Research on the information economy has been based on the assumption that production of software involves low, or even zero, marginal costs. This paper examines this assumption. It argues that the act of driving Internet traffic to an Internet server is an act of distribution and that costs associated with it are actually software production and distribution costs. It suggests a generalized model, the Internet distribution chain, using which the variable and marginal cost of production and distribution can be revealed. Using this model the paper will show that marginal costs associated with the production and distribution of software actually resemble those of traditional products.Contents
Introduction
Distribution of Software as Production of Software
Driving Internet Traffic as Distribution of Software
The Internet Distribution Chain
Internet Distribution Chain Metrics
Marginal Cost of Distribution
Variations in the Internet Distribution Chain
Increasing Marginal Cost of Distribution and Profit Seeking Behavior
Summary
Introduction
The ever growing popularity of personal computing and the Internet in recent years has drawn interest to the economic principles that guide the production and distribution of Information Goods in general and software and other Digital Goods in particular. Observing the economic forces that govern markets for such products, researchers have concluded that the products themselves possess unique characteristics that distinguish them from traditional products.
Most researchers have based their economic analysis on the assumption that "Information is costly to produce but cheap to reproduce" [ 1]. It was noted that many high-tech products, including software, "typically have R&D costs that are large relative to their unit production costs" (Arthur, 1996) and that digital goods "typically have the property that it is very costly to produce the first copy and very cheap to produce subsequent copies" (Varian, 1995). It has been argued that "with information goods the pricing-by-replication scheme breaks down." And that "this has been a major problem for the software industry: once the sunk costs of software development are invested, replication costs essentially zero" (MacKie-Mason and Varian, 1995).
These underlying assumptions are evident in further research on the information economy. Based on them, Varian examined the role Differential Pricing (Varian, 1996) and Versioning (Varian, 1997) play in the profit-seeking behavior of Information Goods producers, Bakos and Brynjolfsson (1997) and MacKie-Mason, Riveros and Gazzale (1999) have explored the effects bundling has on pricing Information Goods and Arthur (1996) has been using them to clarify the concept of Increasing Returns. The same assumption has also been widely used in the business journalistic community to explain the competitive landscape of the software industry (for example, Aley, 1996; and, Romano, 1998).
In this paper I suggest an alternative perspective towards these assumptions. I argue that as far as software products are concerned, production and distribution are parts of the same process and that distribution over the Internet consists mainly of driving consumer traffic to an Internet server. I present a generalized distribution model for software in which analysis of software production and distribution costs can be done. I show that when distribution costs are considered to be variable production costs, software products distributed over the Internet are in fact relatively costly to produce.
Based on this analysis I also show that the economics of software distribution over the Internet is not fundamentally different from those of traditional products where production is constrained by Increasing Marginal Costs and where the price is equal to the Marginal Cost of Production.
Distribution of Software as Production of Software
The production process of traditional goods and services requires a significant input of labor, raw materials or both in order to produce a product. As the costs associated with the Factors of Production are incurred for each and every unit of output, it is easy to classify them as Variable Costs. The production of software on the other hand does not display such a connection between inputs and outputs. Software products require significant input of labor in order to produce the first copy but the inputs in labor or raw materials needed in order to produce additional copies are small or even non-existing. For example if a software product has been uploaded to an Internet server, any copy download by a consumer to her workstation generates an output with what seems to be no input. As there seems to be no input, there also seems to be no input costs associated with this software product.
While this approach toward inputs might be appropriate for analyzing specific scenarios of Information Goods production, it ignores the fact that when looking at software distributed over the Internet, the production and the distribution of such software are parts of a single process.
This can be easily illustrated. Assume that a software product has been uploaded to an Internet server but on a given day no copy of the software has been downloaded from that server. In such a case no production and no distribution has been done. Now assume that on another day a hundred copies of the product have been downloaded. In such a case a hundred units have been produced and distributed. The number of units produced on any given day will solely depend on the number of units downloaded.
In the case of Internet-downloadable software, if it is not distributed then it is not produced. Thus, for software distributed over the Internet, distribution channels serve the very same function production lines serve for traditional products.
This merger of two separate processes - production and distribution - into one is the true unique characteristic of software distributed over the Internet. It results in any costs associated with distribution becoming costs associated with Factors of Production. As distribution is such a crucial part of software production, further analysis of the mechanism that is used to distribute software over the Internet and the costs associated with such distribution is needed.
Driving Internet Traffic as Distribution of Software
When we usually think of distribution we think of it in terms of mobilizing products from the location in which they are produced and stored to the market in which they are consumed (Ganeshan and Harrison, 1995). Distributing software over the Internet however is different. Looking into the way software is distributed, we can see that the market for these goods is formed on an Internet server where the software is downloaded; this is the location where supply is met by demand. The formation of this sort of market depends on consumers having access to a specific Internet server with a capability of downloading a given software product.
To illustrate this further we can look again at the aforementioned example. On the day no consumer accessed the Internet server and downloaded the product, no distribution had been made. On the day a hundred consumers accessed the server and downloaded the product, a hundred units had been distributed.
In the case of Internet-downloadable software, if there are no consumers accessing the server then there is no market, resulting in no distribution of goods to that market. Thus, for software distributed over the Internet, bringing consumers to the Internet server has the very same function as mobilizing traditional products to consumers.
Though sometimes referred to as "distribution" [ 2], the act of downloading by itself does not constitute an act of distribution, as it does not represent the act of bringing goods to a market where consumers can have access to it. Downloading is merely the act of transaction and the method used to complete such transaction.
It can thus be argued that for software distributed over the Internet, the act of distribution does not mean mobilizing products from the location in which they are produced to the market, as is the case with traditional products, and neither does it mean the act of downloading such software. Distribution is bringing consumers to the Internet server from which software products can be downloaded. This is most commonly referred to by those in the Internet industry as the act of driving Internet traffic to an Internet server.
Therefore, the practices used to drive traffic to a specific Internet server are actually production practices and the costs associated with it are actually costs associated with production. This way of looking at software production and distribution over the Internet opens an opportunity to look at the economics of software from a different perspective.
The Internet Distribution Chain
In order to examine the economics of software distribution over the Internet, a generalized distribution model upon which further analysis will be based must first be constructed. The model presented herein will be referred to as the Internet Distribution Chain.
The Internet Distribution Chain is a description of a process. The model defines the various phases involved in distributing software over the Internet, the metrics used in each phase and the quantitative relationships that exist within the chain. Because we regard distribution as an act of production, the model is describing the process of distribution within a distribution channel in a way that resembles the way production models describe production within a production facility. As the model includes both inputs in terms of costs incurred for establishing the chain, and outputs in terms of units sold, it is useful in analyzing the cost structure of software production and distribution.
The Internet Distribution Chain has four phases through which potential consumer traffic is driven. These phases are:
- Attention - Gaining consumer attention for the software product;
- Introduction - Introducing the software product to the consumer;
- Evaluation - Having the consumer evaluate the software product; and,
- Acquisition - Having the consumer purchase and use the software product.
Consumer attention is gained by using various promotion methods. The most common technique is online advertising using banners and text links. Various other business relationships between companies are geared toward achieving the same goal. These include among others co-branding, link sharing, search-engine prioritizing and service announcements to users.
Introducing the consumer to the software product is usually done using the vendor's Web site or any other Web site affiliated with the vendor from which consumers can download the software. Software vendors sometime seek to merge the Attention and Introduction phases of the chain, for example, by providing banner ads that enable direct downloading of software. This however does not change the structure of the chain, only the values associated with each phase.
Evaluation is done when the potential consumer is using the software product in such a way that enables her to assess its value. Generally Information Goods are considered Experience Goods, meaning products that the consumer has to experience in order to determine a specific value [ 3]. Most vendors distributing software over the Internet tend to regard their software as such a product and thus let consumers use the software for free under certain limitations such as time and functionality. However, even in cases where free evaluation is not possible, the consumer is still facing an evaluation process when she needs to make a purchasing decision. The merging of Evaluation with Acquisition does not change the structure of the chain, only the values associated with each phase. Consumer traffic driven through all phases will ultimately result in the purchase of a software product and the Acquisition of a customer.
The purpose each phase in the chain is intended to serve is only potential. When a banner is displayed, it does not necessarily generate consumer attention but rather the potential for gaining such. Only when the consumer clicks the banner or link can it be concluded that her attention has been grabbed. The same rule applies to the Introduction and Evaluation phases, only when the consumer has decided to download the product can we conclude that she has been introduced to it and only when she makes the purchase can we know for sure that she has evaluated the product. Thus, each phase within the chain is linked to the following one by the realization of potential actions. These links are utilization ratios between what was potential and what was realized.
Each phase in the chain has two values associated with it - Quantity and Price. The linking ratios permit the calculation of these two values for each and every phase in the chain. This makes the Internet Distribution Chain a generalized model that can be used in analyzing any observed distribution scenario, regardless of the way the transaction is constructed and to analyze it from an economic point of view.
Attention
Introduction
Evaluation
Acquisition
Quantity
Utilization
RatioQuantity
Utilization
RatioQuantity
Utilization
RatioQuantity
Price
Price
Price
Price
Internet Distribution Chain Metrics
Common metrics are used throughout the Internet industry to measure different variables associated with different parts of the chain (Brown, 2001). The metric used to measure the amount of potential consumer attention generated is the number of Impressions, meaning how many times a link to the Internet server has been displayed to potential consumers at the origin of the distribution chain. The cost associated with generating this potential attention is denoted as CPM (Cost Per Mila), that is the cost of displaying a thousand links to potential consumers at the origin of the chain. The distributing vendor can only assess the scope of attention generated when it has been realized, that is, when a link has been followed. The utilization of Impressions is denoted as CTR (Click Through Ratio), that is the percentage of links followed out of these displayed.
The metric used to measure the amount of potential consumer introduction generated is the number of Hits, meaning how many times the linked Internet server been accessed by potential consumers. The cost associated with generating this potential introduction is denoted as CPC (Cost Per Click), that is the cost of having a single potential consumer, originating from the source of the distribution chain, access the Internet server. The distributing vendor can only assess the scope of introduction generated when it has been realized, that is, when a software download has been made by a visiting consumer. Various names are used to refer to the utilization of Hits. It will be denoted here as DLR (Download Ratio), that is the percentage of downloads made by customers out of those accessing the server.
The metric used to measure the amount of potential consumer evaluation generated is the number of Downloads, or the number of consumers arriving via the distribution chain actually using a specific software product. Various descriptions are used to refer to the cost associated with generating a potential evaluation. Here it will be denoted as CPD (Cost Per Download), that is the cost of having a single potential consumer, originating from the source of the distribution chain, evaluate the software. The distributing vendor can only assess the scope of evaluation, when a software purchase has been made by the consumer. There are a variety of ways to describe the utilization of Downloads; here it will be denoted as AQR (Acquisition Ratio), that is the percentage of purchases made by consumers out of the total evaluating a given software product.
The metric used to measure acquisitions is the number of Customers. The cost associated with generating a consumer is denoted as CPA (Cost Per Acquisition), that is the cost of having a single potential consumer originating from the source of the distribution chain purchase software.
Attention
Introduction
Evaluation
Acquisition
Impressions
CTR
Hits
DLR
Downloads
AQR
Customers
CPM
CPC
CPD
CPA
Marginal Cost of Distribution
In recent years Internet traffic has become a commodity that can be bought or sold much like any other commodity on the market. Software vendors can buy traffic from Internet Web sites serving as distribution channels in any of the forms described in the chain. Impressions,Hits, Downloads and even guaranteed Customers can now be bought. However regardless of the form of the transaction, establishing an Internet Distribution Chain has costs associated with it. Unlike the costs associated with the creation of the first copy of the software, these costs are directly associated with each and every copy of software produced. It is thus impossible to regard these costs as Fixed Costs.
Having a Variable Cost associated with each copy of software enables us to look at software production and distribution from a traditional economic perspective, one in which the marginal cost of production and distribution is significant rather than from a perspective in which it is low, or even zero.
The following example illustrates how applying the Internet Distribution Chain model to a specific case reveals a considerable Marginal Cost of Distribution that can be allocated to the produced copy of software.
Consider a vendor seeking to distribute software and consider a major Internet portal as the distribution channel. The vendor is establishing an Internet Distribution Chain by acquiring banner ads on a specific section of the portal. She is buying 1,000,000 impressions at $1 CPM.
Assuming CTR is 2% the result would be 20,000 hits to the vendor's site.
CPC = Total Cost/Hits = 1,000/20,000 = $0.05Assuming DLR is 5% the result would be 1,000 Downloads.
CPD = Total Cost/Downloads = 1,000/1,000 = $1Assuming AQR is 5% the result would be 50 Customers.
CPA = Total Cost/Customers = 1,000/50 = $20
Attention
Introduction
Evaluation
Acquisition
Impressions
1,000,000CTR
2%Hits
20,000DLR
5%Downloads
1,000AQR
5%Customers
50CPM
$1CPC
$0.05CPD
$1CPA
$20
Using this Internet distribution Chain the software vendor has a Total Cost of $1,000 associated with distribution. The Marginal Cost associated with distributing one copy of the software is $20.
Variations in Internet Distribution Chains
Internet Distribution Chains, much like Factors of Production, vary in their yield and thus in their efficiency. Internet traffic is not uniform and the characteristics of the audience attracted by Web sites vary. As the audience varies so does its response to the message conveyed by each phase in the Internet Distribution Chain. A certain audience might respond well to a specific promotion message while another type of audience might fail to notice it. These variations in characteristics of the audience result in variations in the utilization ratios within the Internet Distribution Chain.
An Internet Distribution Chain originating in one Web site might not result in yielding the same values as another Internet Distribution Chain originating at another Web site. Even within the same distribution channel variations might exist and two Internet Distribution Chains originating in two different parts of the same Web site might produce different yields.
The metric used to measure the efficiency of an Internet Distribution Chain is the CPA, which represents a ratio between input in terms of cost and output in terms of units produced. The variations in yield causes variations in efficiency and in the CPA generated. For software distribution this means that the Marginal Cost of Distribution is different for each Internet Distribution Chain.
The following example illustrates how the Marginal Cost of Distribution varies between chains when the Internet Distribution Chain model is applied to a specific case of distribution via multiple chains.
Consider a vendor seeking to distribute software and consider a major Internet portal as the distribution channel. The vendor is establishing three Internet Distribution Chains by acquiring banner ads on three sections of the portal each with its own distinctive characteristics and audience. She is buying 1,000,000 impressions on each section at $1 CPM. For simplicity we will assume that the only change in the distribution chain would be a change in CTR resulting from variations in the audience addressed:
CTR1 = 2%, CTR2 = 1.2%, CTR3 = 1%
Assuming all other utilization ratios are the same, the Internet Distribution Chains would be:
Attention
Introduction
Evaluation
Acquisition
Impressions
1,000,000CTR
2%Hits
20,000DLR
5%Downloads
1,000AQR
5%Customers
50CPM
$1CPC
$0.05CPD
$1CPA
$20
Attention
Introduction
Evaluation
Acquisition
Impressions
1,000,000CTR
1.2%Hits
12,000DLR
5%Downloads
600AQR
5%Customers
30CPM
$1CPC
$0.12CPD
$1.67CPA
$33.33
Attention
Introduction
Evaluation
Acquisition
Impressions
1,000,000CTR
1%Hits
10,000DLR
5%Downloads
500AQR
5%Customers
25CPM
$1CPC
$0.10CPD
$2CPA
$40
Using these Internet Distribution Chains the software vendor has a Total Cost of $3,000 associated with distribution. The Average Variable Cost associated with distributing one copy of software is $28.60.
Average Variable Cost = Total Cost/(Customers1+ Customers2+ Customers3) = 3000/(50+30+25)=28.6
For the first chain the Marginal Cost of Distribution is $20.
For the second chain the Marginal Cost of Distribution is $33.33.
For the third chain the Marginal Cost of Distribution is $40.
Increasing Marginal Cost of Distribution and Profit Seeking Behavior
Internet Distribution involves the use of scarce resources. There is a finite number of Impressions that can be generated by a Web site acting as a distribution channel within a specific time period and thus there are a finite number of Internet Distribution Chains that can be established at any given moment. Further more, within this time frame there are a finite number of Impressions that can be used as inputs for each Internet Distribution Chain.
From the point of view of the vendor, Internet Distribution Chains are variable Factors of production. Using or not using chains is a short-term decision by the vendor. A vendor seeking to distribute software over the Internet will seek to distribute through Internet Distribution Chains that produce a higher yield resulting in lower CPAs. This behavior resembles the behavior of traditional manufacturers using these Variable Factors of Production that cost less prior to using those that cost more.
However, much like a traditional manufacturer seeking to expand production, the finite number of chains that can be established and the finite yield of these chains would result in utilization of additional Internet Distribution Chains with lower yield and higher CPA. The software vendor thus faces an Increasing Marginal Cost of Distribution.
Under such conditions the behavior of a profit-seeking vendor again resembles that of a traditional manufacturer. Such a vendor will continue to establish new chains and will seek to enlarge the number of Impressions acquired on such chains as long as the chains are profitable. A chain will be profitable if, and only if, CPA for that chain is smaller than or equal to the price of the software product. With such behavior the vendor's profit will be maximized when the Marginal Cost of Distribution is equal to the price of the software.
Summary
In this paper I have proposed the Internet Distribution Chain as a model for analyzing the cost structure of software production and distribution over the Internet. Using this model I have shown that software vendors face significant Marginal Costs of Distribution and that profit-seeking behavior by such vendors will bring the Marginal Cost of Distribution to equal the price of the software product.
Looking at the economics of software distribution over the Internet from this perspective opens the opportunity to look at pricing, competitive scenarios and equilibrium (or lack of such) using classic microeconomic models. It also calls for further research into the economic principles that guide the production and distribution of other Information Goods as well as into the economic principles that guide marketing and distribution of traditional goods over the Internet.
Revisiting the basic assumptions of information economics on a case-by-case base should result in better understanding of how these economics work and what specific characteristics distinguish it from traditional economics.
About the Author
Yaron Ilan is a Class of 99 MBA graduate of the Leon Recannati Graduate School of Business Administration at Tel Aviv University. For the past five years he has been distributing over the Internet software products made by a company he founded in 1996 and has served as an investment manager for Neurone Ventures, an Israeli Seed stage Venture Capital Fund affiliated with Bank of America. He currently is a contributing writer for Ha'aretz, a leading Israeli national newspaper and a marketing and business development consultant to Internet and software companies.
E-mail: yaron@evrit.co.il
Notes
1. Shapiro and Varian, 1999, ch. 1.
References
J. Aley, 1996. "Give it away and get Rich," Fortune, volume 133, number 11, p. 90.
W.B. Arthur, 1996. "Increasing Returns and the New World of Business," Harvard Business Review, (July-August), at http://www.santafe.edu/arthur/Papers/Pdf_files/HBR.pdf.
Y. Bakos and E. Brynjolfsson, 1997. "Bundling Information Goods: Pricing, Profits and Efficiency," Working Paper, MIT Center for Coordination Science, at http://www.gsm.uci.edu/~bakos/big/big96-12.html.
A. Brown, 2001. "E-Commerce and Marketing Dictionary of Terms," at http://www.udel.edu/alex/dictionary.html.
R. Ganeshan and T.P. Harrison, 1995. "An Introduction to Supply Chain Management," at http://silmaril.smeal.psu.edu/misc/supply_chain_intro.html.
J.K. MacKie-Mason and H.R. Varian, 1995. "Economic FAQs About the Internet," Journal of Electronic Publishing, volume 2, number 1 (May), at http://www.press.umich.edu/jep/works/FAQs.html.
J.K. MacKie-Mason, J.F. Riveros and R.S. Gazzale, 1999. "Pricing and Bundling Electronic Information Goods: Evidence from the Field," In: I. Vogelsang and B.M. Compaine (editors). The Internet Upheaval: Raising Questions, Seeking Answers in Communications Policy. Cambridge, Mass.: MIT Press, pp. 277-305.
M. Romano, 1998. "Is there such a thing as a software monopoly?" Salon (November), at http://www.salon.com/21st/feature/1998/11/11feature.html.
C. Shapiro and H.R. Varian, 1999. Information Rules: A Strategic Guide to the Network Economy. Boston: Harvard Business School Press.
H.R. Varian, 1997. "Versioning Information Goods," research paper prepared for Digital Information and Intellectual Property, Harvard University (23-25 January), at http://www.sims.berkeley.edu/~hal/Papers/version.pdf.
H.R. Varian, 1996. "Differential Pricing and Efficiency," First Monday, volume 1, number 2 (August), at http://firstmonday.org/issues/issue2/different/.
H.R. Varian, 1995. "Pricing Information Goods," presented at the Research Libraries Group Symposium on "Scholarship in the New Information Environment," Harvard Law School, Cambridge, Mass., at http://www.sims.berkeley.edu/~hal/Papers/price-info-goods.pdf.
Editorial history
Paper received 26 October 2001; accepted 21 November 2001.
Copyright ©2001, First Monday
The Economics of Software Distribution over the Internet Revisited by Yaron Ilan
First Monday, volume 6, number 12 (December 2001),
URL: http://firstmonday.org/issues/issue6_12/ilan/index.html
