In 2006, Airbus experienced an unexpected two-year delay in the development of its A380 megajet. When preassembled bundles of hundreds of miles of cabin wiring were delivered from its factory in Hamburg, Germany, to the assembly line in Toulouse, France, they failed to fit into the planes. Assembly ground to a standstill. The entire wiring system had to be redesigned and rebuilt. The price tag: $6.1 billion. The reason given for one of the most expensive missteps in the history of commercial aerospace: incompatible design software.1
Industry experts frequently blame disappointing IT performance on a lack of centralized control over IT purchases. While the Airbus example may be an extreme case, many believe decentralization of IT governance has led to ever-growing costs of technology ownership. Yet evidence also exists that centralized IT solutions fail to address the full range of needs within large multidivisional firms. A poor alignment between the technology and local business needs forces business units to reengineer their processes to fit the technology or simply work around it. A Harvard Business School case on IT at Cisco describes a "shadow IT system" that grew up around unmet local demands for customized data and reports.2
How should firms decide whether to centralize or decentralize the selection and purchasing of critical IT infrastructure? This question has persisted in the information systems community at least since the 1970s, with periods of renewed attention whenever significant technical change reshuffled the board. The diffusion of the commercial Internet in the mid-1990s represents one such tectonic realignment. My recent study forthcoming in the Journal of Economics and Management Strategy provides new evidence that, the advice of IT consultants notwithstanding, there is no single best way to structure this process.3 Firms are, in fact, more diverse in their approach than is widely recognized. The best choice depends critically on the specific business context in which the IT will be deployed.
Reality is Complicated
This multi-industry study focused on approximately 3,000 firms in the U.S. manufacturing sector to understand how they allocated authority for IT purchases. The data is from 1998, a time of vigorous investment in IT, but a period as-yet unaffected by the bursting of the dot-com bubble. The IT purchases in question included network equipment, servers, terminals, and enterprise software applications that arguably work best when they are interoperable throughout the firm. The IT governance data come from the Harte Hanks Computer Intelligence Technology Database, which surveys establishments with more than 100 employees on their use of information technology. In particular, they ask whether the authority to make IT purchases resides with local managers or with the corporate parent. This data was merged with the U.S. Census of Manufactures to gain insight into the local business environment of individual plants and the broader firm and industry context in which they operate.
Because of the wide diversity of firms in the sample, one would expect to see a variety of purchasing regimesparticularly among firms in different industries or of different sizes or vintages. A surprising finding, however, was the variation that existed within firms. The accompanying figure shows that firms tended neither to be fully decentralized nor fully centralized, but had local purchasing authority at some locations and centralized control at others. This is in sharp contrast to widespread advice to high-level firm managers to centralize IT purchasing for large organizations.
Adaptation vs. Coordination in Firms
Why was this the case? Were the instances of delegation simply "mistakes" to be corrected? To better understand this surprising diversity of approaches, the study leveraged an important model from the economics literature to explore whether trade-offs between "adaptation" and "coordination" might explain the patterns in the data. According to this type of model, both decentralization and centralization are reasonable outcomes for profit maximizing firms (that is, neither is a "mistake"). Instead, the former makes sense in the presence of particularly compelling local business needs that demand high levels of adaptation; the latter is desirable when coordination is of paramount importance. Ultimate outcomes depend on the magnitudes of these competing needs throughout the firm.
How should firms decide whether to centralize or decentralize the selection and purchasing of critical IT infrastructure?
This is, indeed, consistent with what the data indicate. For instance, if a plant is a major contributor to total firm sales, getting the IT "just right" at that location trumps the benefits of firmwide coordination. Adaptation is also important if a plant produces products outside the main focus of the firm. In this case, local managers may know better than those at headquarters what the plant needs. More-diversified firms also tend to delegate more, presumably because the diversity of IT needs grows with the range of firm products. The presence of legacy systems or growth through mergers and acquisitions also is associated with delegation.
These findings are in line with the notion that centralized managers have too much to contend with in the face of changing IT solutions, upgrades, and customization requests across large, distributed firms. The benefit to the firm of letting local managers make well-adapted choices in economically sensitive or less-common circumstances may outweigh the costs of degraded coordination, overall.
This is not true in all cases, however. In firms where the value of IT interoperability is particularly high, the likelihood of centralized purchasing is much higher. For instance, firms that must coordinate production across locationsand whose IT must be able to communicate across plantsare significantly less likely to delegate. Also, firms apparently do not give up on monitoring what local managers do. In firms that are large enough that central managers may have trouble "keeping tabs" on plant-level purchasing managers, delegation is somewhat less likely. Finally, variation in IT governance structures holds across the range of IT spending: even the biggest IT budgets are often left to local decision-makers.
One "Size" Does Not Fit All
Firms must weigh these many factors in deciding whether to keep the purchasing authority for IT systems at headquarters or farm it out to individual plants. The analysis provides evidence that one approach is not necessarily better than the other in all circumstances. In addition, it points to the need to align the IT governance approach with other key choices at the firm such as its growth and diversification strategy, reliance on legacy IT systems, and the distribution of critical operations within the firm. These findings raise key questions for managers to keep in mind when making this important decision:
- Are the IT needs of this business unit consistent with what exists throughout the firm? Is this a "special case" that needs a different approach?
- Can our central IT division manage the coordination challenges in a firm this size and/or scope? If not, would strategic delegation to some units make better sense?
- If coordination is absolutely critical, do we have the centralized IT resources to manage this well? How can we be sensitive to local needs while adhering to unified standards throughout the firm?
- How important is this one unit to overall firm performanceif a compromise must be made in terms of technology fit, is this the place to make it?
Getting the IT purchasing decision right is a critical one for firm performance. Anecdotal evidence points to the pitfalls that arise from mistakes in both directions. Firms that fail to coordinate when interoperability is critical risk Airbus' fate. Yet, forcing inappropriate standards onto idiosyncraticand economically importantinternal operations is also an unattractive option. Examples of "too much coordination" may make the news less often than a single $6 billion coordination failure, yet the costs of daily workarounds and poor IT fit may be quite costly in the long-run, as Cisco can attest.
Weighing the costs and benefits of a monolithic approach for the entire firm is unlikely to be a fruitful endeavor. Instead, managers should consider a case-by-case approach that may better accommodate the diversity of business contexts that exist within their particular firms.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2012 ACM, Inc.