and iTunes can potentially develop into a mutually beneficial symbiotic relationship. The presence of iTunes has a negative impact on the size of p2p networks resulting in reduced congestion and more efficient file sharing. Better functioning p2p networks, in turn, result in more content exchange, affecting positively iPod sales and Apple's bottom line.
Q: Given that p2p delivers arguably the same music download as Apple's iTunes, and at no charge, why have millions of users chosen to pay a fee using Apple's service rather than download for free from p2p networks?
A: The choice between p2p and iTunes is not trivial. Both models differ on multiple dimensions beyond price, and neither is superior in all attributes. In a world with variance in individuals' needs and valuations, these dimensions are evaluated idiosyncratically, allowing both models to coexist. Legal considerations play an important role as a number of p2p users have been sued by record companies, but other aspects, such as the availability of content, are also relevant. Music from Led Zeppelin, The Beatles, or Radiohead, for example, has been available on p2p networks since Napster's time but is yet to make it to iTunes.
Record companies seem to be applying traditional "brick-and-mortar thinking" in their competition against p2p.
Other differences are related to the "packaging" of content. Digital rights management (DRM) technologies, for example, are used to limit the playback of music purchased on iTunes, while music downloaded from p2p networks has no such restrictions. Although record labels are increasingly experimenting with DRM-less music sales, p2p is superior in this respect. In contrast, metadata (data about data—the indexing data contained in media files such as artist or album name) is superior on iTunes. This allows music collections to be consistently organized by author, album, or genre, and provides a better navigation experience. Digital encoding quality varies widely in p2p networks, but it has continued to improve over the years and in many cases surpasses that of iTunes.
The process of obtaining content, an important part of the experience, also differs. ITunes provides a unified interface that seamlessly integrates the location, purchase, and consumption of content. Users of p2p networks, on the other hand, must navigate a complex environment and endure varying levels of congestion that hinder the quality of the process. ITunes certainly has the upper hand in this area.
The following table compares the strengths and weaknesses of each model:
|It's free||Under constant attack by industry players|
|Variety of content||Downloading time varies|
|No restrictions on content (no DRM)||Congestion problems|
|Constantly improving due to technological race||Sometimes content is unreliable (spoof files)|
|Decentralized—makes it hard||No anonymity|
|Legal||Customers must pay for content (relatively expensive at $0.99)|
|Easy to use||Restrictions on content (DRM)|
|High reliability||Less variety of content|
|Metadata (i.e., information about the files)|
|Works well with iPod|
We built an economic model to improve our understanding of how both forms of digital distribution interact. While this methodology restricts somewhat the scope of the analysis, it allows careful examination of the phenomenon. Specifically, we restricted the comparative advantages of each as follows. Content from iTunes is legally sold at positive prices and downloads are immediate; downloads on p2p are illegal and can take many hours (or even days) to complete, but they are free. Other features such as DRM restrictions and differential metadata add little strategic insight.
Q: One motivation for using iTunes that you identify is the fast download time provided by Apple—you can enjoy the music you download almost instantly. But as broadband speed rolls out to the masses, might download speed become less a competitive advantage for iTunes in attracting new customers?
A: Absolutely. One of the most important comparative advantages of iTunes is that it offers fast downloads because the client-server model allows the operator to manage congestion by simply adding more servers. Congestion in p2p networks, however, depends on the number of peers in the network, the amount of people that share, the state of broadband infrastructure, and the "resolution" of content.
Intuitively, a better broadband infrastructure should make p2p more attractive because, keeping constant the number of sharers, it enables downloads to be faster. While this "direct effect" seems obvious, our analysis also takes into account the following "indirect effect": The behavior of peers may vary when infrastructure improves. As infrastructure improves and downloads become faster, we could have a situation where a smaller number of peers elect to share, and this would end up negatively affecting download speed. We show that this is not the case. It turns out that the behavior of peers is independent of the state of broadband technology. Therefore, only the direct effect is at play. Improvements in digital encoding have a similar effect. Improvements in content resolution (such as the transition to multi-channel audio and high definition video), however, tend to worsen the efficiency of the p2p network, making iTunes relatively more attractive.
Q: Your paper suggests the complex motivations of people using digital networks. For example, p2p users face an initial decision to either "share" the content they download with others on the network, which also reduces overall network congestion, or to "freeride." Freeriding is the easier choice, so why do many users choose to share?
A: To better understand user motivations, we first studied p2p file sharing networks in isolation and later considered interactions with iTunes. It did not take us long to discover that there was a puzzle that we had to solve before introducing competition. Users of p2p networks choose whether to share (offer content that other peers can download) or to freeride (download from others and not offer content for others to download). Sharing content is costlier than freeriding as it entails committing computing resources such as storage space and upload bandwidth to the network. In addition, sharers are more likely to be prosecuted than freeriders. The implication is that nobody "should" share. But if this was the case, then p2p networks should collapse as there would be no content available for download. However, we see that p2p networks are thriving: more than 65 percent of all Internet traffic is content exchanged on p2p networks!
The impact of p2p networks needs be carefully considered when pricing legal downloads.
The easiest and most direct way to solve the puzzle is by assuming that peers are not purely self-interested but that they "feel good" when sharing. While we believe that the "feel good" explanation might be part of the reason why p2p networks exist, we find it hard to justify that peers will develop altruistic emotions in a setting with anonymous transactions. We decided to take a different route and ask whether selfish individuals would ever want to share content in p2p networks when sharing is more costly than freeriding.
The key to solving the puzzle in a world where individuals are selfish is the following. In addition to offering content, sharers also supply "bandwidth." To see this, notice that in a network where one single individual shares content, congestion (or download time) is huge as all members will connect to this one sharer to obtain content. If a second individual decides to share also, then the connections will to some extent be distributed between the two sharers and downloads will be faster. As more and more individuals share, congestion decreases because more bandwidth is supplied. As a consequence, quality of service in the p2p network improves. Because these improvements are enjoyed by all peers, including those who have elected to share, a number of peers are better off sharing rather than freeriding. Therefore, our analysis of p2p shows that sharing prevails in a world where individuals are all selfish, consistent with the evidence.
We should point out that p2p users are well aware of their sharing decisions. It has sometimes been argued that p2p applications enable sharing by default, but this can hardly explain why freeriding is so pervasive. Similarly, some p2p applications such as BitTorrent force users to share, but users can always decide how much to share and ultimately control their contribution to the network.
Q: Your research indicates that p2p networks, while competing with iTunes as music distribution channels, may also benefit from having iTunes in the market. Why?
A: The model predicts that congestion in p2p networks worsens with network size. That is, the larger the number of users in the network, the lower the proportion of users that share. This fact plays an important role in our analysis. If one considers the original Napster network and its evolution over time, congestion worsened as the user base grew. At its peak, Napster's waiting queues were long and downloads slow (when they could be completed). Modern file sharing networks are better designed to cope with scalability issues. Incentive schemes that reward users as a function of their contributions help improve sharing, but they introduce other distortions, and the legal risks of sharing limit their efficacy. The underlying problem remains: Only when upload bandwidth becomes unlimited and legal risks disappear will congestion fade away.
The presence of iTunes provides an alternative to users suffering most from congestion in p2p networks. Some users are better off purchasing from iTunes and enjoying fast downloads. The availability of this outside option drives users away and indirectly improves congestion as networks become smaller. Interestingly, iTunes was not available in the days of Napster and congestion was much worse then. It is in this sense that p2p benefits from the presence of iTunes.
Q: Do you think Apple's pricing strategy of roughly 99 cents per song effectively maximizes profits? Could Apple charge less and capture more market share, or price higher and create more net profit?
A: Apple makes little money from the sale of songs. It is estimated that record companies pocket about $.65 per song sold on iTunes and Apple keeps $.34 to cover the costs of running the service, infrastructure, encoding, dealing with credit card companies, et cetera. Apple's profit comes from the sale of iPods and related products. It is no secret that a large percentage of music files on iPods have not been purchased on iTunes. Most come from users' CD collections, other online stores (such as allofmp3.com), p2p file sharing networks, and other forms of piracy (like sharing between friends). A thriving p2p community acts as an engine for iPod sales.
ITunes's 99 cents per song is ultimately a compromise between Apple and the owners of content, in this case the record companies. We believe that from the point of view of Apple, 99 cents is too expensive. Because profit comes mainly from the sale of hardware, Apple is likely to prefer lower download prices. From the point of view of intellectual property owners, 99 cents is probably too low. Record companies have attempted to renegotiate with Apple to set higher prices for new, more popular content. Our analysis suggests that this may be a bad idea because it is precisely for popular content that p2p is a better substitute for iTunes. Rare content, on the other hand, is where p2p does not seem to work well as there are fewer peers offering it. With this initiative, record companies seem to be applying traditional "brick-and-mortar thinking" in their competition against p2p. But this is surely the wrong mindset to deal with p2p.
At the end of the day, optimal pricing is an empirical question that cannot be answered without access to proprietary data that is outside our reach. Our model, however, highlights a few factors that should be taken into consideration in the determination of optimal price. Most importantly, our research demonstrates that record companies should explicitly consider the competition from p2p networks when making pricing decisions; this is something that they do not appear to be doing presently. We show that for a large portion of potential customers the baseline comparison is not "buying at price $x vs. not buying" but "buying at price $x vs. downloading from p2p."
Of course, pirated content has been available prior to p2p networks. The cassette recorder, for example, allowed individuals to generate unauthorized copies and to illegally share copyrighted content, lowering potential revenues to content providers. But cassettes and other forms of analog content replication were subject to quality degradation and required physical exchange, which confined sharing to relatively small social networks (family and friends). By eliminating these restrictions, peer-to-peer file sharing technology has increased the accessibility and attractiveness of unauthorized content replication. The threat of p2p is not different in nature, but of much larger scale as it does not require the exchange of a physical support. As a result, the impact of p2p networks needs be carefully considered when pricing legal downloads.
Our analysis reveals that, contrary to intuition, prices low enough to "kill" p2p are not optimal in large markets. The industry is better off setting higher prices and attracting those consumers ready to pay due to congestion. Coexistence with p2p, however, does result in lower prices than would otherwise be observed. We also find that legal attacks result in less sharing, harming p2p networks and helping sustain high prices. In any case, we do not expect p2p file sharing networks to disappear anytime soon. So far, they have proven resilient to technical and legal attacks due to their decentralized nature. The content industry must find ways to embrace the new technology and profit from it.
Q: What are the practical "takeaways" for managers in related areas?
A: At a broad level, in designing new business models, managers must carefully consider how robust a given design is to models of other industry participants with which they interact. Business models built with consideration only of how they work in isolation of those of other players will often exhibit poor performance. How well iTunes works as a channel for digital content distribution depends not only on the intrinsic operation of the model but also on how it interacts with p2p. Clearly, the extent to which two business models interact is not exogenously given but a result of choices made by the designers. Conversely, managers must also ponder how aggressive their business models are toward those of other players and ask whether or not complementarities are exploited. While this point might seem obvious, the academic and practitioner communities have so far offered little insight on how to think about interaction between business models. Our paper makes a first step towards a general theory.
At a more concrete level, given that p2p file sharing networks are likely to improve in performance as Internet infrastructure develops, the content industry must make tough choices regarding their revenue models. Moves towards monetizing products not subject to costless replication and distribution, such as live concerts and merchandising (for music) and product placements (for movies and network shows), will become essential for the financial health of media companies. This will involve setting up "creative" contracts with artists to share value created.
Our analysis also suggests that the "scarce" resource in digital goods distribution through p2p networks is not content, but bandwidth. This fact has been repeatedly documented by empirical studies on p2p networks. Studies on the Gnutella network, for example, have found that over half of its users do not contribute. As a consequence, we expect ISPs to have a more visible role in shaping industry structure going forward.
Finally, to compete effectively against p2p, online digital distribution must strive to become accessible and attractive to consumers. Online content providers are in a unique position to optimize and deliver new experiences to consumers that cannot be easily matched by decentralized, self-sustained peer-to-peer networks. ITunes provides a better customer experience than file sharing networks for similar content, and this allows record companies to charge positive prices and make a profit. While the potential industry-wide revenue implications of p2p are still uncertain, our analysis suggests that there is scope for profit-maximizing online distributors and content producers to compete effectively against unauthorized file sharing.
Q: What are you working on now?
A: Peer-to-peer file sharing remains a fascinating field for research that has not received much attention outside computer science. We believe that there are great opportunities for research here and we are doing some more work in this area. In particular, we are studying how different incentives schemes built into p2p networks such as BitTorrent and eMule affect the types of content that are likely to be found in these networks and the average life of that content. This will allows us to finesse our analysis of optimal pricing of profit maximizing firms that offer similar content on a client-server architecture.
Andres is working on a model of consumer search with recommendations, motivated by the proliferation of online recommender systems. The model can explain the impact of this technology on markets for experience goods, such as music, cinema, or books. This research relates to the recent debate on the Long Tail, and discusses the strategic implications for online retailers.
Together with Barry Nalebuff (Yale School of Management) and David Yoffie (HBS), Ramon is working on a model of competitive interaction between Microsoft, Intel, and AMD. The paper, titled "Competing Complements," introduces a mechanism by which firms may discipline the behavior of complementors at the value capture stage.