The cloud computing economic model is expected to bring significant rewards – apparently. Those rewards may be possible but the quality of analysis to demonstrate that the cloud paradigm will yield an ever-growing margin is far from assured.
The assumptions underlying the economics of the cloud are tenuous and therefore the promotions and promises should be treated with caution. Supporting claims from vendors with commercial interest do not add greater clarity and their use of concepts about how ‘the firm’ operates are not empirically grounded.
In many ways the economic model of cloud computing production is akin to a new mechanization, such as the development of factories. But the hype conceals challenges, typical in any first phase transition to another method of working.
In essence cloud computing offers the potential for greater productivity from more efficiently used and deployed assets and resources. The production architecture of cloud computing could simultaneously harness the following forces:
• Enormous economies of scale once investment is operational, and;
• Network effects, which is the power of large groups using similar services or products where leveraged productivity benefits are accrued e.g. Microsoft Office
• Network effects of data in common structures, allowing sharing of information between systems (and even organisations) that were previously isolated.
In such an environment the opportunity cost of moving to cloud computing is paramount because it entails a commitment in ‘sunk costs’: those are costs that underpin the venture. This must be viewed together with potential yield, i.e. the carrot in the hype.
In the cloud computing model the scalability of an operation removes some capital cost (at the single organization level) and labor. Therefore in the case of Amazon it can service clients at a lower ‘use on demand price’ than the client would pay had it invested in the same services. Boston Consulting has enumerated related advantages to cloud computing below:
• Reduced capital requirements for up front investments in IT since the enterprise is able to utilize the infrastructure, applications, and platforms in the cloud
• The flexibility to meet sudden changes in demand and peaks and troughs
• Significant cost savings in selected situations notably when the scale of an enterprise computing resources is relatively small compared with that of cloud providers
While these advantages seem ironclad, when real companies are surveyed, the realities of competition demand curves and price are not identical with the arguments that underpin the features listed above.
How firms really operate
A 1998 US survey demonstrated a number of salient features of business practice and the management of costs and prices. It also showed how prices change in the wider marketplace:
• “First, about 85 percent of all the goods and services in the U.S. non farm business sector are sold to “regular customers” with whom sellers have an ongoing relationship … And about 70 percent of sales are business to business rather than from businesses to consumers.
• Second, and related, contractual rigidities … are extremely common … about one-quarter of output is sold under contracts that fix nominal prices for a nontrivial period of time. And it appears that discounts from contract prices are rare. Roughly another 60 percent of output is covered by Okun-style implicit contracts which slow down price adjustments.
• Third, firms typically report fixed costs that are quite high relative to variable costs.”
The survey showed a high degree of stable competition and fixed prices. Instant demand peaks and hyper-flexibility are not everyday concerns. Prices are stable and customer relationships static. There are some firms that require or can use flexibility, but not commonly, and that has wide-ranging effect on the aggregate outcome.
Real firms’ view of capital and operating expenditure are divergent from theory. Approximately 90 per cent of their output is produced under conditions of constant or falling marginal cost, and rising economies of scale.
The Fallacy of Composition and Total Cost of Ownership
Stable and constant business to business commerce is not normal in arguments or case studies supporting cloud computing. Cloud vendors tend to overstate the unusual instance and extrapolate it to a so-called typical condition of operations. This type of argument and logic at first sight appears quite convincing. It actually confuses any issues.
Independent studies and most, if not all, vendor claims on cloud savings use a simple formula, viz: the evidence from a single study is extrapolated to several or all other players within an industry or sector. The cumulative savings from this methodology, notably in government services are potentially large. Both Microsoft and Google offer the same formula when they present total cost of ownership savings for an individual firm or even for multiple organisations based on single firm or study.
The intention in demonstrating cloud savings has occurred is to rationalize the need for a process of migration and change. The trouble is that the purported savings are not realizable in the typical methodology that is presented by vendors and consultancy firms. This is a problem for organisations looking to achieve similar savings e.g. government agencies, the education sector where many individual organisations share and are joined by common requirements when they share many services. It is also a consideration for organisations joining private clouds i.e. smaller agglomerations of users usually within a specific sector.
In the first instance claimed savings across several organisations is subject to the Fallacy of Composition which refers to errors in logic that arise “when one infers that something is true of the whole from the fact that it is true of some part of the whole (or even of every proper part)”. It holds that what is true for one individual will be true for all others. A common example of the fallacy is that thrift is good, which may be true of an individual but if an entire nation or market practice it then demand falls and consequences, such as unemployment, and falling investment may result.
If something is logically invalid it may be considered either irrelevant or just acceptable, depending on attitude, because not everything can be made perfectly logical in real business environments.  Most vendors emphasize that the examples they offer of purported savings are an illustration and therefore to be read as indicative, not absolutely definitive. While such a position is understandable, and flexible, it is still wrong because it is an inference about something that is supposedly true, or is understood to be in some way true.
The forecast value of any cloud deployment based on the inferences of prices is incorrect. In addition, an illustration of something that is fundamentally wrong is not useful, it is still wrong, it illustrates an error. It means the methodology is wrong; the argument that coheres the purpose of the savings are also wrong and most of all it cannot be used to build a business case.
Applying the fallacy of composition to proposals and data that present savings is instructive into what can really be achieved. In reference to total cost of ownership (TCO) the fallacy of composition can be applied to all vendors’ TCOs as the claimed savings is the element of individual thrift of one firm applied, and/ or inferred to other firms in equal proportion and even across whole industries . For single firms planning on choosing Microsoft’s 365 cloud offerings or Google Apps the fallacy of composition may be a useful test.
The research on cloud computing has mostly used single instances, when larger groups of organisations ought to be analyzed. One of the key aspects to examine is aggregate demand and utilization in which a single organization would be compared to several. It is aggregate utilization and network effects across several organisations, not necessarily within the same industry, which will reveal how the cloud alters costs in relation to productivity.
For larger groups of organisations, such as school districts, or government agencies a more substantial method of analysis is needed to assess future value. This must balance individual organisations with the aggregate. It may also understate the achieved value in either savings or additional productivity.
The conundrum is known as the Aggregation Problem. The aggregation problem refers to the difficulty of treating an empirical or theoretical aggregate as if it reacted like a less-aggregated measure, for example, about the behavior of an individual agent. In other words taking a single instance and extending its results to other instances. This problem is considerable but ignored.
In most models of cloud savings: Public cloud, Private cloud et.al., as the metrics used to evaluate individual cost benefits for each scenario is costed and benefited seem to apply a uniform cost without attention to aggregation. The results of those models are therefore inaccurate. Most of the material (both independent and from vendors) dealing with government cloud savings demonstrates it is never referenced, never solved, and therefore remains a problem.
It is not possible to offer solutions here to such a complex analytical problem.
There are, however, two ways to consider this issue:
1. Vendors and any consulting firms must demonstrate that they have understood and dealt with the problem and have an appropriate algorithm or other mathematical model that answers the issue;
2. Anyone attempting this to produce a verifiable business case ought to start on a small-scale e.g. 2-4 firms or agencies and also ensure they are a similar types of organisations i.e. size, operations etc.
If organisations are going to realize financial benefits in cloud computing the vendors of those products must do a better, more comprehensive and thorough job at explaining them.
Any organization seriously reviewing the return on investment and wider changes of cloud computing is necessarily going to have to assess their own organization over a long period. And that process will require the skills of outside expertise to complete such a task.
• Appraise the value and the inflection points of business that would improve.
• A productivity evaluation is very complex but a yardstick is critical to make.
• Where is the industry heading and how quickly: what are the forces and incentives to change and how extensively?
• Be skeptical about claims and also thorough with measurements.
1. Blinder surveyed a representative weighted sample of US non-agricultural corporations with annual sales of more than US$10 million; a 61% response rate resulted in a study of 200 corporations whose combined output represented 7.6% of the USA’s GDP. The interviews were face to face, with Blinder and a team of Economics PhD students conducting the interviews; the interviewees were top executives of the firms, with 25% being the President or CEO, and 45% a Vice President.
2. A causal relationship: http://en.wikipedia.org/wiki/Okun%27s_law
3. Asking About Prices: a New Approach to Understanding Price Stickiness, Russell Sage Foundation, and New York. Blinder, A.S., Canetti, E., Lebow, D., & Rudd, J., (1998). p. 302
4. Another contemporary example of a fallacy which appears correct is related to the reconstruction of Queensland after the floods. This situation is presented as a potential boon for the economy yet the idea of economic growth from destruction is an error known as the fallacy of broken glass and identified 150 years ago.
5. This is commonly done in anecdotal pieces in IT media, e.g. Firm X migrated to ABC product and if every firm did the same then there would be a saving of $XX.
6. An aggregate in economics is a summary measure describing a market or economy.