Seldom before has any technology promised so much to so many. If you believe what you read, so-called "client-server" computer systems will cut the cost of information technology, make computers easier to use, provide unprecedented computing power, speed systems development, and transform the shape of organizations as well as the nature of management.

To winnow the truth from these claims, we undertook an in-depth study of five large corporations, each at a different stage of implementing client-server solutions, in five different industries: consumer packaged goods, insurance, regional banking, transport, and investment banking. In each case, we interviewed both the CIO and line managers, and analyzed the costs and benefits of in-place client-server systems. We then validated our findings through a series of roundtable discussions with managers from another 20 large and medium-sized companies.

What we found, in brief, is that client-server technology does create unique opportunities to implement ambitious programs of corporate change. It makes possible computer systems that are more flexible and easier to modify than their predecessors. And it eases the task of redesigning the way decisions, especially cross-functional decisions, are made within an organization. Inevitably, however, significant challenges stand in the way of those who would grasp these new capabilities.

Not only does building client-server systems demand the mastery of new technologies; it also requires a new relationship between information technology and the businesses it serves. Instead of being simply an asset that produces benefits measured largely in terms of cost savings and productivity, information technology becomes a service that promotes and enables change. This, in turn, means that IT professionals must transform themselves from "engineers," whose prime responsibility is to keep IT assets working at maximum efficiency, to "coaches," whose role includes helping an organization to identify and develop the people and skills it needs to manage change successfully.

To evaluate client-server in terms of its technological costs and benefits alone—as many companies now do—is to miss the point

Given this scope of promise, to evaluate client-server in terms of its technological costs and benefits alone—as many companies now do—is to miss the point. In our experience, at least as many client-server projects have run into difficulties because of a failure to manage the organizational consequences of the technology as have faltered through problems with the technology itself. Although the risks and benefits of these projects will naturally vary from company to company—depending largely on whether a company needs to change its information technology, its business processes, or both at once—all contain an organizational as well as a technological element. Both must be managed together.

What client-server is—and isn’t

Three characteristics set client-server computing apart from the rest of the IT world:

1. Division of labor. Client-server systems divide up the work of computing among many separate machines. This contrasts with mainframe-based systems, which centralize the work in one place. With client-server, there is often a central "server" responsible for maintaining data (as well as the databases which access that data). Desktop "clients" then draw the data from the server and perform calculations upon it. A server might, for example, hold sales data, which desktop clients then use to do forecasting.

Networks transport data between the various machines that work on it, just as roads and railroads transport components between factories

2. Network infrastructure. Given this division of labor, client-server systems rely on networks to weld together disparate computers into a cohesive whole. At a minimum, the networks provide for transportation of data between the various machines that work on it, just as roads and railroads provide for transportation of components between factories. Many companies are taking this approach a step further, and building commonly used computing services directly into the infrastructure. Providing, say, document management from a central location, and making it accessible over the whole network, can simplify the management of information systems and reduce costs.

3. Emphasis on the user. Much of the power of client-server goes into making new applications engaging and easier to use. It provides users with greater latitude in choosing software presentation, and software designers with greater latitude in creating programs that are pretty, readily customized, and modern—such as user interfaces with multiple windows, graphics, and a point-and-click style of entering commands. More important, changes to a program on a "client" desktop need not affect programs or data on the server. As a result, improving systems is much quicker and simpler than when all computing has to be provided through a one-size-fits-all package on a mainframe.

Dispelling the myths

Numerous claims about client-server are not borne out by the facts. One myth is that the move to client-server architectures is mostly driven by a wish to reduce costs. True, many of today’s personal computers and workstations offer processing power comparable to that of a minicomputer or a mainframe at a significantly lower price. In practice, however, money does not seem to be the primary concern of those moving to client-server.

Most of the systems we encountered had been chosen because they could do things that mainframe systems could not

Although we did find some examples of client-server systems being used to replace obsolete, expensive to run minicomputers and mainframes, most of the systems we encountered had been chosen because they could do things that mainframe systems could not. Either they provided new capabilities, or they allowed systems to be developed more quickly—or they did both at once. Cost savings, when they did occur, came as a bonus.

A second familiar claim is that client-server technology is capable of supporting only lightweight applications like decision support, not heavyweight, mission-critical functions like transaction processing. This alleged limitation rests on a too limited definition of the technology. Many companies do, in fact, employ client-server for complex transaction processing.

A major investment bank, for example, uses client-server to trade a variety of financial products across a variety of offices. Indeed, the system is capable of managing trades all the way through to cash settlement with the Federal Reserve. Although most of the transaction volume is still handled on mainframes, these mainframes have simply become components in the larger client-server system.

A final claim is that the technologies underlying client-server are still too immature to be scaled up to support really big applications. Although most of the executives we interviewed believe that client-server systems are inherently more complex to manage than those based on a single mainframe or mini-computer, they also feel that their complexity is manageable—and that their benefits far outweigh their risks. Part of the reason is that the software used to manage client-server is maturing very quickly.

One of the most ambitious implementations we studied involves the core booking and logistics system for a large transport company. It is using client-server to integrate several disparate older systems into a common platform for 17 major new applications. This massive integration and development project is being accomplished in under three years—primarily with off-the-shelf tools and products.

A focus on organizational change

Underlying all these false claims is an even greater misconception. Much of the discussion about client-server assumes that the technology is best evaluated in technological terms. Not so. One of the strongest influences on the decision to adopt client-server is its potential impact on organization.

Client-server changes the distribution of decision-making power within an organization

"Information," runs today’s cliché, "is power." By freeing information from the straitjacket of centralized, mainframe-based systems, client-server changes the distribution of decision-making power within an organization. It can, for example, give salespeople a complete overview of a particular customer relationship via their notebook computers—thus allowing them to make their own judgments about that customer’s most important needs, instead of following standard order-taking procedures. Equally, the availability of extra information can enable production workers to take more responsibility for scheduling work and improving quality.

In the back office of a major bank, the mainframe-based system for trading securities transactions was designed around 24 "control points"—each representing a decision that could be taken by a manager regarding the progress of a transaction. With the installation of client-server technology, every one of these control points changed.

Companies are moving to take advantage of the new organizational capabilities created by client-server...

...by trying to improve the ability of their organization as a whole to make better decisions. Often, this is a matter of making the whole at least as good as the sum of its parts—that is, ensuring that well-informed islands of decision-making prowess are not lost in a sea of ignorance. This may require decisions to be made more rapidly, or by different people in different places. It may also mean changing the kinds of decisions that get made.

...by trying to bring together information from different departments, different functions, and different companies. Providing a comprehensive solution to a business problem often depends on assembling information from a variety of sources and making it available within different (usually shorter) time frames. The ability to respond to unexpected demands for information, perhaps from unusual places, is crucial.

Having redesigned decision making once, managers usually realize that they will want to do so again

...by trying to exploit the additional flexibility inherent in the technology to build information systems that can change as fast as the underlying business does. Having redesigned decision making once, managers usually realize that they will want to do so again—or indeed continuously—as their business evolves. Client-server architectures make it relatively easy to change the information that servers bring to the desktop as well as the computations performed there.

Capturing the benefits

The most successful client-server systems we encountered created benefits through the business changes that they made possible, rather than through the technology itself. In fact, all the successes we observed focused on solutions that enabled different decisions, in different locations or by different people, within different time frames. One particularly successful system provided the salesforce of a consumer goods company with more and better information on the products they sell. As a result, the salesforce has taken over—to good effect—some decisions previously made by retailers.

In addition, this system permits greater customization and flexibility in marketing. Previously, information on the results of product promotions like coupons or special prices was gathered weekly, region by region, on a centralized system. The new technology collects this information daily, store by store. Decisions that were formerly the province of regional directors can now be made by salespeople able to tailor focused promotion strategies to each store. Total costs of the system were over $30 million, but the base-case return on investment is over 50 percent (with substantial further upside if sales hit more aggressive targets).

Conversely, companies that did not take advantage of the changes in decision making made possible by client-server failed to reap the benefits of their new systems. At the same consumer goods company, for example, a client-server solution was deployed to automate the collection of statistical process control data for a production line in a regional factory. The system allowed line workers to input changes in process parameters (product size, height, color) directly into personal computers stationed on the factory floor. This data collection had previously been carried out manually and input by clerical staff overnight.

Although the new system changed the timing and location of data collection, it had only a limited effect on the decisions made by line supervisors. This was because the major decision-making benefits had already been captured when manual statistical process control was introduced. Thus the immediate benefits from client-server were quite small and did not justify its incremental costs. However, since much of this investment was in infrastructure—personal computers, networking hardware, and basic computer training for the plant workers—the company is now trying to run additional applications on the same platform, which could dramatically alter its return on investment.

Striking the balance

Companies seeking to adopt client-server must manage change along two dimensions: technology and organization. Simply making client-server computing work means installing new server hardware, end-user computers, and software; building network infrastructure; and creating new skills within the IS department. Reaping the business benefits, however, means harnessing the capabilities of that new technology to change organization structures and decision-making processes. But it also means fostering process and/or organizational change within the business to exploit those new capabilities. Any client-server strategy must, of course, pay close attention to both these dimensions of change. The tactical risks and benefits for any business will depend on the relative urgency of technological and organizational change.

There are four possible ways of getting this balance right:

Opportunistic. Some companies decide to adopt client-server before they face an urgent need either to improve their basic technology or to change their business processes. This approach implies that they have two or three years in which to gain experience and tap the benefits of client-server on specific projects of their own choosing.

Infrastructure-led. Other companies find themselves driven toward client-server by encountering the barriers inherent within their existing information systems. The most urgent problems typically concern network infrastructure. Existing technology often cannot gather or distribute the information that managers need to make timely, accurate decisions. It is also common for problems to exist in building the essential skills within IS for managing new systems development. In such cases, the creation of new network infrastructure and new skills becomes the priority in positioning an organization to benefit from client-server.

Application-led. Still other companies find that, although their existing technical skills and network infrastructure can support migration to client-server systems, their business processes require more rapid change. They may, for example, need to develop new products more quickly, or create cross-functional teams to bring a range of skills to bear on customers’ problems. For these firms, getting in place new solutions—and the organizational change that must accompany them—is the priority.

Parallel. Finally, a number of companies find themselves besieged on both fronts. They have to change business processes as well as technology—and change them fast. They may, for example, need to respond to sudden price competition or to a major new opportunity. Here, client-server can become a do-or-die proposition.

Each of these approaches strikes a different balance, presents different risks, offers different rewards, and implies different management priorities.

The opportunistic approach

The great risk of the opportunistic approach is that a company will continue its leisurely approach for too long

The great advantage of the opportunistic approach to client-server is that a company can initially address the task with leisurely deliberation. The great risk, however, is that the company will continue its leisurely approach for too long and let value-adding opportunities slip away. The situation of one regional bank illustrates this dilemma.

The bank in question relies on mainframe systems and has excellent skills in managing them. It has also succeeded in creating a few small but effective client-server systems. Its choice of what to do next, however, is finely poised between inertia and change. For any given systems requirement, it will be cheaper for the bank to extend its existing mainframe skills and technology. Yet managers fear that a rival will soon begin to use client-server technology to create innovative financial products that might lure away the bank’s core customers. Their challenge is to convert that vague worry into purposeful action before it comes true.

A useful tactic for senior executives in this kind of position is to examine the results of their own companies’ early experiments with client-server. If these first solutions have been enthusiastically welcomed by line managers and are delivering clear business value, the best step is to build on that enthusiasm and shift toward an application-led approach. Indeed, if there is little response to line managers’ enthusiasm for the business benefits that client-server can bring, the company’s IT strategy may quickly be placed at risk: line managers will go outside the firm to get the systems they want, with little regard to their "fit" with the overall direction of IT.

On the other hand, if initial experiments with client-server reveal weaknesses either in networks or in the company’s systems-building skills, then senior managers may want to use that evidence to build momentum for an infrastructure-led approach. Whichever the case, the opportunistic approach is a luxury best enjoyed briefly.

The application-led approach

An application-led approach is often the most straightforward. Companies in fast-moving service industries, for example, find it comes naturally—particularly when information technology is already deeply embedded in the products and services they offer to customers.

Driven by business change, the process often focuses on new initiatives to take advantage of new market opportunities. Because these opportunities are clearly defined in business terms, each project should provide a healthy return with no more than the usual level of risk associated with capital investment. Moreover, since the systems to be built are demanded by—not imposed on—their ultimate users, there should be few difficulties in gaining acceptance for new technology. As in any change process, a series of early, highly visible successes can quickly build momentum.

One nagging problem is the temptation to overspend on each initiative

There are, however, several dangers. One nagging problem is the temptation to overspend on each initiative. Because projects so readily justify themselves individually, it is easy to let duplication and excessive complexity slip into systems design.

Among the companies we studied, some estimated that they spent as much as 25 percent more than necessary in pursuing an applications-led approach. The only company that consistently avoided overspending did so by building, within its central IS function, the capability to discern potential economies of scale among the client-server projects proposed by individual business units. It also provided financial incentives for these business units to cooperate with central IS in tapping those scale economies.

The greatest danger of this approach is that inattention to infrastructure will suddenly create obstacles to progress

The greatest danger of this approach is that inattention to infrastructure will suddenly create obstacles to progress. Networks will no longer be able to cope with the flows of information demanded by burgeoning, enterprise-wide client-server applications. And managers who have grown used to the freewheeling ease of justifying projects on their own merits often find it difficult to accept the imposed discipline of building robust infrastructure. They will easily agree neither to limit their technical choices for the common good nor to help pay for shared facilities.

Rapidly increasing systems costs are a good sign that it is time to switch the emphasis of client-server migration toward infrastructure building. (At one bank, for example, costs were rising by more than 25 percent a year.) One possible solution is to institute a "federal" organization within IS, taxing business units to fund centrally-constructed infrastructure. Inevitably, however, organizational politics will tend to dominate the transition from an applications-led to an infrastructure-led approach to client-server systems.

The infrastructure-led approach

Some companies—particularly those in capital-intensive industries like oil, transport, and steel—are already governed by the discipline of centralized capital investment budgets and long planning cycles. For them, it may be quite natural to approach client-server by building the necessary network infrastructure first. For others, it may well be the sheer scale of the technical challenge that dictates an infrastructure-led approach. One consumer goods company, for example, determined that it could obtain significant business benefits from client-server. But with 3,000 widely dispersed sales representatives, it had to build a new network infrastructure before it could even begin to deliver the information required to tap that potential.

So long as infrastructure needs can be predicted with reasonable accuracy, building infrastructure before applications offers several advantages

So long as eventual infrastructure needs can be predicted with reasonable accuracy, a migration to client-server that builds infrastructure before applications offers several potential advantages. It allows a company to exploit latent economies of scale in network building. It also makes it possible to design a network that is technically elegant and resistant to obsolescence. In other words, it lowers the risk that early client-server migration will come to a sudden halt for lack of infrastructure.

A key problem, however, is that if managers miss the mark in their predictions, they will build an infrastructure that is inappropriate or unwanted. Two tactics can help to mitigate that risk. First, even as it focuses on building infrastructure, the IS department must involve and enthuse the managers who will conceive and demand the applications that will use that infrastructure. And second, a company should develop a portfolio of potential client-server applications. The success of any one may pay for the investment in infrastructure. Even in a less than ideal world, this portfolio approach can help limit the costs and risks of network infrastructure.

The parallel approach

The key to managing a parallel migration is to break the task into manageable pieces

Not surprisingly, doing things in parallel—that is, pursuing a migration approach that requires leading with infrastructure and applications together—combines the risks and benefits of both these approaches. But it also compounds the risks because of the sheer scale and complexity of the changes involved. The key to managing a parallel migration is to break the task into manageable pieces—a collection of smaller, self-financing projects, each of which can be justified in its own terms—and to place a premium on speedy implementation.¹ Although each project should adhere to an overall vision and to corporate-wide technology standards, it is impossible to achieve the technical elegance of an infrastructure-led approach amid the near-chaos of parallel change.

Ultimately, much will come down to the leadership skills of the senior executive leading the effort—usually the CIO. Many CIOs assume an essentially passive role, offering to build systems to support whatever individual businesses want to do. But strategically managing the opportunities created by client-server in the fast-moving world of parallel change requires a proactive CIO—fully supported by the CEO—who can bring an understanding of technological risks and opportunities to the discussions that will determine the shape of the business.

A new role for the IS department

Any move toward client-server will sooner or later require dramatic changes in how technology is managed

The parallel approach to client-server migration poses the greatest challenge to an IS department—and, indeed, to a business as a whole. Still, any move toward client-server will sooner or later require dramatic changes in how technology is managed. Not only will the IS department have to master new technologies, it will have to take on fundamentally new responsibilities. Ironically, the most dramatic changes are often the easiest to recognize and manage.

Client-server brings a real-time focus to information technology. Traditionally, IS departments managed major projects with multi-year paybacks. Investment priorities and budgets were set at annual meetings after in-depth study and careful preparation. With the possible exception of the initial construction of network infrastructure, most projects in a client-server world simply do not fit into this framework. Applications are short-lived, and are better thought of in expense terms than as a capital investment. Many will essentially be perishable. If it is to avoid becoming a constant drag on the business changes enabled by client-server, IS will have to speed up its own decision-making processes.

But that is not all. IS must also make different decisions in different ways. Client-server forces IS into a new balancing act. It must, for instance, work alongside—if not inside—business units to help build applications quickly, get information where it is needed, and support line managers’ efforts to get the most out of the technology. But it must also exert a central authority to maintain infrastructure, tap economies of scale in technology, and reduce the complexity of that technology.

A first step in getting this balance right is simply to recognize the conflicting demands. A second, however, is to acknowledge that, contrary to prevailing assumptions, many aspects of client-server’s distributed style of computing can be successfully centralized. End-user support, for example, is one area where companies can capture scale economies.

At a healthcare products company, the IS department and business units took three steps to identify the cost savings that could be derived by centralizing client-server support activities. They defined appropriate service levels and performance goals. They identified business-specific requirements. And they capitalized on opportunities to share resources by highlighting three levels of central administration and support. Taken together, these measures cut IT costs by more than 25 percent.

Many successful firms offer their business units a choice: either pay a high price for support of a unique technology, or enjoy savings on a standard product

Closely related to the centralization issue is the question of standardization. To control the complexity inherent in enterprise-wide client-server systems, successful IS departments work with users to standardize technologies, products, and vendors. Many successful firms offer their business units a choice: they can either pay a high price for support of a unique technology, or enjoy savings on a standard product. In this way, IS encourages each operating unit to determine the true business value of nonstandard or unique products. Where there is minimal loss of flexibility or major cost savings, the IS department may wish to standardize more extensively.

Although exercising central control comes naturally to an IS department accustomed to building systems single-handed, working closely with individual business units does not. Changing the ingrained attitudes and values of IS staff—not to mention boosting their interpersonal skills—is a long, hard process. Successful companies, however, have taken advantage of the overall organizational change that client-server brings to take steps in that direction.

At one investment bank, for example, the adoption of a client-server system led to a major overhaul of financial control systems. In the process, the boundaries dividing IS staff from accountants were blurred in several ways. Not only did the two begin to share offices, but IS staff were also given responsibility for some aspects of the execution of transactions. Similarly, some accountants were put in charge of some aspects of the new technology.

Despite the challenges, we expect the trend toward client-server computing to continue—indeed, to accelerate. Client-server makes full use of modern technology: fast microprocessors and high-speed data networks. It also helps to create modern organizations: fast-moving and quick to grasp new business opportunities. For many companies, only client-server offers the possibility of incorporating more information (and more information technology) into products and services while retaining the organizational ability to change as swiftly as markets do. Whatever route managers decide to take toward that goal, such capabilities are well worth the effort.