An empirical study on the drivers of management control
systems' design in new product development
Tony Davila*
IESE, University of Navarra, Avenida Pearson, 21, Barcelona 08034, Spain
Abstract
New product development has changed signi®cantly over the last decade and management control systems have played an important role in this transformation. This study draws on Galbraith's concept of uncertainty and investi-gates the relationship between project uncertainty, product strategy and management control systems. It also explores whether these systems help or, as argued in the innovation literature, hinder product development performance. Results support the relevance of the project uncertainty and product strategy to explain the design of management control systems. They also show that better cost and design information has a positive association with performance, but that time information has a negative eect.#2000 Elsevier Science Ltd. All rights reserved.
1. Introduction
New product development has become a central dimension in the strategies of many companies (Brown & Eisenhardt, 1995; Clark & Fujimoto, 1991, p. 6; Grant, 1996; Gupta & Wilemon, 1990; Schilling & Hill, 1998). Current emphasis on ®rst mover advantages, fast product introductions, more demanding product functionality, and shortening life cycles has put greater pressure on new product development (Cooper, 1998). While manufacturing has traditionally been a key repo-sitory of core competencies (Hayes & Abernathy, 1980), outperforming competitors in product development has emerged as a relevant source of competitive advantage.
As the process has gained importance, aca-demics as well as practitioners have voiced the importance that management control systems play in coordinating and controlling this process (Cooper & Kleinschmidt, 1987; Zirger & Maidique, 1990). For example, Clark and Fujimoto (1991), in their study of the product development process in the auto industry, argue that:
Today's eective product development organ-ization is characterized not only by creativity and freedom, but also by discipline and con-trol in scheduling, resource use, and product quality (...) The challenge in product devel-opment is not so much unilateral pursuit of organic structure and permissive management style as a subtle balance of control and free-dom, precision and ¯exibility, individualism and teamwork (Clark & Fujimoto, p. 169).
0361-3682/00/$ - see front matter#2000 Elsevier Science Ltd. All rights reserved. P I I : S 0 3 6 1 - 3 6 8 2 ( 9 9 ) 0 0 0 3 4 - 3
www.elsevier.com/locate/aos
However, this emphasis on a structured product development process contrasts with the traditional view supporting a hands-o approach (Lothian, 1984; McNair & Leibfried, 1992). According to this latter view, successful new products result from devoting adequate resources to the process and avoiding control procedures that could restrict the freedom of engineers. The impact of management control systems in product develop-ment performance is, therefore, unclear.
So far, management accounting literature has devoted scant attention to new product develop-ment. Most studies have looked at the relevance of management control systems to the broader pro-cess of R&D (Abernethy & Brownell, 1997; Birnberg, 1988; Brownell, 1985; Hayes, 1977; Kamm, 1980; Rockness & Shields, 1984, 1988). These studies mainly characterize management control systems as hindering or, at most, being irrelevant in R&D settings. In contrast, Nixon (1998) oers a rich case description of a product development process where ®nancial control plays a signi®cant role.
The importance of new product development requires the allocation of accounting research resources in order to understand the phenomenon. This study seeks to extend this line of inquiry. Using a contingency approach, the study investi-gates the design of management control systems1
to understand how companies adapt their sys-tems to the particular characteristics of each pro-duct development eort. Moreover, the study brings new evidence to the unsettled issue of the relevance or, alternatively, the lack of relevance of management control systems in product development.
Several characteristics distinguish this study. In contrast to previous research, the unit of analysis is the product development project rather than the R&D project. Because R&D projects are very
heterogeneous (National Science Foundation, 1976), focusing on one type of project increases the power of the research design. The study also goes beyond the narrow de®nition of management control systems around ®nancial information to add formal but non-®nancial information (Kaplan, 1983; Banker, Potter & Schroeder, 1993). Moreover, the theoretical foundation of the study leads to an interpretation of management control systems dierent from previous studies and to a dierent set of independent variables.
The study focuses on the medical devices indus-try to keep the external factors as constant as possible and avoid confounding eects that may come from dierences across industries. This industry has several attractive characteristics. First, product development is an important pro-cess: R&D over sales averages more than 5% for the industry and new products are constantly introduced. Therefore, companies have well thought-out product development processes. Second, the industry is characterized by a lot of technological diversity. Some products Ð syringes, for example Ð use well-established technology, while others Ð CT systems, for example Ð com-pete by bringing to the market the latest technol-ogy developed. Finally, product strategies are also diverse; even products belonging to the same company and serving the same product-market have to adapt their value proposition to dierent market segments ranging from price sensitive to performance oriented customers. X-ray products include machines designed to take static images of parts of the body, where price is the key purchas-ing criteria, as well as sophisticated machines that scan the whole body from dierent angles, where performance and customer interfaces are the key competitive dimensions.2 Both diversity in
tech-nology and product strategies suggest that com-panies manage product development dierently.
The remainder of this paper is structured as fol-lows. The next section reviews previous research
1 The term management control systems is used to name the
design as well as the use of measurement systems in an organi-zation. Therefore, leaving out other formal procedures that the organization may use to alter behaviour (Flamholtz, 1983; p. 154). An alternative term is management accounting systems. However, management accounting systems are sometimes interpreted as conveying ®nancial information only, while this paper also investigates non-®nancial measures.
2 The companies in the study include a wide range of medical
on the design and role of management control systems in R&D and presents the theoretical fra-mework underlying the study. Section 3 describes the phenomenon studied through the description of four representative cases. These cases illustrate the variables in the research as well as the hypotheses of the study. Section 4 develops the hypotheses for the empirical test based on the theory as well as case ®ndings. Section 5 describes the research design for the survey study. Section 6 discusses the results of the paper and Section 7 reaches conclusions.
2. Theory development
2.1. The product development process
The objective of product development is to translate an idea into a tangible physical asset. The process is structured around well-de®ned phases; each phase ends with a decision-making meeting where management decides about the future of the project. A typical product develop-ment project starts with a planning phase to establish the requirements of the project. During this phase, the organization de®nes the target market and the characteristics of the product. These characteristics include functionality, price, performance, and expected release time. The out-come of the initial phase is a broad description of these characteristics. The second phase Ð concept design Ð goes into more detail to specify the pro-duct speci®cations and the requirements of the development project: target costs, technological performance, customer interfaces, market release dates, and organizational resources. The third phase Ð product design Ð is the actual develop-ment of the physical product. It is in this phase when trade-os get resolved and information is transformed into a tangible product. The last two phases Ð testing and production start up Ð con-®rm that the product meets its objectives and pre-pare it for release. The process, even though described as linear, is an iterative process: product speci®cations or even the product concept can be re-evaluated in light of new information generated throughout the process.
2.2. Literature review
Past work on management control systems in R&D follows two approaches. One line of research focuses on how R&D departments use ®nancial measures (Brownell, 1985; Hayes, 1977; Rockness & Shields, 1988). The consensus from these studies reveals that ®nancial measures do not have an important role in R&D departments other than signaling the commitment of the organiza-tion to its R&D eorts. The perceived importance of budgets ``decreases monotonically from plan-ning to monitoring, monitoring to evaluating, and evaluating to rewarding'' (Rockness & Shields, p. 571).
Another line of research adopts a broader view of control systems (Abernethy & Brownell, 1997; Kamm, 1980; Rockness & Shields, 1984). For example, Kamm de®nes control as ``the set of criteria, policies and procedures established to standardize operations and to make possible mea-surement of performance to insure achievement of organizational objectives'' (p. I-12, I-13). Rock-ness and Shields (1984) study the relationship between types of control and project character-istics. Following Ouchi's framework3 (Ouchi,
1979), they classify R&D projects according to the level of knowledge of the transformation process and the measurability of the output. Next, they predict a relationship between these characteristics and the type of control used: input, behavior, and output control. These authors ®nd only marginal relationships between control systems and project characteristics. Similarly, Kamm (1980) concludes that ``researchers do not necessarily exhibit more innovative behavior when they perceive relatively loose administrative control than when they per-ceive tight control'' (p. IV±11). Abernethy and Brownell (1997), use Perrow's model (Perrow, 1970) that relates type of control with task analyzability and number of exceptions. These authors conclude that ``reliance on accounting controls has signi®cant positive eects on perfor-mance only where task uncertainty is lowest'' while ``behavior controls appear to contribute to
performance in no situation'' (p. 245).4 This
evi-dence suggests that management control systems have, at most, a minor role in product development. Nixon (1998) provides a rich case description on the balancing role of the controller in assisting engineers during the development of a new copper rod production machine. In contrast to previous studies, the author reports that the ``®nancial component of the system serves to integrate the disparate perspectives'' (p. 343).
However, management control systems have proven to be useful tools in environments char-acterized by high levels of uncertainty. For example, Khandwalla (1972) ®nds that reliance on formal control systems increases with the intensity of competition. Similarly, Simons (1987) reports that high performing prospectors rely on the information provided by frequently updated for-mal control systems to drive organizational learn-ing (Dent, 1990). Additional research shows that managers who perceive a higher level of environ-mental uncertainty tend to use broad scope and more timely information (Chenhall & Morris, 1986) as well as more external, non-®nancial, and ex-ante information (Gordon & Narayanan, 1984). Kren (1992) ®nds that participation in the budgeting process is related to better performance for high uncertainty tasks. Finally, Chenhall and Lang®eld-Smith (1998) report that dierent stra-tegic priorities emphasize dierent formal control systems regardless of the uncertainty faced by the organization.
A possible explanation for the apparent contra-diction between the results for R&D environ-ments, where management control systems seem not to be relevant, and other environments is a dierent interpretation of management control systems. R&D studies interpret these systems as control tools to reduce goal divergence rather than as information tools to deal with uncertainty. These ®ndings are in line with the concept of clan control (Ouchi, 1979). Clan control emphasizes informal control mechanisms and relies less on
formal systems. When uncertainty is high, clan control is preferred to solve goal congruence problems (Merchant, 1982).5
In line with the alternative interpretation of management control systems as tools to manage uncertainty, studies on target costing all concur on the role of target costing procedures as commu-nication, problem solving, and learning devices (Cooper, 1995; Kato, Boer & Chow, 1995; Koga, 1998; Sakurai, 1989; Tani, 1995). Koga and Davila (1998) ®nd that target costing ful®lls an information role to facilitate learning and experi-mentation, yet they ®nd no support for target costing being used to address goal divergence problems or coordination issues.
2.3. Theoretical framework
2.3.1. Management control systems and the concept of uncertainty
Product development is an uncertain process. For example, Gupta and Wilemon (1990) report that technological uncertainty is mentioned as a reason for delays by 58% of project managers surveyed. Each new product development process presents a dierent set of problems and organiza-tions need information to solve uncertainties as they emerge. The theoretical background of the paper is based on the concept of uncertainty as ``the dierence between the amount of informa-tion required to perform a task and the amount of information already possessed by the organiza-tion'' (Galbraith, 1973, p. 5). This paper, in con-trast to previous work in the ®eld, assumes that the main role of management control systems in product development is to supply information required to reduce uncertainty rather than to reduce goal divergence problems. This alternative perspective is intended to reconcile the tension that exists between the sparse empirical evidence available with the strong recommendations by practitioners and academics in the product devel-opment ®eld. The concept of management control
4 Accounting control is similar to Rockness and Shields'
(1988) ®nancial measures, while behavior control is related to the level of formalization of the organizational structure.
5 The dual role of management control systems is common
systems used in this study,6 following Chenhall
and Morris (1986) and Gordon and Narayanan (1984), goes beyond the narrow perspective focused around accounting numbers Ð cost, pro®tability, and budget Ð to include a broader information set (Kaplan, 1983) capturing cus-tomer, product design, and time-related measures. Management control systems in new product development are viewed as sources of information that are used to close the gap between ``the infor-mation required to perform a task and the amount of information already possessed''. This view is consistent with Tushman and Nadler (1978) who argue that management control systems are eec-tive tools to manage uncertainty because they supply the data needed to reduce Galbraith's ``information gap''.
However, management control systems are not necessarily the optimal sources when the informa-tion that they deliver is not matched to the uncer-tainty facing the product development manager. The relevant information may be obtained from alternative sources. For instance, it may be obtained through experimentation (Pisano, 1994) or informal communication (Allen, 1977; Dou-gerthy, 1990); if this is the case, then management control systems may not have any role in the pro-cess and, consequently, not be related to project uncertainty.
Research in new product development (McGrath, 1995; Shenhar & Dvir, 1996; Von Hippel, 1988; Wheelwright & Clark, 1992) has identi®ed three main types of uncertainty (or ``information gaps'' according to Galbraith's de®-nition): market-related uncertainty, technology-related uncertainty, and project scope. These three types of uncertainty shape the design of manage-ment control systems. In addition to the uncer-tainty characterizing the project, the design of management control systems depends on the strategy (Govindarajan & Gupta, 1985) as well as the organizational structure (Bruns & Waterhouse,
1975). Cooper (1995) reports that companies place dierent emphasis on target costing procedures depending on product strategy. Certainly, the value of a piece of information (for instance, cost information) is contingent upon the importance as well as the uncertainty related to the competitive dimension addressed (cost leadership). Similarly, organizational structure aects the size of the project team that is associated with the level of formalization (Mintzberg, 1979, pp. 230±235) and the project manager's responsibilities that aect the allocation of uncertainty. For instance, if the marketing department is responsible for dealing with market uncertainty, then the project manager will be insulated from it and he will not demand customer-related information, even if it may be critical to the success of the project.
2.3.2. Management control systems and project performance
The eect of management control systems upon new product development performance is dicult to predict. If management control systems supply information relevant for coordination and learn-ing, then a positive relationship between perfor-mance and the use of management control systems is expected. Some evidence in the product devel-opment ®eld exists pointing in this direction (Koga & Davila, 1998, Nixon, 1998). But argu-ments as well as evidence (Eisenhardt & Tabrizi, 1995) exist suggesting that such a relationship does not exist or is negative. Management control systems, by imposing rules and constraining behavior, reduce the level of creativity required from product development and, thus, negatively aect performance (Amabile, 1998).
3. Case studies
To understand how project managers use man-agement control systems, I visited 12 business units in seven companies both in Europe and the United States. During each of these visits, I inter-viewed one or two project managers, the market-ing manager, the R&D manager, and the general manager for each business unit as well as the per-son in charge of the design and implementation of
6 Simons (1995, p. 5) de®nes management control systems
the new product development process guidelines. Because existing literature in management control systems in product development is still sparse, I chose to do an exploratory study using case studies as the preferred methodology to build knowledge about the phenomenon (Yin, 1988). Interviews were structured around a set of ques-tions about the formal systems and the product development process itself. The questions were open-ended which allowed me to adapt the inter-view to the expertise of each manager without losing the overall direction. Appendix A presents the protocol that I used for the interviews with project managers. Similar protocols were used for the interviews with other managers.
I interviewed an average of ®ve managers in each business. The use of multiple informers allowed for a triangulation of the data. When a manager's explanation did not agree with the description given by previous managers of the same organization, the dierences were explored until the reason for this divergence was fully understood.
Next, I present four illustrative cases on the diversity of product development projects and the design and use of management control systems.
3.1. Project manager A
Project manager A worked in an anesthesia monitoring system. This product was designed to work together with the company's anesthesia delivery system. The company's strategy was ``to work very close to the customer, in that sense we are not a low cost producer but we focus very much on customer needs and facilitate customer interface with the product. We want to be special in the sense of adapting to the needs of the custo-mer and understanding the custocusto-mer well''.
At the beginning of the project, the manager signed a three page contract with eight goals: schedule (phases and review points), quality, usability, manufacturing cost, project budget, simple description of intended functionality, and contact points with other projects (the anesthesia delivery system). The purpose of this contract was not to evaluate performance ex-post, but to gain the personal commitment of each person involved
in the project. The contract brought together the expectations of the various people involved in the project rather than establish goals to increase extrinsic motivation.
Project goals were clearly de®ned except for product speci®cations related to the customer interface. The product's strategic advantage came from meeting customer needs and developing the appropriate customer interface. The ``information gap'' to be closed during the product design phase came from the market, in particular from customer needs.
Because of the relevance of customer informa-tion, management built ¯exibility into project goals to incorporate this information during the execution instead of freezing it at the beginning of the project. The decision to sketch only certain product speci®cations at the start of the project was intended to adapt as much as possible to cus-tomer feedback: ``there is a need to expose the product and product concept to the customer and be ready to change and adapt features and appearance to their reactions.'' Uncertainty was purposely left unresolved on the customer dimen-sion to adapt during the development process, but it was clearly bounded: ``there is a need to de®ne ¯exibility dimensions up-front (and freeze other dimensions)''.
During the execution, the project was divided into smaller sub-projects including ``moving from the traditional two measures captured in a tradi-tional anesthesia monitoring system to several measures, developing the frame to integrate the various modules of the product, and writing the product's software''. The project also had mar-keting sub-projects like ``the product launch pro-ject including training distributors, promotion material and marketing concept communication''. The project manager directly supervised engineers and marketing people. He was also frequently in touch with manufacturing people to prepare pro-duction ramp-up.
3.2. Project manager B
Project manager B had developed a new clip used for brain surgery. Medical doctors used these clips to keep blood vessels closed while performing brain surgery. Existing clips were metallic. This material had the advantage of providing the right mechanical properties like torsion and resistance as well as being cost eective. But metal had a signi®cant drawback for certain types of surgery. When the doctor, while performing surgery, had to do a scan of the patient's brain, the clip created shades in the picture and, more important, mag-netic ®elds could move the clip with possibly devastating consequences for the patient. In doing some tests on his own, project manager B found that a new material, titanium, could solve this problem. Titanium was more expensive but it would become the only product available to per-form scans during surgery. The company esti-mated a signi®cant market for the product and funded the project.
According to the project manager the main question during the project was to get the mechanical properties right: ``in this product, technology was critical''. Technology was the main source of uncertainty as well as the key fac-tor for product success. He did not care about the cost of the product, because it would have a vir-tual monopoly in its segment: doctors did not have alternative products and competitors were unlikely to develop the required mechanical know-how to copy the clip in the short term.
At the beginning of the project, the project manager talked to doctors and was present in several surgeries to see how the clip was used. These visits allowed him to understand customer needs. In addition, the project was not subject to time pressure because no other company was investing in a similar product. Only when the technology was well understood, did the company decide on a deadline. During the 4 years that the project lasted, all the attention of the project manager was focused on ®nding the right combi-nation of materials and the appropriate design to meet the mechanical requirements: ``because it was intense in technology, it was hard to see problems and it was also hard to calculate timing''.
Because technology was the paramount variable in this project, project manager B worked together with a team made up of researchers. Only a mar-keting manager was supporting the team to facilitate contact with doctors. The project plan was simple, the timing for the various phases of the product were loosely speci®ed as was the budget and the expected product cost. The fact that the CEO had come from the R&D function and kept close con-tact with product development people reinforced an informal control on the project. Through monthly meetings, the CEO evaluated whether the project was moving according to expectations without the help of a formal project plan.
The management control system for project manager B was almost non-existent. He got all the relevant information from prototyping ``to assure manufacturability''. He built more than two thousand prototypes before he found the right mix of materials and design. Any other information, like timing or cost, was irrelevant to him. The new clip was a success when it hit the market.
3.3. Project manager C
Project manager C worked for the same com-pany as project manager B. He was in charge of developing a hip endoprothesis for an Asian country. The product was similar to an existing one, but the marketing department had found that the body geometry of people in the main ethnic group of the country was dierent. The company saw this fact as a relevant dimension for competi-tive advantage. The project took a year to develop. Because the product was similar to an existing one, few doubts existed regarding product cost and technology: ``we knew a lot about the structure of the development of this product''.
to compare project costs and budget, it was impor-tant to reduce as much as possible the number of prototypes to save development money. There was a trade-o between safety and investment''.
The main task for project manager C was to coordinate the eort of engineers to meet the tight schedule. The source of uncertainty came mainly from project scope. The management control system provided detailed information on how the project progressed in terms of schedule and budget.
The project manager did not have direct contact with the customer. In fact, his supervisor talked to the marketing people in the Asian country and the marketing people talked to doctors. The project manager did not see this lack of direct access to the customer as a problem because the product was well understood, the only relevant issue being a change in geometry. Moreover, the contacts with the marketing people were mostly related to pro-duct launch, not to customer needs.
Finally, product costs were also well under-stood. However, the manufacturing person involved in the development team periodically estimated product cost to make sure that it was on target: ``the project would have stopped only if manufacturing costs had been too high''.
Management control systems in this project were focused around time-to-market and project budget. The latter information was re¯ected in the project manager's decisions concerning whether to build a new prototype. Product costs, even if critical to project success, were managed by exception.
3.4. Project manager D
Project manager D developed a multi-purpose X-ray machine. The product had two critical components, the X-ray camera and the examina-tion table for the patient. The technology for the X-ray machine was well understood and devel-oped in-house. But, the table was a complex mechanical device. Because the machine allowed an X-ray picture to be taken of any part of the body, the table was large and, as a consequence, hard to develop. In addition, the doctor could choose the angle for the picture that (s)he con-sidered most appropriate. This capability meant that the table had to move at least 180 degrees in
each of the three spatial axes with a high degree of precision. The main source of uncertainty for this product came from mechanical technology.
The X-ray division had recently reassessed its strategy after several years of disappointing ®nan-cial performance. According to the marketing manager: ``We are stripping down the number of products because now there are too many and it is expensive to deliver and service such an extensive line of products. We are not satisfying customers per se, we are also looking at pro®tability. The current product line is based only on satisfying customer needs and this is why there is so much proliferation of products''. This new emphasis on cost aected project manager D, even if technol-ogy was the key source of uncertainty.
Product development was a linear process at the division. It started in the marketing department with product de®nition, then customer require-ments were translated into system speci®cations, system speci®cations into component speci®ca-tions, then components were integrated at the sys-tem integration phase, and ®nally the product was launched. The role of the project manager was limited in this division to the supervision of com-ponent development. His main task was to break down the project into small work packages fully speci®ed in terms of budget, time, component speci®cations, and component cost and make sure that plans were met. In the terms of Wheelwright and Clark (1992), he was a ``lightweight'' project manager with no people reporting directly to him, but only coordinating the development eort. The project manager mentioned: ``I never talk to cus-tomers, they talk to the marketing people but not to me''.
supplier, provide support to their people, and try to minimize the eect of these problems in project scheduling and cost. He did not care as much about budget because ``development cost is similar to time because it is basically time multiplied by price''. Even if the product achieved its objectives, the project was not considered a success because of delays and budget overruns.
3.5. Discussion of case studies
The previous cases provide a diverse set of pro-duct development experiences and dierent roles for management control systems. Each project manager required dierent information depending on product strategy and type of uncertainty. For project A, meeting customer needs was the key success factor as well as the main source of uncer-tainty. Management purposely left customer-related uncertainties to be resolved during the development process through close contact with the customer. The structure of management con-trol systems emphasized customer interaction. Time, budget, and product cost were managed by exception. Because the project never hit these constraints, the project manager devoted his attention to customer information. The project team integrated both engineers and marketing people with a looser coordination with the manu-facturing function. This structure re¯ected the management belief that the project manager should be in charge of marketing.
In contrast, project B was all technology. Time was not a constraint, nor was budget nor product cost. In fact, the formal systems were loose com-pared to the detailed project plans and review points used in the other projects. The project manager focused his attention on prototyping as the most ecient way of coping with technological uncertainty. Project B exempli®es a situation where detailed formal management control sys-tems could undermine performance. Prototyping gave project manager B the information that he needed Ð any other source of information would have been a burden and undermined performance. His team was composed of R&D people only and he reported to the CEO who had a background in R&D.
Project C illustrates the development process most similar to a manufacturing process where uncertainty resides in coordination Ð project scope. The cause±eect relationships were well understood and product functionality was well de®ned. Project manager's attention was mainly devoted to time-to-market and budget. He did not interact with customers, nor did he devote much attention to costs (controlled by exception), but he was constantly thinking what needed to be done to meet the deadline and assessing whether he could save development costs by reducing prototyping. It is interesting to observe how project manager C used a non-®nancial measure Ð number of proto-types Ð as a substitute for a ®nancial measure Ð project investment. Again, this project manager was in charge of an R&D team. Interestingly, his contacts with marketing were not related to customer needs but to product launch because of its importance to the strategy of the product.
Finally, project manager D worked at a com-pany where costs had become a key dimension because product proliferation had led the com-pany into disappointing ®nancial performance. This emphasis was translated into frequent cost estimations. Unfortunately, the main source of uncertainty for project manager D came from technology. The design of a key part of the pro-duct was subcontracted out and ran into prob-lems. Project manager D had to devote most of his attention to this unexpected issue that aected the timing, functionality, and budget of the pro-ject. In this case, management control systems informed the project manager about technology only by exception even if it may have required more frequent updating. Project manager D did not have a team reporting to him, he only coordi-nated the technical part of the project. Table 1 summarizes these ®ndings.
4. Development of the research hypotheses
4.1. Uncertainty and the design of management control systems
design and use of management control systems. Case discussions illustrate how the sources of uncertainty vary across projects. Also, product development literature and management account-ing literature identify dierent types of project uncertainty. To be as close as possible to the phe-nomenon studied, I rely on the classi®cation of uncertainty used in the product development lit-erature. Uncertainty is classi®ed as market uncer-tainty, technological unceruncer-tainty, and project scope (Shenhar & Dvir, 1996).7
Von Hippel (1988, chapter 2) describes the importance of the organization's experience with the targeted customer segment. When the
organi-zation already serves the target customers, their needs and requirements are well understood and uncertainty is low. In contrast, when the organiza-tion enters a new market, uncertainty surrounding
Table 1
Summary of the case studies
Project manager A were clearly de®ned except for customer interface.
Technology-related uncertainty
The project manager built more than 2,000 prototypes.
Project scope Pressure came from coordinating eorts to meet the expected market introduction date.
Product strategy Customer-focused strategy ``We focus very much on customer needs and because they had already started to sell the product.''
Low cost strategy ``Product costs are estimated every time new parts become available.'' systems were designed to focus management main vehicle to reduce uncertainty.
Information purpose Management control systems used constantly to monitor schedule and by exception for cost and budget.
Information purpose Management control systems used by exception to detect potential problems.
Performance The alignment between project uncertainty, customer-focused strategy and management control systems' design led to a successful project.
The low emphasis on time, cost, or customer information allowed project manager to focus on experimentation and allowed attention to be focussed on time-to-authority led to poor performance re¯ected in problems with an OEM supplier.
7 A parallelism can be established between both
customer preferences increases (market uncer-tainty). In the latter case, information about cus-tomers is expected to help in reducing market uncertainty.
H1a:8 Customer information is used more
intensively as market uncertainty increases. The sources of product technology can range from existing, well-known bodies of knowledge (low uncertainty) to unknown and yet-to-be developed technologies (high uncertainty) (McGrath, 1995; Shenhar & Dvir, 1996; Wheel-wright & Clark, 1992). When technology is the main source of uncertainty, project team members focus their attention on resolving the problems associated with technology. Product design and functionality information can help in addressing this type of uncertainty. However, case study B suggests that project managers may obtain the relevant information from experimentation and prototyping (Clark & Fujimoto, 1991; Pisano, 1994), and then the relationship between technol-ogy uncertainty and the use of management con-trol systems is non-existent or even negative.
H1b: Management control systems are used less intensively as technological uncertainty increases.
Finally, project scope is related to eort that the project manager has to devote to coordinating the input from dierent constituencies. Project scope depends on the number of people involved in the project. A small project, possibly because the pro-duct is simple or because it only involves a small group of engineers, will have low demands on formal systems for coordination. In contrast, a large project with ®fty people dispersed in several departments around the company will need to rely much more on formal systems for coordination (Mintzberg, 1979).
The coordination eort will also depend on the project manager's responsibility. For example, project manager A was responsible for customer interaction as well as technology development, while project manager B only supervised R&D people. There is ample evidence on the relation-ship between organizational structure and the
design of management control systems (Baiman, Larker & Rajan, 1995; Bruns & Waterhouse, 1975; Merchant, 1981). Therefore, the empirical tests need to control for the organizational structure.
H1c: Management control systems are used more intensively as project scope increases.
4.2. Product strategy and the design of management control systems
The relationship between strategy and manage-ment control systems' design has been well docu-mented at the business strategy level (Govindarajan & Fisher, 1990; Kaplan & Norton, 1996; Lang-®eld-Smith, 1997; Merchant, 1985; Simons, 1987). The ®ndings of these studies are robust in terms of the typology of strategy used. Simons (1987) uses the strategy types de®ned by Miles and Snow (1978); Merchant (1985) follows the typology sug-gested by MacMillan (1982); while Govindarajan and Fisher (1990) rely on Porter's (1980) concept of competitive strategy. If these results are gen-eralized to product development, then it is expec-ted that product strategies will be relaexpec-ted to management control systems' design. However, this relationship is only a conjecture empty of any empirical evidence. Even if cost may be critical to the success of a product competing on price, meeting initial speci®cations may satisfy this objective and the project manager can safely ignore cost information. The typology of product strategies selected for the research is based on Miller and Roth (1994) who identify price, time-to-market, and customer focus as dierent product strategies.9If management control systems provide
useful information to deal with relevant project uncertainties, then project managers designing low-price products will value product cost infor-mation more highly, while time inforinfor-mation may be more valuable for products that would stand to bene®t from ®rst mover advantages. The following hypotheses capture these arguments:
8 Hypotheses are stated in positive terms for clarity, but the
no-hypotheses are tested.
9 Technology-based strategy is sometimes included as an
H2a: Cost information will be used more inten-sively as the importance of a low cost product strategy increases.
H2b: Time information will be used more inten-sively as the importance of a time-to-market pro-duct strategy increases.
H2c: Customer information will be used more intensively as the importance of a customer focused product strategy increases.
4.3. Management control systems and project performance
The aim of most managerial activities is to improve the performance of the organization. Therefore, it is relevant to know whether manage-ment control systems aect project performance.
Notice, however, that the absence of a relation-ship between management control systems and performance does not necessarily mean that these systems are irrelevant. An alternative interpreta-tion is that companies have optimally designed systems. If all companies have precisely the man-agement control systems that they require, then performance will not be related to these systems. In contrast, if such a relationship exists, then it can be concluded that management control sys-tems are related to project performance and that some companies are not using optimal systems.
The relationship between management control systems and project performance will be positive if projects bene®t from more structured systems. On the other hand, if systems are too structured and sti¯e the ability of the development team to respond to demands particular to the project, then the relationship will be negative.
Moreover, the relationship between manage-ment control systems and project performance may be contingent upon certain project character-istics.10In particular, strategy has been frequently
identi®ed as aecting the design of management control systems (Govindarajan & Gupta, 1985; Lang®eld-Smith, 1997). The following hypotheses capture the main eect (H3a) as well as contingent relationships (H3b,H3c,H3d).
H3a: More intense use of management control systems has a positive eect on project performance.
H3b: More intense use of customer information has a positive eect upon performance for pro-ducts following a customer-focused strategy.
H3c: More intense use of cost information has a positive eect upon performance for products following a low cost strategy.
H3d: More intense use of time information has a positive eect upon performance for products following a time strategy.
Finally, the detail reported at the beginning of the product design phase may also aect project performance. However, existing evidence is con-tradictory. Eisenhardt and Tabrizi (1995) ®nd that the amount of planning has no eect upon devel-opment time. In contrast, Gupta and Wilemon (1990) report that the ®rst reason for product delays is a poor de®nition of product requirements (71% of the respondents). A more general argu-ment supporting the importance of planning is provided by Bruns and McKinnon (1992) who found a positive association between clear goals and improved performance. The last hypothesis captures these arguments and relates them to pro-duct development.11
H3e: Detailed project objectives are associated with improved performance.
5. Research and survey design
Management control systems in product devel-opment vary over time and across the organiza-tion's hierarchy. They vary over time because information needs are dierent for the planning, concept design, product design, and testing and start up phases. Similarly, management control systems span the whole organization, from the formal systems used by top management, to the routines that shape the work of a recently hired engineer. This variation in the research setting can
10 I thank one of the referees for pointing out this interesting
extension.
11 Similarly to the discussion for project performance, the
decrease signi®cantly the power of the research design. To increase as much as possible this power, the research design includes three speci®c decisions. First, the study focuses on the product design phase only. Because this phase requires more structured information to evaluate trade-os, the relationships predicted by the theory will be especially strong in this phase. Also, the start and the end of this phase are clearly de®ned, thereby eliminating ambiguity about which data are required from project managers. Moreover, focusing on one phase reduces the noise that would result from asking for and interpreting data related to multiple phases.
The second research design decision is to specify the hierarchical level inside the organization. The project manager is the person in charge of moving a product development project from an idea to a physical object and thus, (s)he is the key person for the success of the project. This person is selec-ted as the unit of analysis.12
The third design choice is to limit the study to the medical devices industry. Eleven companies participated in the second part of the study. A contact person in each company selected a group of projects as heterogeneous as possible in terms of size and product strategies. The data were col-lected using a questionnaire mailed to project managers that had recently ®nished the develop-ment of a new product.
The questionnaire was designed to collect as much quantitative data as possible to avoid per-ceptual biases. However, recall bias Ð possibly driven by ex-post rationalization Ð could be a threat to the integrity of the data.
The response rate was 77% (56 out of 73 mailed questionnaires). This high response rate was accomplished by following several procedures (Dillman, 1983). The questionnaire was initially pre-tested among academics with previous experi-ence in questionnaire design. Some of the items were shifted to facilitate answering the questions, and to avoid, as much as possible, respondents
rationalizing their behavior. Then, a group of ten project managers tested the questionnaire. Two of these managers had the questionnaire administered in person. In the other cases, man-agers completed the questionnaire and commented on it in a telephone conversation. Only minor wording changes were necessary after the second pre-test.
Each questionnaire was personally addressed to the project manager. The package included a cover letter, the questionnaire, a pre-paid envel-ope, and a copy of an article for practitioners that could be of interest to the respondents as a token of appreciation for their eort Ð completing the questionnaire took 35 to 45 min. The letter oered a copy of the aggregate results of the study to the companies as well as to each respondent. The support of the contact people in the companies was also a very important element in achieving the high response rate.
5.1. Dependent variables
Preliminary interviews with product development managers identi®ed the six types of information most frequently reported through the organiza-tions' formal systems: product cost, product design, time-related, customer-related, resource input (budgets), and pro®tability. The design of management control systems for each of the six types of information is measured through three characteristics (Merchant, 1981; Simons, 1995):
1. Level of detail in the information reported is measured on a ®ve-point scale with three anchor points exemplifying measures ranging from low to medium and high detail. For example, cost information has low detail if the system only reports material and labor costs, and it has high detail when, in addition, the systems include related manufacturing, marketing, and administrative costs. Simi-larly, customer information has low detail if it comes only from an initial assessment of the marketing department, and it has high detail when, in addition, the project team interacts directly with the customer. Appen-dix B reproduces the anchor points used.
12 Results not reported in this paper show that the project
2. Frequency of information updating is mea-sured, for each type of information, on a ®ve-point scale ranging from (1) weekly or less, (2) twice a month, (3) monthly, (4) quarterly, and (5) longer than quarterly. 3. Usage pattern of information is measured on
a ®ve-point scale anchored with two sentences: ``the information was used to monitor the project, but it was not discussed with my team except when it reported events that fell below plans or expectations'' (diagnostic system) and ``the information was used con-stantly in the interactions with my team. Frequently it was the main topic of our con-versation'' (interactive system).
These three characteristics have the same pur-pose: providing information in order to reduce uncertainty. Therefore, they represent the same underlying latent variable. This variable is identi-®ed using principal component factor analysis as described in Table 2.13 Table 2 also describes the
variance explained, the ®rst eigenvalue, as well as the Cronbach alpha measure of reliability (Cron-bach, 1951). The inter-item reliability estimates meet Nunnally's (1967) standards for exploratory research.
Project performance (Perf) is a multidimensional variable (Shenhar, Dvir & Levy, 1997) and the importance of each dimension changes across projects. Meeting cost objectives can be critical for certain products while secondary for others. Moreover, the success of a product may not be correctly assessed for a long time after its market introduction. Financial success is not a good measure of performance (Cooper & Kleinschmidt, 1987): consider companies entering new markets Ð their early products are intended to facilitate learning rather than to make big pro®ts.
A set of questions developed by Shenhar and Dvir (1996) were adapted to measure project per-formance. The instrument includes eleven items
that capture dierent aspects of product develop-ment (see Appendix C). The respondent rates the importance of each item on a seven-point scale from ``not important'' to ``extremely important''. (S)he also rates performance for each item on a seven-point scale from ``extremely poor'' to ``extre-mely good''. The overall measure of performance (Perf) is the weighted average of these 11 items.
The drawback of using a self-reported measure is that it may be aected by perceptual biases. On the other hand, it has the advantage that it cap-tures the dimensions most relevant to the project and takes into account expectations for the pro-ject. For example, a delay of one month in intro-ducing a new product is bad for time-sensitive projects, but it is not important for projects focused on other dimensions.
5.2. Independent variables
To measure product strategy, principal compo-nent factor analysis with varimax rotation is used on nine questionnaire items intended to measure these variables. One set of items asks the respondent to allocate 100 points among dierent possible strategies. The other six items require respondents to rate the importance for the company and for the customer of each strategy in a seven-point scale ranging from ``not important'' to ``extremely important''. Three factors are identi®ed re¯ecting three possible strategies: cost-related questions load onto the ®rst factor, this factor identi®es the importance of cost strategy; the second factor re¯ects time strategy; and the third factor represents the importance of customer strategy (see Table 3).
Project uncertainty includes three variables: market uncertainty, technological uncertainty and project scope. Market uncertainty (Mkt-X) and technologi-cal uncertainty (Tech-X) are multidimensional con-cepts, constructed both as dummy variables. When the project is below the median in each of the questions that de®ne Mkt-X (Tech-X), the variable takes a value of zero, it takes a value of one if one of the questions is above the median, and so on (see Appendix D for a description of the questionnaire items) (see Table 4 for descriptive statistics).
The number of people involved in the project (People) and the number of new parts in the
13 Both, principal factor analysis and maximum likelihood
product (New-Parts) represent project scope. However, New-Parts may also re¯ect technologi-cal uncertainty if it is argued that products with more parts are also technologically more complex. I use three variables to control for organiza-tional structure. The ®rst one is the level of
cross-functional integration (Function) that exists in the project team. This variable is measured by the number of functions reporting to the project manager. The eect of this variable on perfor-mance has strong support in the product develop-ment literature (Clark & Fujimoto, 1991; Zirger &
Table 2
Principal factor analysis for the construction of management control systems' variablesa
Variable Name of variable
Items in questionnaire Loading on ®rst factor
Variance explained
Eigenvalue Cronbach alpha
Customer information CUSTI .Detail of customer info. 0.70
.Updating frequency of customer info. (*) 0.74 0.55 1.655 0.60 .Interactive use of customer info. 0.74
Product design DESI .Detail of product info. 0.83
information .Updating frequency of product info. (*) 0.78 0.57 1.721 0.61 .Interactive use of product info. 0.62
Time information TIMEI .Detail of schedule info. 0.65
.Updating frequency of schedule info. (*) 0.82 0.61 1.823 0.67 .Interactive use of schedule info. 0.76
Cost information COSTI .Detail of cost info. 0.86
.Updating frequency of cost info. (*) 0.76 0.62 1.859 0.68 .Interactive use of cost info. 0.62
Resources information BUDI .Detail of resources info. 0.66
.Updating frequency of resources info. (*) 0.68 0.58 1.747 0.64 .Interactive use of resources info. 0.83
Pro®tability information PROFI .Detail of pro®t info. 0.81
.Updating frequency of pro®t info. (*) 0.85 0.68 2.043 0.76 .Interactive use of pro®t info. 0.72
a Loadings based on the principal factor analysis, this speci®cation is more robust to the underlying properties of the probability
distribution of the variables. One factor is retained for each construct. Loading signs for questions that are worded in reverse (denoted by * in the table) have been changed. In all cases the value of the second eigenvalue is less than 1.
Table 3
Factor analysis on independent variables
Items in questionnaire First factor (Cost-Str) Second factor (Time-Str) Third factor (Cust-Str) Uniqueness
Design a low cost product 0.74 ÿ0.16 ÿ0.28 0.35
Meet unit cost objectives 0.79 0.04 0.06 0.38
Target customers value price 0.82 0.02 0.12 0.31
Reduce time to market ÿ0.11 0.74 ÿ0.27 0.36
Meet timing goals ÿ0.03 0.81 ÿ0.05 0.35
Target customers value time 0.06 0.74 0.20 0.40 Design a customer friendly product ÿ0.17 ÿ0.25 0.64 0.50
Ful®ll customer needs ÿ0.10 0.22 0.77 0.34
Target customers value ease of use 0.17 ÿ0.16 0.81 0.29
Eigenvalue 2.11 1.97 1.65
Maidique, 1990). The second variable represents the hierarchical level of the project manager's superior (Hierarchy). This variable takes values from one to four depending on the position of the superior (see Appendix E). This variable is rele-vant because arguably managers with a higher hierarchical position may be busier and thus dele-gate more decision making to the project manager. The last variable is the authority of the project manager over marketing decisions (Mkt-Dec) (0:78). The questionnaire items used to
mea-sure marketing authority are adapted from
Keat-ing (1997), the respondent evaluates his authority over a set of decisions on a seven-point scale ran-ging from ``I (or my team) took action without consulting other people in the company'' to ``other people in the company decided what action to take, my opinion was not solicited, but the deci-sion was explained to me''.
The detail of project objectives (Plan) is mea-sured as a weighted average. First, the respondent is asked to evaluate the ``level of detail in the pro-ject plan prepared before the start of the design phase'' for each of the six types of information on
Table 4
Descriptive statistics on variables and related questionnaire itemsa
Theoretical Actual
Min Max Min Max Mean Std. Dev.
Project performance 1 7 3.3 6.3 4.67 0.72
Management control systems design
Updating of customer related information 1 5 1 5 3.16 1.15 Updating of product design information 1 5 1 5 2.00 1.12 Updating of product schedule information 1 5 1 4 1.98 0.93 Updating of product cost information 1 5 1 5 3.73 1.03 Updating of product resources information 1 5 1 5 2.85 1.19 Updating of pro®tability information 1 5 1 5 4.17 0.85
Product strategy
Low cost strategy (%) 0% 100% 0% 50% 16.1% 14.1% Time-based strategy (%) 0% 100% 0% 75% 23.3% 16.2% Customer focused strategy (%) 0% 100% 0% 70% 32.7% 17.6%
Project uncertainty
Technology uncertainty 0 3 0 3 1.2 0.80
Market uncertainty 0 3 0 3 1.2 0.93
Percentage of new parts 0% 100% 10% 100% 56.1% 27.5% Number of people in the project 0 1 0 106 16.8 21.4
Organizational structure
Functions under the poroject manager 0 1 0 4 1.3 0.94
Position of supervisor 1 4 1 4 2.6 0.95
Authority over marketing decisions 4 28 12 27 18.0 3.70
Detailed project objectives
Plan 0 5 0.14 4.48 2.62 0.84
a Project performanceis the weighted average performance including the dimensions described in Appendix C.Updating of
the same ®ve point-scale used for the level of information reported. Then, the respondent rates the importance of each type of information. Plan is the weighted average of the detail for each type of information.
6. Results
6.1. Descriptive statistics
Table 4 gives descriptive statistics on variables and representative items in the questionnaire. Time-information receives the most attention (with a mean of 1.98 which means that this infor-mation is updated more frequently than twice a month). This result, corroborated during company visits, indicates that management control systems in product development, following project man-agement techniques, are focused around time. It is also important to observe that traditional accounting measures Ð cost and pro®tability information Ð are the ones used less frequently (with means of 3.73 and 4.17 respectively). In particular, pro®tability information Ð even if it is the ultimate goal of a product development eort Ð is on average updated quarterly or even longer and it is the least discussed measure in meetings. Project managers explained this apparent paradox arguing that the ®nancial attractiveness of a pro-ject is studied before the actual development starts; once the development eort is under way, ®nancial performance is expected to follow from sound non-®nancial performance. Project man-agers also mentioned that they do not explicitly include pro®tability issues when evaluating design trade-os.14 Also notice that this observation
reinforces existing evidence (Abernethy & Brownell, 1997; Brownell, 1985; Rockness & Shields, 1988) regarding the low importance of traditional accounting measures in these types of organiza-tional processes.
Even if time information is used most often, time strategy (exempli®ed by the relative impor-tance of the various strategies) is not as important as customer focus (32.7% for customer focus ver-sus 23.3% for time and 16.1% for cost). Achieving low cost has little importance for the sample studied, which suggests that price pressures in the health industry have not yet aected product development in the medical devices companies.15
On average, more than 50% of the parts designed are new and the number of people involved ranges from 0 (nobody devoted full time to the project) to 106, with an average of 17. Finally, the number of functions reporting to the project manager is only 1.3 (median=1) indicating that companies still use a functional structure even if current research advocates for cross-functional teams. The lack of cross-functional integration was also con®rmed in ®eld visits. It is quite common to have project managers supervising engineers only and reporting to the R&D manager. How-ever, cross-functional teams did exist; for example, in one of the companies visited not only were teams formed with people from dierent func-tions, but also the project leader could come from any function including human resources or accounting.
Table 5 presents the pairwise correlation matrix among independent variables. Technological uncertainty (Tech-X) is correlated with customer strategy (Cust-Str) (0.29) indicating that this strategy is likely to require higher product perfor-mance compared to time and cost strategies. Pro-jects with high technological uncertainty (Tech-X) are also positively correlated with the scope of the project (People) (0.33). Companies in the sample have strong technological capabilities with sizable R&D departments. These capabilities allow them to tackle technologically complex projects by assigning more people.
Managers supervising projects with a high number of new parts (New-Parts) report to a per-son more senior in the hierarchy (0.25) possibly because these products tend to represent a more
14 There is little discussion in the literature as to why
non-®nancial measures are used more often than non-®nancial measures. If the ®nal goal of organizational decisions is pro®tability, it seems reasonable to expect decisions to be made according to this criterium. The evidence suggests otherwise.
15 The two additional product objectives included in this
signi®cant eort by the company, thus requiring top management attention.
Authority over marketing decisions is correlated with market uncertainty (0.23) indicating that project managers perceiving a more complex mar-ket also have more authority to deal with this type of uncertainty. Finally, authority over marketing decisions is correlated with the hierarchical posi-tion of the superior (0.23). This correlaposi-tion may just con®rm that as the project manager's span of attention expands (in this case to include market-ing decisions), (s)he is supervised by a more senior person. Or, alternatively, projects with high mar-ket uncertainty are newer to the organization and, as such, they demand more attention from top management.
6.2. The design of management control systems
To test the hypotheses relating management control systems to project uncertainty and product strategy, I use the following regression model:
Management Control Systems Characteristics =f(Company Dummies, Product Strategy, Product Uncertainty, Organizational Structure)
Table 6 shows the results from OLS regressions for the six types of information reported in the man-agement accounting system. The variance in¯ation
factors and the condition indexes are within the expected ranges Ð thus, multicollinearity is not a problem.
Hypothesis H1a predicted a positive relation-ship between more intense use of customer infor-mation and market uncertainty. The ®rst column in Table 6 (CUSTI) supports this claim and the coecient for market uncertainty (Mkt-X) is positive and signi®cant (the coecient has a value of 0.362 and is signi®cant at the 1% level). In addition, the coecient for authority over mar-keting decisions (Mkt-Dec) is also positive and signi®cant (0.363). In other words, the project manager uses customer information more often when he is responsible for marketing decisions possibly because he faces a higher degree of market uncertainty.
HypothesisH1bpredicted a negative relationship between technological uncertainty and manage-ment control systems. Supporting this relationship, I ®nd three regressions (DESI, TIMEI, and BUDI) where the coecient for technological uncertainty (Tech-X) is negative and signi®cant. This result is in line with management control systems being a poor vehicle to reduce technology-related uncertainty.
Hypothesis H1c related project scope with management control systems being more detailed
Table 5
Correlation matrixa,*
Cost-Str Time-Str Cust-Str Tech-X Mkt-X New-parts People Function Heirarchy Mkt-dec
Time-Str 0.00
Cust-Str 0.00 0.00
Tech-X ÿ0.03 0.17 0.29**
Mkt-X 0.03 0.10 ÿ0.09 ÿ0.07
New-Parts 0.01 0.12 0.09 0.35** 0.12
People ÿ0.05 ÿ0.05 0.11 0.33** ÿ0.08 0.00
X-Function ÿ0.06 ÿ0.03 0.02 ÿ0.15 ÿ0.02 0.00 ÿ0.21
Hierarchy ÿ0.09 ÿ0.05 ÿ0.19 0.26* 0.21 0.25* ÿ0.07 0.04
Mkt-Dec 0.01 0.08 ÿ0.01 ÿ0.20 0.23* 0.15 0.03 0.02 0.23*
Plan 0.25* ÿ0.15 ÿ0.06 ÿ0.04 0.01 ÿ0.17 0.03 ÿ0.09 0.06 ÿ0.14
*10% Con®dence level; **5% Con®dence level.
a Cost-Str: importance of cost to the success of the product, Time-Str: importance of time to the success of the product, Cust-Str:
and more intensively used. The evidence in Table 6 is weak. Only the coecient for new parts (New-Parts) is positive and signi®cant for design infor-mation (DESI) and the coecient for the number
of people (People) is signi®cant for budget infor-mation (BUDI) (value of 0.013 signi®cant at the 5% level). The signi®cance of the number of new parts may just re¯ect the fact that more
Table 6
Results on the design of management control systems in product developmentd
Dependent variableb CUSTIa DESIa TIMEIa COSTIa BUDIa PROFIa Variancecin¯ation factors
Intercept 0.147 0.349 0.369 ÿ0.04 0.423 0.101
Prob(T) 0.74 0.51 0.47 0.94 0.43 0.85
Product uncertainty
Tech-X 0.145 ÿ0.578*** ÿ0.378* 0.018 ÿ0.400* 0.194 1.90
prob(t) 0.44 0.01 0.09 0.93 0.07 0.39
Mkt-X 0.362*** ÿ0.020 ÿ0.288* 0.034 ÿ0.129 0.102 1.42
prob(t) 0.01 0.90 0.06 0.84 0.44 0.55
New-Parts 0.003 0.017*** 0.081 0.003 0.008 0.002 1.45
prob(t) 0.62 0.00 0.12 0.57 0.16 0.74
People 0.004 0.005 0.000 0.006 0.013** ÿ0.003 1.52
prob(t) 0.51 0.43 0.99 0.41 0.05 0.67
Product strategy
Cost-Str ÿ0.160 ÿ0.152 ÿ0.031 0.331** ÿ0.031 0.305** 1.06
prob(t) 0.16 0.25 0.81 0.02 0.81 0.03
Time-Str 0.287** 0.188 0.361*** 0.026 0.012 0.021 1.22
prob(t) 0.02 0.21 0.01 0.86 0.94 0.89
Cust-Str 0.208 0.212 0.191 0.203 0.194 ÿ0.027 1.41
prob(t) 0.14 0.20 0.24 0.23 0.24 0.87
Organizational structure
Function ÿ0.077 0.046 0.045 0.073 0.423 0.065 1.11
prob(t) 0.53 0.75 0.75 0.61 0.44 0.66
Hierarchy ÿ0.160 ÿ0.128 0.034 ÿ0.174 ÿ0.097 ÿ0.213 1.44
prob(t) 0.24 0.44 0.82 0.32 0.56 0.21
Mkt-Dec 0.363*** ÿ0.191 ÿ0.040 0.119 0.175 0.349** 1.20
prob(t) 0.01 0.18 0.77 0.411 0.22 0.02
R2 54.5% 43.3% 39.5% 30.6% 37.1% 28.1% AdjustedR2 38.5% 22.8% 20.4% 10.5% 14.4% 8.8%
N 51 50 51 50 50 52
a The condition index in all regression is around 11.70 (small dierences are due to dierent data points). *10% con®dence level.
**5% con®dence level. ***1% con®dence level. In all the regressions, dummies are used to control for companies with more than 5 projects in the sample.
b CUSTI: use of customer information, DESI: use of product design information, TIMEI: use of time information, COSTI: use of
cost information, BUDI: use of budget information, PROFI: use of pro®tability information, Cost-Str: importance of cost to the success of the product, Time-Str: importance of time to the success of the product, Cust-Str: importance of functionality (customer demands) to the success of the product, Tech-X: level of technological uncertainty, Mkt-X: level of market uncertainty, New-Parts: percentage of new parts, People: number of people involved in the project, Hierarchy: hierarchical level of project manager's superior, Mkt-Dec: project manager's authority over marketing decisions, Function: level of cross-functional integration.
c Variance in¯ation factor is de®ned as the inverse of 1 minus the correlation of the independent variable on the rest of independent
variables. Multicollinearity is considered to be a problem when the variance in¯ation factor is above 100 (A & Clark, 1990, p. 162). d Condition index (or number) is the square root of the relationship of the largest to the smallest eigenvalues of the normalized
information on design is required as the number of parts increases. Therefore, this ®nding should be interpreted with care. The signi®cance of number of people in explaining the use of budget infor-mation (BUDI) suggests that project managers use budget information to coordinate and control as the project grows (Lukka, 1988).
Next, I turn to the hypothesized relationships between product strategy and management con-trol systems. Hypothesis H2a related cost infor-mation with the importance of low cost product strategy. In support of such a relationship, the coecient for low cost product strategy (Cost-Str) is positive (0.331) and signi®cant (at the 5% level) in the regression for cost information (COSTI). Also in support of hypothesisH2b, the coecient for time-to-market product strategy (Time-Str) is positive (0.361) and signi®cant (at the 1% level) for time-related information (TIMEI). In contrast, Table 6 shows no support for hypothesis H2c. If customer information is related to customer strat-egy, then the coecient for Cust-Str would be positive and signi®cant in the ®rst regression (CUSTI). Contrary to the hypothesis, the coe-cient is non-signi®cant.
6.3. The eect of management control systems on project performance
To test for the relationship between manage-ment control systems and project performance, I use the following regression model:16
Project Performance=f(Company Dummies, Plan, Product Strategy, Product Uncertainty, Organizational Structure, Management Control Systems Characteristics, Interaction Terms)
Table 7 presents the results relating manage-ment control systems and project performance.17
The table includes the variance in¯ation factors and the condition indexes to test for multi-collinearity. Because they are within the expected
ranges, multicollinearity can be dismissed as a threat to the results.
The ®rst regression presents the main eect. Supporting the main eect hypothesis (H3a) between management control systems and improved product development performance, the coecients for design (DESI) and cost informa-tion (COSTI) are positive and signi®cant. How-ever, the coecient for time information (TIMEI) is negative (coecient ÿ0.225 signi®cant at the
10%).18 This last ®nding is against hypothesis H3a and agrees with the argument that manage-ment control systems can be detrimanage-mental to project performance. The widespread recommendation that decreasing development time is ``always good'' to gain competitive advantage (Patterson, 1993) may not always hold.
The second regression includes interaction terms to test for contingencies. Two interaction terms are signi®cant. More intense use of customer information for products following a customer-focused strategy has a positive impact on perfor-mance (as hypothesisH3bpredicted). In a similar way, more intense use of cost information is asso-ciated with better performance as the importance of a low cost strategy increases (hypothesisH3c). In contrast, I ®nd no support for hypothesisH3d
relating time information and performance as the importance of time-to-market increases.
As predicted in hypothesisH3e, detailed project planning is associated with improved performance (signi®cant at the 5 and 10% level). Finally, it is relevant to point out that the coecient for Func-tion is positive and signi®cant (0.300 and 0.290 both signi®cant at the 1% level), in line with pre-vious research that found that cross-functional integration bene®ts product development (Clark & Fujimoto, 1991).
7. Discussion
This study sought to explore the drivers of management control systems design in new pro-duct development. The theoretical foundations are
16 I also tested the model using Euclidean distances as
pro-posed by Drazin and Van de Ven (1985). Results do not change.
17 Only relevant independent variables are included to save
degrees of freedom, results are robust to alternative speci®ca-tions. None of the product uncertainty variables was relevant.
18 I also included a quadratic term for TIMEI to test for
