Information technology implementation in organizations has gone from automating back-office clerks to supporting the complex tasks of autonomous knowledge workers. The research I report here is not about new organizations or those transcending a deep crisis; rather, it concerns the push and pull of managers attempting to implement a new technology. It tells the story of companies trying to change the behavior of employed but fiercely independent “revenue producers” while, at the same time, trying not to drive away the best performers. The research investigates how different executives try to make the same people in the same roles work with a new tool.
My research was inspired by the notion that the work of salespeople (or lawyers or doctors) might be as radically transformed by technology as was the work of tinsmiths, hoopers, and portrait painters when technology impinged on their lives and livelihood. Traditionally, the knowledge worker has had more autonomy than the laborer, thus challenging the manager who is attempting to “automate” knowledge work. However, as technology infringes on the domain of symbolic, abstract work, the interaction between user and tool becomes more complex. Although automating knowledge tasks has proved a noxious process, organization after organization has tried to gain more influence over knowledge workers. The promise of productivity in this domain is a powerful force, drawing entrepreneur and bureaucrat alike into new, more comprehensive attempts.
The increasing sophistication of the type and degree of information technology has heightened this tension. More often, computer technologies are not passive but active tools that manage the process of work. A doctor’s notes on a patient, once a document of record, now are an interactive “protocol-driven data-capture device”—both supporting and constraining the doctor’s activities. The salesperson’s “pitchbook” is replaced by a multimedia offering, and the old order form is replaced by a configuration system based on laptop computers and their software. An aggressive form of software is expert systems, which specifically aim at codifying knowledge and creating a specific method to do a task. The challenge of getting people to use expert systems is an interesting example of the general problem of influencing knowledge workers’ behavior with computer-based tools.
Creating a model that will predict the successful implementation of any new technology is almost as challenging as creating a general-purpose thinking machine. Neither has come to fruition. Despite the vast literature on technology implementation and the universal recognition that we have entered a “postindustrial” economy, there are only a few examples of systematic research on the implementation of new tools by knowledge professionals in a field setting.1 There are singular case studies, of course, but with little cross-case analysis. Pentland’s extensive study of the Internal Revenue Service professionals’ use of laptop technology is one exception. Pentland studied laptop implementation in four districts, using interviews and a questionnaire survey of roughly 1,000 agents. He found radically different adoption; only two of the four districts had high usage.2 The high-use districts were characterized by high-quality training, rumors that laptop use would be mandatory, and supportive attitudes toward the central automation office. The low-use districts did not exhibit these characteristics.
Thus practicing managers who are concerned with implementation are left with scant evidence from rather distant examples that they must then extrapolate to their own situations. In this study, I try to add to the knowledge base by examining an expert system that helps field salespeople sell more insurance.
The Research Challenge
There is no theory that can predict the outcome of an implementation process. Markus and Robey, in an interesting synthesis of the implementation literature in the MIS field, characterize three categories of implementation: variance, process, and emergent.3 Variance theories assume an invariant relationship between independent and dependent variables. These factors often include top management support, an effective champion, and training. Other researchers have focused on implementation processes, i.e., how people adjust to the technology during its implementation.4 Some complex models incorporate both process and factor variables. Lucas et al. constructed a complex model of implementation with twenty-seven variables.5 In what Markus and Robey call the emergent school, there is a more phenomenological argument in which the implementation process emerges, as do the important variables for analysis. These occur as the implementation process unfolds. Kling and Iacono thought through the variety and richness of the phenomenon of interest in an even more elaborate model.6 Although informative, few if any of these theories have been tested for predictive validity.
Because there is no theory with predictive validity and because this research examines a new technology in a field setting, I designed a two-phase approach. In phase one, I did early case and field work to understand how early adopters used the technology and to qualify the most important independent and dependent variables. In phase two, I performed a cross-case analysis to investigate the relationships among these variables.
Phase One: Field Research
Knowledge-intensive tasks, by definition, cannot be generalized. Knowledge work is specific; indeed, the more knowledge intensive, the more specific it is. To fathom the complexities of an elaborate, abstract, and leading-edge tool such as a large-scale expert system in the context of the fluid, idiosyncratic process of insurance sales, I needed to understand the technology itself, the insurance industry, and the insurance sales process. Following the suggestions of Bonoma and Cook and Campbell, I created an initial, in-depth case study to understand the basic industry dynamics, the products, and the technology.7 I tried to understand the nature of the business, the competition, the products and services, the regulations, and the global insurance environment. Preliminary research into the use of the technology in a financial planning organization and early field research into the use of the Profiling expert system by potential and existing clients provided considerable insight. I also interviewed the technology developers and managers at the company that created the software, thus gaining intimate knowledge of the system and its purpose.8
Phase Two: Comparative Case Studies
Previous research has shown that environmental context can have a significant influence on the likelihood that an organization will adopt new technology.9 The nature of the industry and the evolving patterns of product and process design have an important relationship to the nature of innovation, especially in information technology.10 Company strategy shapes the nature of technology use as well.11 All these different contextual factors — control variables in the study —influence the adoption of new technology. Within individual firms, the organization structure, distribution structure, compensation systems, and product and service mix are important variables: training, job roles, and potential career paths all interact with the adoption of a new technology.12
I designed the study to follow the implementation of the same technology in similar but distinct organizations.13 More specifically, I chose companies in the insurance industry that had similar strategies and product and market emphases and were located in countries that had broadly similar regulatory and economic contexts.14 Each company sold a full line of life, disability, and insurance investment products (e.g., annuities and pension plans), and each had a direct salesforce that was the intended target user of the technology. Thus the control variables were company, product, regulatory environment, organizational structure, distribution method, compensation systems, and role of the salesperson. The technology was the Profiling expert system, which I will describe in detail later.
After reviewing the vast literature on diffusion, I focused on four independent variables: (1) the role of the technology sponsor, (2) the role of the champion, (3) the design of the implementation, and (4) the rationale for the radical change.15 (A technology sponsor is willing to support the technology’s implementation. The champion is the active participant and leader of the implementation effort.) All four independent variables have been identified as integral components of achieving “fit” between the project and the overall organization. In a number of large-scale research projects on the implementation of computing in local government settings, researchers at the University of California at Irvine have shown that management rhetoric on technology has been central to successful implementation.16 Walton, in a study of new IT in organizations, has noted the importance of how management frames the meaning of new technology.17 Orlikowski, in a study of adoption of radically different CASE software for software developers, found that the successful projects made a conscious rhetorical link between the current task and the strategic organizational goals.18
In this study, I revisited the fundamental relationships among these important variables in an “apples to apples” comparison.
Because Profiling is a batch system (i.e., it is centrally controlled, with Profiles coming in and out on a batch basis), it was possible to capture every single use of the technology.19 I also collected data on individual sales performance, demographic background statistics, organizational statistics, and geographic statistics.20 Next I describe the technology and the four companies and their similarities in order to identify the control variables. I then show the radical change that Profiling represented for selling insurance at all the organizations in the study. Finally, I look at the four implementation variables and their impact on Profiling use at the companies.
What Is Profiling?
In 1986, Applied Expert Systems of Cambridge, Massachusetts, created Profiling to perform comprehensive financial planning in cash management, risk management, income protection, general insurance (e.g., property or casualty), education funding, wills, credit management, investment planning, and retirement planning. Profiling also provides a networth restructuring statement to guide clients to a healthier financial balance sheet through asset diversification. Its primary market is couples, families, and individuals with incomes from $30,000 to $150,000, although the system’s logic can accommodate significantly higher incomes.
Profiling uses an extensive questionnaire, a personal financial profile (the client’s report), and an agent’s report. On the questionnaire, the client assembles financial and personal data, including future goals like buying a new house or putting children through college. Profile then makes generic recommendations. At the same time, Profile uses computed text to prepare a plan seemingly written to the client’s personal situation instead of “boilerplate” text with numbers inserted. The agent’s report matches specific company products to the recommendations. (See Tables 1 and 2 for samples of a client’s report and an agent’s report that one company adopted.)
In short, with the data on the questionnaire, the expert financial planning logic of the system (with more than 2,000 rules) presents a comprehensive financial picture and reveals the client’s insurance and investment needs. The client receives a document detailing the financial situation and pinpointing areas for improvement, plus a strategy for meeting future goals. The agent works from this document and the agent’s report.
Companies that implement Profiling hope that it will provide significant self-reinforcing benefits. Profiling can enable a clearer focus on the customer and create a total financial plan presenting the best products in an integrated manner. This, in turn, enables salespeople to sell more products on each sales call. Since customers with many products from the same company are much less likely to drop their coverage, Profiling can be important for retaining customers. The potential profitability implications of increased client retention are enormous.21 In addition, the better compensated agents who resulted from this process are less likely to leave, thus helping to decrease turnover. Achieving these high expectations and promises, of course, necessitates radical changes to how products are sold and how agents are trained at each site.
The Research Sites
- Lutheran Brotherhood, headquartered in Minneapolis, Minnesota, was established in 1917.22 A fraternal organization, the company sells only to Lutherans and their families, a nationwide population of approximately 14.5 million people. Owned and operated for the benefit of its members, Lutheran Brotherhood boasts an uninterrupted history of dividend payments on its life insurance. In 1987, it issued $3.5 billion in new business with $20.9 billion of insurance in force, placing it in the top 5 percent by size of all U.S. life insurance companies. In 1989, the company estimated assets under management at $7.5 billion and about 1 million members.
- National Mutual, established in 1869, is the second largest insurance company in Australia, with offices in New Zealand, Great Britain, the United States, and elsewhere, although most of its market, especially in personal insurance, is in Australia.23 With a premium income of $3.8 billion in 1990, National Mutual has an extensive field salesforce of 3,500 captive agents, 1,200 associate agents, and more than 100 specialty planners.
- Sun Alliance Insurance Group PLC is the oldest and largest personal insurer in Britain, with one of every five households holding one of its products.24 Its two major distribution companies — Sun Alliance Life and Sun Alliance UK — cover the personal insurance market, and property and casualty for companies, respectively. In 1990, it had more than ¥10 billion under management, about 8,000 employees, 2,000 company representatives, and more than 9 million policy holders.
- The Prudential Insurance Company was originally incorporated as the Widows and Orphans Friendly Society in 1873.25 The company has four core businesses: insurance and investment for individuals, and asset management and employee benefits for institutions. Prudential is the largest insurer in the United States for both individual life premiums and the amount of insurance in force, and is divided into four major geographical areas. I studied the western region.
· Similarities among Sites.
Each site has three critical functions: sales, underwriting, and investment. In life insurance companies, the underwriting, product design (on which the sales, underwriting, and investment areas collaborate), and investment activities are highly centralized to maximize control and concentrate expertise. (See Table 3 for similarities and differences among the companies.)
Each company fields a geographically diverse sales-force with a central home office. At three sites, the home office is in one location; due to its size, Prudential has five home offices, which report to corporate headquarters in New Jersey. The types of products, especially life insurance, that each firm offers are similar, as are agent relationships with parent organizations. However, the amount of independence grew as agents became more successful. Very successful agents and large agencies often arranged special relationships with the parent company.
Compensation systems are also similar in the four firms. Each provides a small stipend at the outset of an agent’s tenure, usually while she or he is in training, to encourage the individual to become self-supporting on commission income. All four firms are subject to government regulation. In the United Kingdom and Australia, regulation is at the national level. In the United States, the regulatory power is largely with the states, but the federal government heavily regulates the investment aspects. Finally, salespeople across the organizations have similar training, job responsibilities, and potential career paths.
The Context for Insurance Companies
During the late 1970s and 1980s, deregulation hit many of the world’s financial markets, including those in the United States, the United Kingdom, and Australia; managers, including those in the insurance industry, responded with many product innovations. At the same time, clients became more sophisticated as companies educated consumers about options and as various products became available. Hence, experienced and novice salespeople faced external pressures and an avalanche of internal demands on their knowledge, not to mention ongoing changes in tax laws and regulations. The insurance industry experienced what some termed “knowledge overload.” Sales-people often compensated by focusing carefully on a well-defined market and mastering a few products.
During this same period, the cost of face-to-face selling time rose, with the typical personal sales call reaching $150 or more.26 Competition from other distribution forms (e.g., direct mail) was also increasing, as were new distribution outlets like banks. The United Kingdom enacted legislation providing commission disclosure on life insurance products. Such factors, individually and together, pushed the managers of large life insurance companies to consider how their agents could provide more value-added services to justify the expense of a personal selling effort.
Adding to the cost pressure was the large field salesforce turnover, which, in this industry, ranges from 70 percent to 95 percent during a five-year period; high customer turnover often accompanies high agent turnover. Most life insurance contracts are heavily front-end loaded, with commissions paid to the agent early in the life of the contract; products typically do not become profitable until the third year, at which point commissions, underwriting costs, and initiation expenses are recovered in premium payments. Thus the profit implications of even small changes to product retention promise significant short- and long-term rewards.
The Selling Process
Based on previous research on the impacts of financial planning software in general, I judged that Profiling was likely to necessitate a radical change in the selling processes at all four companies.27 (See Figure 1 for the companies’ minimal sales process with and without Profiling.)
- Without Profiling. In the selling process with a single-need approach to a new client, the agent’s only required paperwork is to fill out the insurance application, which then typically goes to underwriting for pricing and approval. Underwriting, from a sales perspective, is frequently a source of unwanted delay; commissions are rarely paid at the time of application but are usually paid only on issued business.
- With Profiling. In the sample companies, using Profile lengthened the sales process in two ways. First, it necessitated at least one and probably two more sales calls on the customer. The agent needed to fill out the questionnaire; clients needed to provide information about previous life insurance policies, benefits packages from their jobs, banking and checking account information, and so on, to detail their financial situations. The Profiling system had only eleven required fields of data to make a plan, but the more complete the questionnaire, the more accurate the plan. Employees at the local sales office entered the questionnaire information into a computer. Every day, the data was sent by phone to a central facility, where the Profile was produced and mailed, along with the agent’s report, back to the agent.
The underwriting process, which came next, frequently caused a delay. Profile plans often recommended buying disability insurance because people are much more likely, in the short run, to become disabled rather than to die. However, life insurance salespeople are often loath to suggest disability coverage because the premiums are high and the underwriting is much more involved.
Because of its comprehensiveness, Profiling also required considerable product knowledge. In addition, it changed the access to client information and salesperson activity. Before Profiling, both new and established agents often had proprietary access to their own client files. Information for underwriting, product issuance, and billing purposes passed from field to home office, but the client’s needs, habits, and even financial situation stayed with the local agent. Established agents often “owned” the clients and successfully took their business as they moved from one firm to another. After Profiling was implemented, however, all the detailed client data were fed straight to the home office. Hence, the company adopting the system could track its salespeople at a higher level of detail by looking at their Profiling activity.
Analysis of the Implementation and Use of Radical Technology
In the second part of the study, I assessed the levels of technology use in the four companies and the impact of the four variables — the role of the champion, the role of the sponsor, the design of the implementation, and the rationale for the radical change. It is important to note that the technology had very different levels and patterns of use in these similar companies. From a research standpoint, I was fortunate to find such wide variance.
The successful organizations designed two very different implementations. Prudential completely changed the agent’s role and job design. It folded the technology into that redesign and altered task and technology together, an approach I call redesign. Sun Alliance did not change the nature of the task, but rather, through concerted championship on the part of senior and local line management, achieved adoption by constant emphasis and management effort, an approach I call focus. The common characteristics in both companies were an office-by-office rollout strategy and a significant degree of implementation momentum.
Interestingly, failure (defined as lack of adoption) had similar aspects in the two firms that did not succeed in implementing Profile; both relied on the caveat emptor approach. Lutheran Brotherhood and National Mutual both had thorough approaches to the implementation process. Both firms began with lead users across the organization and then progressively widened availability of the technology to other participants. Experts in training, technology, and financial planning led the effort. As rational as this approach was — and as scientific as the pilot design attempted to be — both companies failed to achieve adoption.
Overall, Sun Alliance and Prudential used a focused office-by-office approach, driven by senior line managers, to implement Profiling. National Mutual and Lutheran Brotherhood took the approach that the technology should be tested and made available on a wider and wider basis to the professionals. (See Table 4 for a comparison of the three implementation strategies.)
Different Levels of Use
At its zenith, National Mutual reached 642 Profiles in a single month, only to end Profiling completely in January 1992. Use at Lutheran Brotherhood peaked at 68 Profiles in September 1989 and steadily declined until the pilot was stopped in November 1989. In contrast, Sun Alliance had approximately 4,760 Profiles in November 1991, and Prudential, by January 1992, had 744 per month, with the numbers steadily increasing. (See Figure 2 for comparative use statistics.) At National Mutual, the number of agents who could use the technology was approximately 3,600; at Lutheran Brotherhood, 150; at Sun Alliance, 1,100; and at Prudential, 175. On an individual agent basis, the statistics are even more striking. Thus management at Sun Alliance considered that it had achieved a successful implementation. Prudential also achieved solid success. At Lutheran Brotherhood and National Mutual, on the other hand, there was little adoption; both ended up dropping the technology altogether. Why were the levels of use so different?
Each site had its own constellation of the four variables: the champion, the sponsor, the design, and the rationale for the change. First I examine the redesign implementation at Prudential.
The Profiling effort at Prudential was part of an overall strategy to redesign the selling process through the tied salesforce channel. (The agents were not independent but were tied to Prudential.) Senior sales management analyzed sales activity and discovered that a significant percentage of a salesperson’s time was spent on nonsales and nonservice activities. This caused the salesforce to be less productive than management desired. To make the direct sales channel more productive, John Tymochko, one of two senior managers running the company’s western region (which generates a significant percentage of the entire firm’s premium income), led an overall strategy to redesign the salesperson’s entire job, an initiative called Locally Deployed Agent (LDA).
The concept of the LDA was to give salespeople better administrative and technology support simultaneously.Prudential gave agents a budget for a full-time assistant to answer phones, make appointments, and perform other client servicing. The agents set up shop in storefront locations, where they literally hung out the Prudential shingle. The technology support had two major components: (1) the salesperson had a suite of office productivity tools such as call management, calendar software, and database management software to enable efficient marketing operations at the office and effective communications to the Prudential information systems; and (2) the internal Prudential systems were modified to make it easier to get product illustrations and client information from the company’s central computer systems. Profiling was designed to give both local and remote support. The local office conducted Profiling and then shipped the Profiles to the salesperson (just as the other firms in the study did). In return for this additional support, Prudential expected the agents to be twice as productive in their sales quotas.
Prudential’s pilot design called for an office-by-office rollout. The company rejected the idea of having a firm-wide or regionwide initiative. Instead, it focused on one region at a time and worked to get the new form deeply accepted before moving to a new location (see Table 5). Local managers at each location supported the implementation process; Tymochko had selected them because of their willingness to redefine the sales process.
Prudential’s CEO was the sponsor; the day-to-day champion was Tymochko, a line manager with significant responsibility for the sales region; the nature of the pilot was a focused redesign of the entire sales process, and the management rationale was to create an entirely new distribution mechanism — one that was more efficient than the traditional sales channel. This implementation strategy is “redesign” because several major factors in the salesperson’s job changed. The nature of day-to-day activity was more focused on selling and less on administration. The access to technology and its character also changed to support the sales process more directly. Profiling, as a full-service tool, matched the strategy of providing a more service-oriented and complete relationship with the Prudential client.
· Sun Alliance.
Sun Alliance created an extremely targeted implementation approach that aimed for significant adoption by building up support in each local office. This approach is “focused” implementation because: (1) management assumes that the performance of the task needs changing, but the fundamental task does not need to be redesigned; (2) the technology will facilitate the change and aid in both enabling and stabilizing the new state of the task; and (3) the manager’s role is to achieve sufficient use of the technology so a new pattern and new approach to selling is established.
At Sun Alliance, Paul Tebbutt, a senior line manager in sales, championed the Profiling implementation and was deeply involved in the overall training for its use. Although he was the marketing manager and, subsequently, sales director, Tebbutt was known throughout the organization as a high-volume, high-quality salesman. He continued to sell insurance, even in his management role. As one agent put it, “Paul knew how to communicate with the sales-people because he had lived it himself.” Tebbutt commented on the early challenges of implementation: “The first time through, we made a big mistake. We brought in one consultant from each branch and made him or her the Profiling expert. When these “experts” went back to the field, they did not have the support of their peers. They were alone. At the head office, we thought that the logic was right, therefore it should work. We did not have the champions of change in place.”
Sun Alliance’s focused approach succeeded because it introduced the technology as a way to help reinforce cultural change and transform tasks. For example, in the two-day training sessions for agents, one day was dedicated to Profiling and one to basic motivation and selling techniques, with Profiling as the key concept. Tebbutt led an effort to change Sun Alliance’s mission statement to read: “We will be the best financial Profiling company in the U.K.” Moreover, with the passage of the Financial Services Act (FSA) in the United Kingdom, senior managers wanted to ensure that their agents provided the best quality advice. Profiling adoption was not a “natural” consequence of the FSA, but management seized on such external pressure to support introduction of the radical technology into the selling process.28
I could not easily identify the project sponsorship at Sun Alliance. The top management group described the initial introduction of the technology, but in more than two dozen interviews with senior management, I could not find a single sponsor. While the managers saw Profiling as a group decision, its champion was clearly Tebbutt. The pilot design was focused; its purpose was to achieve penetrated adoption, not just trial. The management rationale was to completely change the nature of the relationship with the client from “policy sales” to “advice relationship.”
Sun, however, did not change its compensation structure or administrative support. The Profiling effort was integrated into the existing management structure and systems. There were no changes to the fundamental job design as at Prudential or any additional budget for administrative or clerical support for the field agents. Sun managers simply pushed this way of doing business onto the existing roles and structure.
· Other Explanations for Success.
A number of alternative hypotheses might explain the success at Sun and Prudential. Each organization’s strong culture may have helped to reinforce the implementation process. This cultural effect could be at the country level (e.g., the English insurance industry is more likely to adopt Profiling due to its history of heavy consumer focus or regulation). Culture could also be an important variable at the company level (e.g., the employees at Sun Alliance might be more willing to follow management directions). Although we cannot measure culture in a way that can rule out these alternative explanations, there are some other indicators. First, the two successful implementations were in different countries, which lends credence to the claim that this is not an issue of culture. Second, within Sun Alliance, there was a second sales organization, the executive agency division (EAD), which also tried to implement Profiling. (See Figure 3 for a comparison of the two Sun divisions.)
Paul Tebbutt directed SAPLIS, one of the Sun Alliance divisions that adopted Profiling. However, he did not direct EAD, where little adoption occurred. EAD chose an implementation strategy similar to caveat emptor. There was no line leader; a staff person at EAD was responsible for the implementation. Profiling was available to anyone who wanted to use it, but there was no office-by-office focus on the product’s rollout. The data from this division seems to support the idea that Profiling’s success was not due to a specific cultural issue within Sun Alliance or to the Financial Services Act; otherwise, there would have been more adoption at EAD.
Explaining the Failure to Implement
At both organizations where Profiling did not succeed, the CEOs were the sponsors. At National Mutual, George Meier, chief executive of National Mutual Group, sponsored the project. At Lutheran Brotherhood, President Bob Gandrud was the sponsor and paid for the project from special funds allocated to him for new initiatives. The head-office staff people championed the Profiling initiative; their main method of persuasion was a rational analysis that explained how the technology had helped other salespeople sell more. The champions’ role was to make the technology available to a geographically dispersed group of representative agents simultaneously. They did not attack the method of selling directly.
When Lutheran Brotherhood and National Mutual managers designed the pilot, they drew on an existing implementation method similar to the way they introduced new products. In the decision to launch Profiling like a new product, senior managers did not consciously examine the implications of instituting a radical technology as if it were an incremental technology. They simply thought that if the technology itself had value, agents would use it. In addition, both organizations drew on previous implementation efforts as a model.
Thus, when new products were to be rolled out to the field salesforce, Lutheran Brotherhood’s head office created the products, gave the agents introductory training, and encouraged them to use the products; this was the same approach they took with Profiling. Vicki Obenshain, the head office manager most involved with Profiling, had been in charge of the successful rollout of laptop computer support and the designer of a number of successful marketing initiatives.
At National Mutual, Neville Mears, the head-office manager in charge of introducing Profiling to the field sales agents, had more than twenty-five years with the firm and a track record of innovation in administrative support. Among his accomplishments was an innovation giving agents better access to the client information in the firm’s various databases, an effort that earned him the managing director’s Award for Excellence. The primary support group for the Profiling integration was a central head-office group. “Product consultants,” who constantly pushed out or “sold” new products to the field, provided distribution.
· Caveat Emptor Implementation.
To understand the caveat emptor approach, we need to understand the nature of the implementation approach, the management rationale, and the championship. For instance, in the implementation of Profiling at Lutheran Brotherhood and National Mutual, the pilot design of the technology had significant, interrelated implications for subsequent management actions. The field salesforce’s perception of the role of technology was particularly important because the company viewed agents as “independent” businesspeople. That independence was reflected in the organization’s compensation systems and policies; for example, agents had to buy their own software and hardware. While they had a small stipend for office support, any additional support came from their own funds. In addition, project championship was embedded in the pilot design decision, so the champions were head-office staff people, not sales managers who were credible to the field sales agents and managers. Further, implementing Profiling as a new product rather than a new process further emphasized the agents’ discretionary adoption. That is, the agents were not obliged to sell all the products available or to purchase all the sales technologies they were offered.
A number of connected factors constitute caveat emptor implementation. First, the individual (the agent) is the best judge of whether to adopt. Second, a small number of trials give the agent ample information to make that judgment. Third, the manager’s role as implementer is to make the technology available for the agent to judge.
Lutheran Brotherhood and National Mutual used this approach to introduce new products and most other tools to their agents and implemented Profiling this way. Specifically, both companies constructed the pilots to ascertain a representative sample of agents. Lutheran Brotherhood chose six offices reflecting a wide range of markets to measure whether the technology would work in the broad market that the firm served. It also used two agencies as a control for overall economic factors. At one point in the pilot design, Obenshain and the vice president of sales discussed the choice of pilot offices, debating whether the pilot should concentrate only on “friendly” offices, i.e., where the local managers would actively push the technology. They determined, however, that this would overly bias the sample toward positive results, so they chose geographically dispersed offices in representative markets, with only a few agents experimenting with Profiling in each.
National Mutual employed a similar logic. For its pilot, it chose thirty-six agents from each of the seven states in the country. Internal analysis of pilot results suggested that the average sale increased from $1,400 to $2,900 (in Australian dollars), adding impetus to the idea that the technology would be efficient and that a laissez-faire implementation strategy would work.29 Thus the extended pilot also strove to achieve the same singular penetration in a number of geographically dispersed offices; that is, one agent in each office was designated the Profiling agent. The company did not train or encourage other agents in that branch to use the tool.
This philosophy extended to National Mutual’s “limited deployment” phase, in which it designated one Profiling agent in each of 150 branch offices. Given National Mutual’s motivation and previous history of rolling out products to a wide selection of field agents, the caveat emptor strategy had mechanisms inherited from a management rationale created for product introductions.
Interestingly, by early 1990, when adoption volumes were lower than expected, the comments of National Mutual managers suggested that they were committed to transforming the sales process. The head of the agency remarked at that time: “We won’t know for four more years if Profiling is a success or not. I won’t entertain a review of the Profiling decision for at least three years. We are going to do Profiling because it is central to our strategy. Cultural change takes years, and we are going to have a go at it.30
Unfortunately, after another year of low adoption volumes, management support for the project waned, and, with the deepening of the worldwide recession, it ended in January 1992. Thus, even though the rhetoric of change and transformation was present, there was little direct line-management involvement in the implementation.
Implications for Managing the Implementation of Technology
The implementation of Profiling provided a delightfully complex challenge to managers. On the one hand, the potential gains to the organization from implementing this new method of work were tremendous. The strategic implications for profitability, customer satisfaction, and agent retention were potentially lucrative; moreover, if well implemented, the organization could achieve significantly better information about its customers and their use of products. If well executed, a successful implementation of Profiling might give the adopting organization a significant competitive advantage. On the other hand, management faced the task of getting autonomous knowledge workers to change their behavior. Knowledge workers already have much more power and autonomy than clerical workers. By providing current revenue to the organization, they enhance their power base even more. Implementing a new technology tool at the heart of their task process is difficult.
· A Constellation of Actions.
Faced with these challenges, the successful managers in this study took different approaches. Paul Tebbutt of Sun Alliance used strong leadership and a focused implementation approach. At Sun, managers simply pushed the technology into the organization, a sobering reminder of the power of impatient, credible line management. Prudential looked more comprehensively at the core assumptions and design of the task, started with a clean slate, and redesigned.
At a quick glance, these findings are not particularly striking. We have always known that leadership matters. What is interesting is that radical technological change may necessitate a constellation of actions by the manager to achieve success. To claim that there is only one solution is an overstatement of current understanding. Such hubris is potentially more damaging if it anchors the manager’s planning and thinking in the incorrect assumption that there is one best way to implement a new technology. My data indicate that a better starting point is to acknowledge, a priori, that there may be a number of (or at least more than one) equally effective means to achieve the desired end state. More specifically, the research indicates that redesign and focus implementation strategies may substitute for each other. Very different processes may have equivalent outcomes. To begin with a frame of mind and analytic approach that precludes this isomorphism is to fundamentally misframe the task and hamper a manager’s options.
The job of a manager is to think broadly about options and perhaps search for possible approaches, which may be surprisingly different. From this research, it seems that the strategizing must be done early in the implementation process, because once the first decisions are made, the implementation process takes on a life of its own, and the core design variables may be difficult or impossible to change. Just as the dominant design of manufacturing products and processes constrains the alternatives from supply through distribution, so does the path of new technology implementation rarely return to its roots and core assumptions.
A striking finding of the study is that the senior managers rarely thought of the different options as isolated activities but rather in patterns or constellations. Practically speaking, this means that while the senior managers were figuring out what to do, they were also figuring out who should do it, what to expect, and how to do it.31
· Political Ramifications.
Managers were also trying to assess the nature of the potential political ramifications of implementing a new technology. Each managerial lever for change had significant constraints and meanings that were predominantly local. For example, events concerning technology can influence potential users’ perceptions of the next technology, which often vary from firm to firm. Thus Profiling and the accompanying total-needs analysis embedded in its approach resembled Lutheran Brotherhood’s earlier effort, “Financial Dimensions,” which the head office brought to the field during the mid-1980s. Financial Dimensions was a selling approach emphasizing the customer’s total needs, complete with a sales pitch and training system. Employees saw it as a failure. During the implementation of Profiling, some agents reported seeing it as another Financial Dimensions and immediately predicted failure. At Prudential, technology per se did not seem to have this negative reception.
The nature of the supporting organization also has an impact on implementation. For example, any selling system that the head office brings to the field has a very different flavor than one that field sales agents introduce. Behind the bland phrase “I’m from Corporate and I’m here to help you” lies the strong implication that the field organization is pitted against the head office, especially when administrative innovations are involved. Consequently, who the “owner” is becomes tremendously important because it has nothing to do with the technology itself but everything to do with its reception. At Lutheran Brotherhood and National Mutual, the prime movers in the implementation process were from the head office; they were not senior line managers in sales. This too affected adoption of the technology. Even though there were sponsors at the top of the organization and the day-to-day champions were capable managers, the target audience for the technology did not identify with them. The “owner” of the innovation was a factor loaded into the value of the technology before it was even tried. The more successful managers seemed better at sensing these complexities and assembling the right constellation of people and resources. The less successful managers allowed a scattered approach and support that was not credible to the field personnel.
By the end of implementation at National Mutual, Profiling had become a symbol similar to Lutheran Brotherhood’s Financial Dimensions. National Mutual’s program manager, installed in late 1991, wanted to ensure that none of his efforts were labeled “son of Profiling.” Other National Mutual managers used anti-Profiling positioning. One began a field salesforce meeting by announcing, “The first thing I want to say is that this idea has nothing to do with Profiling.”
In addition to their strong constellation of choices, the senior managers at Sun and Prudential consciously managed the momentum of the project. Tyre has discussed the windows of opportunity for changing manufacturing methods.32 In implementations with radical organizational implications, the issue of momentum takes on even more significance because training and local management support has a half-life, and the novelty of a new tool soon wanes.
In this study, the two examples of success each exhibited a particular momentum. This variable has been implicitly dealt with in many implementation studies but rarely measured as an explicit management concern. While adoption rates have been carefully tracked and managed in consumer product consumption, project momentum is rarely if ever measured in research on software implementation for knowledge professionals.33 But, in the design of the two successful efforts, managers’ concern was to create a sequence of events that would build momentum for implementation.
Many organizational theorists argue that radical change is best effected quickly. But if a manager faces a powerful cadre of knowledge workers, the swiftness and decisiveness of the implementation process can be constrained by the need to keep the implementees happy and employed. This implies a need to act swiftly and make a concerted effort to manage momentum over time. The more radical the software, the more important it is to manage organizational momentum.
· Economies of Scale.
What would cause a need for organizational momentum? What would make the caveat emptor process fail so miserably in the case of a tool that would help the end users? Why do the focus and redesign approaches work better? There may be economies of scale and scope in sales knowledge. If minimum efficient scale for a new method is not reached, it will fail after early experimentation. In the case of a complex sales tool, people’s knowledge must be created and shared. If only one person uses such software as Profiling, the intellectual cost of “inventing” all this local knowledge might be too much to manage efficiently. This factor places a heavy load on sales personnel who are the sole users of the technology. By contrast, staying on top of the learning and the changing nature of a complex tool is easier if a substantial majority uses the tool.
Managers who intuitively understand these economies of scale may see their roles as champions to be less like scientists and more like change agents for rapid achievement of economies of scale in local knowledge. Implementing radical software seems to need a significant amount of organizational and managerial effort to cause adoption. This also works in concert with a manager’s concern to manage momentum. The more momentum, the faster you get to minimum efficient scale in the knowledge work.
The economies of scale may be even more challenging to achieve if they have delayed benefits. In many interviews, sales agents and managers reported that people took significant time becoming comfortable with Profiling. The process also delayed the payment of commissions, even though the commissions were expected to be higher. If the radical software technologies have delayed benefits, focus or redesign are important for sustaining or creating the knowledge necessary to use the tool practically. As managers move from automating the clerical tasks to supporting knowledge work, the issue of creating and managing economies of scale in knowledge work will become central to a firm’s value-creating activities. When introducing technologies that are incremental and not radical, the concern about minimum efficient knowledge scale may not be as great. But minimum efficient scale of knowledge surely arises when radical software is involved.
There are a number of tasks for researchers. First, if the idea of implementation isomorphism is correct, they need to construct models that can accommodate non-obvious, isomorphic substitutions. One constellation of variables may have the functional equivalency of another set, but the nature of these trade-offs may not be immediately apparent. Traditional, single-form models do not identify these factors. Larger samples do not solve this specification problem, which may explain why there has been so little progress in creating implementation theory with any predictive validity.
Second, researchers need to create tools to measure momentum. In the change process, the speed and surety with which change moves forward has tremendous impact. This issue is especially acute if significant learning is needed to use the new software, because lack of momentum can turn into delays. Delay provides time to forget, and forgetting itself slows momentum.
Senior management must understand that a single solution is not the only solution. There is probably no optimal way to implement a technology. Successful implementation of radically new software in autonomous work necessitates significant motivation — perhaps beyond short-term individual economic benefit. Senior managers are personally responsible either to participate in the implementation process or to change the nature of the organization so that the new technology can be absorbed. The natural propensity when implementing technology to support autonomous professionals is to adopt a caveat emptor strategy, because it implicitly relies on independent judgment for the technology’s acceptance and subsequent use. But radical software technologies also need concentrated action. Projects have life and momentum.34 The managers of a project must assume a strategy of “diffuse or die.” If the project does not grow, it will probably end.
1. For the pioneering work of Fritz Machlup in categorizing and articulating the nature of the knowledge economy, see:
F. Machlup, The Branches of Learning (Princeton, New Jersey: Princeton University Press, 1982).
Daniel Bell’s work set the path for many to follow. See:
D Bell, The Coming of Post-Industrial Society: A Venture in Social Forecasting (New York: Basic Books, 1976).
2. B.T. Pentland, personal conversation.
3. M.L. Markus and D. Robey, “Information Technology and Organizational Change: Causal Structure in Theory and Research,” Management Science, volume 34, May 1988, pp. 583–598.
4. See M.L. Markus, and M. Keil, “If We Build It, They Will Come: Designing Information Systems That People Want to Use,” Sloan Management Review, volume 35, Summer 1994, pp. 11–25.
5. H.C. Lucas, M.J. Ginzberg, and R.L. Schultz, Information Systems Implementation: Testing a Structural Model (Norwood, New Jersey: Ablex Publishing Corporation, 1990).
6. R. Kling and S. Iacono, “The Control of Information Systems Development after Implementation,” Communications of the ACM, volume 27, 1984, pp. 1218–1226.
7. T.V. Bonoma, “Case Research in Marketing: Opportunities, Problems, and a Process,” Journal of Marketing Research, volume 22, May 1985, p. 199–208; and
T.D. Cook and D.T. Campbell, “Validity,” in Quasi-Experimentation (Chicago: Rand-McNally, 1977), pp. 37–94.
8. See J. Sviokla, “PlanPower: The Financial Planning Expert System” (Boston: Harvard Business School, Case 186–293, 1986); and
J. Sviokla, “Expert Systems and Their Impact on the Firm: The Effects of PlanPower Use on the Information Processing Capacity of the Financial Collaborative,” Journal of Management Information Systems, volume 6, Winter 1989–1990, pp. 65–84.
9. J.E. Etillie, “A Note on the Relationship between Managerial Change Sector Firms,” R&D Management, volume 13, October 1983, pp. 231–244.
10. See J. McKenney, Waves of Change: Business Evolution through Information Technology (Boston: Harvard Business School Press, 1995).
11. See L. Applegate, J. McKenney, and F.W. McFarlan, Corporate Information Systems Management (Homewood, Illinois: Irwin Publishers, 1996).
12. For a comprehensive model of implementation factors, see:
Lucas et al. (1990).
13. In a seminal book on case research, Yin suggests that when trying to isolate specific causative variables, researchers should choose sites with control variables as similar as possible and independent variables with as much variance as possible, which he terms “case replication logic”; each case is a “test.” See:
R.K. Yin, Case Study Research: Design and Methods (Newbury Park, California: Sage Publications, 1989).
14. For example, I had the opportunity to include a Japanese site in the study, but I refrained from including this site because the distribution of life insurance in Japan is much different from that in England, Australia, or the United States. In Japan, the tradition of door-to-door selling of life insurance — with frequent collection of premiums (mostly by women) still is a thriving distribution mechanism. I also had the opportunity to, but consciously avoided looking at, this same technology in the context of a direct sales insurance provider or in a large financial services firm that also rolled out the technology. I wanted to keep the basic industry, organizational, and individual context as comparable as possible.
15. See D. Leonard-Barton, “Implementation as Mutual Adaptation of Technology and Organization,” Research Policy, volume 17, October 1988, pp. 251–267;
J.F. Rockart, “The Line Takes the Leadership — IS Management in a Wired Society,” Sloan Management Review, volume 29, Summer 1988, pp. 57–64; and
E. von Hippel, The Sources of Innovation (New York: Oxford University Press, 1988).
16. J.L. King, V. Gurbarani, K. Kraemer, F.W. McFarlan et al., “Institutional Factors in Information Technology Innovation,” Information Systems Research, volume 5, June 1994, pp. 139–169.
17. R. Walton, Up and Running: Integrating Information Technology and the Organization (Boston: Harvard Business School Press, 1989).
18. W. Orlikowski, “CASE Tools as Organizational Change: Investigating Incremental and Radical Changes in Systems Development,” MIS Quarterly, volume 17, September 1993, pp. 309–340.
19. I encountered a number of significant methodological challenges when trying to assess the introduction of Profiling technology and its subsequent use and effect on sales. Perceived usage and end-user satisfaction, two often-used measures, have been severely criticized for their lack of reliability. See:
N.P. Melone, “Theoretical Assessment of the User-Satisfaction Construct in Information Systems Research,” Management Science, volume 36, January 1990, pp. 76–91.
My study utilizes actual use statistics gathered from the information systems, down to each individual transaction as the measure of use.
Many studies stop at satisfaction or have only a tenuous link to performance; this study, by contrast, uses product sales and commission data as sources of actual performance of individuals: the organizational metrics are the aggregation of the individual measurers.
20. I conducted structured interviews with more than 100 salespeople, sales managers, and senior managers both before and after Profiling’s implementation on the selling approach, the types of technology in current use, and Profiling’s fit with current methods. I compiled a year-by-year comparative history of each firm’s environmental situation and tracked the major internal organizational changes and business initiatives. In addition, I created a technology time line that documented the entire implementation effort at each organization in terms of important meetings, dates, and personnel changes. I constructed a comparative time line across each organization. And I regularly met with senior managers to review impressions and verify findings.
21. See F.F. Reichheld and W.E. Sasser, “Zero Defections: Quality Comes to Services,” Harvard Business Review, volume 68, September–October 1990, pp. 105–112. This article articulates the significant financial implications of keeping retailing customers. See also:
J. Sviokla and B. Shapiro, Keeping Customers (Boston: Harvard Business School Press, 1994).
22. J. Sviokla, “Lutheran Brotherhood and the FSNAR+ Pilot” (Boston: Harvard Business School, Case 190–163, 1990).
23. J. Sviokla, “Profiling at National Mutual (A), (B), and (C)” (Boston: Harvard Business School, Cases 191-078, 191-101, and 191-102, 1991).
24. J. Sviokla, “Sun Alliance Insurance Group, PLC” (Boston: Harvard Business School, Case 192-073, 1992).
25. J. Sviokla, “Client Profiling: The Prudential Insurance Company of America” (Boston: Harvard Business School, Case 193-084, 1993).
26. Estimate by senior sales managers at Lutheran Brotherhood.
27. J. Sviokla, “Managing a Transformational Technology: A Field Study of the Introduction of Profiling” (Boston: Harvard Business School, Working Paper 93-059, 1993).
28. In 1987, the passage of the Financial Services Act in Great Britain required any insurance agent to show “knowledge of the customer” and “best advice.” For Sun Alliance, this meant filling out a fact-finding questionnaire for the client to sign or not. (The client also had the option of not sharing data; that exercised right would be noted on the form.) Most insurance organizations dealt with the regulation by having individual agents fill out a generic questionnaire that was then filed at the local sales office.
29. An Australian dollar equals approximately US$ 0.80.
30. Sviokla, “Profiling at National Mutual (B),” p. 2.
31. Jeanette Lawrence documents how judges consider the goals, means, and feasibility of decision options simultaneously. This is similar to the behavior we saw in the managers concerning implementation. See:
J. Lawrence, “Expertise on the Bench: Modeling Magistrates Judicial Decision Making,” in The Nature of Expertise, ed. M. Chi, R. Glaser, and M. Farr (Hillsdale, New Jersey: Lawrence Erlbaum Associates, Inc., 1988), pp. 229–259.
32. M.J. Tyre and W. Orlikowski, “Windows of Opportunity: Temporal Patterns of Technological Adaptation in Organizations,” Organization Science, volume 5, February 1994, pp. 98–118.
33. For an interesting exception and specific discussion of the mechanisms of momentum, see:
G. Moore, Crossing the Chasm: Marketing and Selling Technology Products (New York: Harper Business, 1991).
34. R.E. Walton, “The Diffusion of New Work Structures: Explaining Why Success Didn’t Take,” Organizational Dynamics, Winter 1975, pp. 3–22.