One of the central problems in developing high tech systems is that there appear to be unavoidable trade-offs between managing the scale and connectedness of emerging high tech systems. Everything depends on everything else, even well bounded tasks are complicated by unexpected dependencies on hardware or other technologies. What do we need to know in order to identify, describe, and address these various complications and difficulties adequately as they arise? We will begin the process of understanding the problem domain and approaches to addressing its difficulties by illustrating economic aspects of information goods, some work aspects of software engineering (digital production), and the classic dimensions of project management.
New media and information industries are refining if not redefining our knowledge of the economics of markets, products, services and production. Broadly the challenges involve information goods, high tech systems and bases or markets of user/consumers. However, while the business models, technology and material foundations of these new ‘goods’ are constantly changing, the principles of economics do not. There are many dimensions along which information goods and systems are different from purely material products and services. Information goods are ‘experience goods’ (‘consumed’ by experiencing or operating), they are subject to the economics of attention (if you are not paying for the product you are the product), and the technology itself is associated with pronounced production side scale, product feature scale, and greater potentials for user ‘lock-in,’ ‘switching costs,’ and ‘network externalities’(Shapiro and Varian, 1998). For the purpose of this section we will focus on the economic case for production side scale economies of software. The same logic applies to other digital media and products with digital media components.
Unlike many physical goods, information-rich products have some distinctive economic qualities and characteristics. Like all information-rich goods (products like computer hardware, books, newspapers, film, television) the initial design and production of the first copy of a software product demands a huge up front investment in development before there can be a payout. Unlike physical goods manufacturing, the mass reproduction of a digital or information rich good is a simple trivial act of copying. For pure digital goods there are vanishingly small incremental costs in terms of the energy used, storage space and time taken to duplicate. In software manufacturing (if such a term can even apply in the current era) the development costs far outweighs the reproduction costs. Development costs dominate the economic cost characteristics of software.
The following presentation is adapted from Oz Shy’s book, The Economics of Network Industries (Shy, 2001). The argument goes as follows; that as sales/consumption of the product increase the average cost of product approaches the cost of producing and delivering the next unit (the marginal cost). In the case of a purely digital good (software, information, media, etc) the marginal cost of production is very small and often carried by other parties (e.g. the broadband service). Applying the logic of cost based pricing model to your product suggests a strategy of effectively giving it away.
The total cost of production at a particular level (TC) is the sum of the sunk R&D costs plus the cost of producing and shipping ‘q’ units. By definition the total cost of production at a particular level (q) is the sum of the cost of R&D (cost of developing, testing and releasing software) plus the cost (μ) of shipping one copy to the customer.
1. TC(q) = θ + μq
If we define the average cost (AC) of production of a product as the cumulative total cost of production at a particular production level divided by ‘q,’ the quantity produced (ideally also sold) at that level.
2. AC(q) = TC(q)/q
The average cost becomes:
3. AC(q) = θ/q + μ
The marginal cost at a particular product level, or additional cost as a result in a small increase in the production level is the incremental additional cost divided by the change in quantity produced.
4. MC(q) = ∆TC(q)/∆q
And in the limit (the differential wrt q of equation 1).
5. MC(q) = μ
A graphical analysis (below) of average and marginal software production cost as functions of quantity demonstrates that the average and marginal costs converge at high output levels (Shy, 2001).
Figure: Cost and price characteristics of software (adapted from Shy, 2001)
The implication of this analysis is that for every price you set there exists a minimal level of sales for which any additional sale will result in a profit. One conclusion from this argument is that ‘cost based pricing’ is not a viable strategy for software because there is no unique break-even point (Shapiro & Varian, 1998). In effect the logic follows that the more units you sell the lower you can set your price. The logic of cost based pricing suggests you should charge very very little, or give software products away (if there is a large potential market for them). So software markets are subject to huge economies of scale.
“the more you produce the lower your average cost of production.”
(Shapiro and Varian, 1998)
HIGH TECH PRODUCTS AND PLATFORMS AS ECONOMIC SYSTEMS
I previously characterised the dominant aspect of high tech products to be: their intrinsic complexity and propensity to change. Both aspects lend high tech products to exhibit ‘systemness’ or systematicity within their environments. For example software itself may be both a product and a platform. To illustrate; interdependencies arise whenever a software program makes functions accessible via an API (Application Programming Interface). APIs allow other programs to use the first program. The consequence is that a combination of the two programs allows us to accomplish something new that we couldn’t do with the separate programs. These effects are termed complementarities and it gives rise to system-like effects in the computing environment and in the market for high tech products and services (Shapiro and Varian, 1998).
Complements and Combinatorial Innovation
When goods that are complements are produced the combination of two products become more desirable and valuable to users than the products alone (Figure below).
Figure: Compatibility drives profits (and utility)
Furthermore we can show that the economic effect of complements dictate that
“aggregate industry profit is higher when firms produce compatible components than when they produce incompatible components.” (Shy, 2001)The reason being that the sunk cost of R&D can be averaged over a larger market; and larger markets are generally better for all firms, even competitors regardless of their market share.
“the firm with smaller market share under compatibility earns a higher profit under compatibility.” (Shy, 2001)This is because the market itself is generally larger, thus the marketing strategy question ‘do you want a large piece of a small pie, or a small piece of a much larger pie?’ Why is this relevant? It is relevant because it is one of the dynamics that drives change in the operating environment of organizations. Synergies in Internet services and platforms have driven constantly expanding integration and adaptation, change and innovation. The internet boom of the 90s through to today is largely a consequence of 'recombinant growth' or combinatorial innovation of general purpose technologies (Varian et al., 2004). The idea of combinatorial innovation accounts in part for the clustering of waves of invention that appear whenever some new technology becomes successful. The ubiquity of one program can act in turn as a platform for other programs; for example the mutual complementarities between Twitter, Bit.ly, and Facebook. Much of what is termed Web 2.0 computing can be thought of as leveraging complementarities of different technologies that in turn creates clusters of innovation.
Compatible Products are Driven in Turn by Market Standards
Markets incorporating complements and compatible products welcome technological standards (Varian et al., 2004). Standards are desirable because they facilitate complements and compatibility. Open standards are better because of the free availability of technical rules and protocols necessary to access a market. However even a closed or proprietary standard is preferable to none as it provides an ordering influence, providing rules or structures that establish and regulate aspects such as interoperability or quality. Network effects arise from the utility consumers gain from combinations of complementary products (Shapiro and Varian, 1998, Shy, 2001, Varian et al., 2004).
The very simplest network effect can be illustrated by the example of fax machines. The first purchaser of a fax machine has no one to send a fax to. A second fax machine bought by the first buyer’s friend allows them to send faxes to each other, which is somewhat useful. However if there are thousands of fax machines, in firms, government agencies, kiosks and people’s homes then the fax machine becomes more useful to everyone. As the market becomes larger the usefulness, or utility, of fax machines as a class of technology becomes greater.
The principle applies equally to single categories of networked technology like fax machines as it does to families of technologies that can interoperate. Network externalities arise between automobiles and MP3 players if auto manufacturers install audio jacks or USB ports to connect the car’s sound system with the MP3 player. The utility of both cars and MP3 players increases. Standards and openness drive further growth and innovation (and lock-in and switching costs etc). Standards enable software markets that in turn enable hardware sales that enable software etc. all enabled by a standard.
Software has a unique role as the preeminent enabling technology for hardware and has unexpectedly led to software becoming the platform itself. Software – operating systems or execution environments like browsers and browser-based ecosystems like Facebook – enables developers to achieve a degree of independence from the hardware. Such platform software becomes essentially a new type of standard that may itself be open or proprietary and the same economic models dealing with complementary goods and compatibility apply. Therefore the same kinds of innovation clustering producing waves of combinatorial innovation, can be seen to occur with successful platforms.
A software platform benefits from the variety of add-in software written for the platform and this in turn generates a virtuous cycle of value growth and further innovation as products are re-combined and used in novel ways. In summary the economic characteristics and market logic of software products drive them towards interdependency with other software, standards play a huge role in enabling this (closed or open). The whole context of software production exists within an ecosystem of different products and services, which are in effect environments or platforms themselves and these arguments explain in part the ever-expanding ubiquity of software within technological systems.
The various engineering professions (civil, mechanical, chemical etc.) typically separate design work from production work, treating production via either the project management perspective for once-off style constructions, or via the process control perspective for managing operational environments. However software software production (design and development) has proven difficult if not impossible to control via predominantly construction perspectives or as manufacturing processes, why is this? Why shouldn't software engineer lend itself to the kinds of management instruments that proved so successful in the classical sense of Fordist production? Why isn't software like more like civil engineering for example?
Well, the digital economy is subject to some interesting essential and intrinsic characteristics that, while not absent in physical goods markets, occur in greater and lesser extent in comparison with physical goods. In the case of digital production the process of manufacturing the end product is a trivial exercise of electronic duplication with marginal costs of manufacturing additional copies being effectively zero. Therefore the production cost characteristics of software and many high tech goods shift to a focus on the process, effort and cost of producing the first unit. Software is costly to develop but cheap to reproduce. Multiple copies can be produced at a constant or near zero per-unit cost. There are no natural capacity limits on producing additional copies. The costs of software production are dominated by the sunk cost of R&D. Once the first copy is created the sunk costs are, 'sunk'! Software production costs will therefore be dominated by employee/human costs (salaries and servicing the working environment) rather than material costs (computers).
This initial analysis seems to suggest that software development efforts should be treated like stand-alone projects, i.e. time bounded design and development of a finished product. This is indeed a characteristic of many industry settings, e.g. for device/hardware software in telecommunications, for robotics, for mission critical systems in aviation and aerospace, for critical infrastructure such as energy distribution and core or internet backbones.
Software design and development produces little if any substantial material assets or residues. Software production models should therefore emphasize design activities rather than manufacturing activities. Software R&D (the cost of developing, testing and releasing software) is a human knowledge intensive activity. The consequence therefore is that while a software firm’s strategic advantage is manifest in its products, its competitive capability is bound up in its employees’ design knowledge and experience.
But software and high-tech yields a new kind of cornucopia, a wealth of value that is becoming more significant and more freely available. Software begets software and systems support other systems. The whole technological infrastructure of microprocessor led, computer driven, software and high-tech device innovation has kept producing value and benefits for organisations, markets, and society at large for 50 years or more. The fact that it continues to evolve and is still implicated in societal transformation suggests it will continue for a while longer.