NEW PRODUCT DEVELOPMENT STRUCTURES: THE EFFECT OF CUSTOMER OVERLOAD ON POST- CONCEPT TIME TO MARKET
1. INTRODUCTION
Time to market is widely viewed as a key source of competitive advantage, particularly in "fast cycle" industries [18,19,20] where product life cycles are often two years or less. Three considerations are critical in formulating new product development strategies to improve time to market. First, for a multiple-product firm with frequent new product introductions, cross-product learning among designers is vital for rapid product introduction [8,10,17]. Second, when customers have different needs and their input guides new product specifications, proximity of designers to customers is a key factor in determining time to market [9,15]. Third, with time pressures in product introductions, close coordination between product designers and process engineers facilitates timely production [2,1 1].These considerations suggest sharply different structures for new product development in a global company. For a firm with multiple international manufacturing locations, no structure can simultaneously favor all three factors. When cross product learning is the dominant concern, a firm can facilitate such learning by locating all product designers in one facility. The resulting concentrated new product development structure, however, may not provide product designers adequate proximity to customers and process engineers. In contrast, a distributed structure disperses the new product development effort to multiple manufacturing locations. This structure affords closer interaction of designers with local customers and process engineers, but interaction among designers is reduced by the geographic separation of multiple design facilities. In general terms, compelling reasons can be given in support of either of these two product development strategies [1,5,16]. (It is important to note that the distinctions we recognize in this study pertain to engineering structures essentially and do not reflect overall organizational structural differences.)
In the computer industry and in particular the electronic component industry, manufacturers have adopted vastly different product development structures. We examine the new product development structures of three firms in the data network interface segment. Together, these firms account for nearly 80% of the total sales in that category. The firms have introduced more than 200 new products that have gone to volume production1 in the past five years. All three firms operate globally and, aside from some minor variations, sell standardized products. Although the firms are similar in companysize, market growth and customer base, their new product development structures are very different.
Companies 1 and 2 concentrate their product design teams at one location in the United States and maintain their production facilities in the Far East to take advantage of low labor costs. In this structure, the physical separation of product designers from process engineers reduces interfunctional coordination. However, it fosters cross-product learning among designers. Hence, the concentrated new product development structure is high on cross-product learning and low on learning from customers and product-process coordination (Figure 1).
Company 3, in contrast, has developed a structure in which product designers and process engineers are co-located at several manufacturing facilities, some of which perform sub-assembly operations. Company 3 has established manufacturing and design organizations not only in the Far East for low cost labor, but also in Europe for capital equipment grants for automated processes and for proximity to European customers. It also maintains manufacturing and design teams in Mexico for fast response to the needs of the North American market This network of co-located teams fosters close coordination of design and process engineering at the local plant level. However, the structure inhibits cross-product learning across facilities. Hence, the distributed new product development structure, in the spirit of focused manufacturing, is low on cross-product learning and high on product-process coordination and customer proximity (Figure 1).
We examine how too much customer input affects the two structures. Is the development process slowed by over-listening to the customer? That possibility is of particular concern with the distributed structure, which is able to generate a larger number of customer inputs on new product projects than the concentrated structure [12]. A sufficiently large number of such inputs can potentially overload the new product development process, causing confusion, coordination problems, duplication of effort, and resource constraints and thereby extending the time to market. We therefore address two questions: How much does time to market increase in each structure as the amount of customer input is increased? Does the effect differ at different levels of customer input?
By controlling for the number of projects and varying the number of customers across the two structures, we show that up to a certain level of customer input, the distributed structure is superior in time to market because of its higher product/process coordination. However, beyond that level, this structure's greater time increase shifts preference to the concentrated structure.
2. RESEARCH SITES AND DATA COLLECTION
2.1 Research Sites
For our study, we chose three international manufacturers of electronic components for the computer and data communications industries. Each company has less than $100 million in annual sales, but maintains extensive international manufacturing and distribution capabilities (Table 1). All three are headquartered in the United States and are publicly traded on the NASDAQ exchange.- Table 1 here -
The companies produce interconnect filter modules, a product used in every piece of data processing and data communications equipment that is connected to a local area network (LAN) or wide area network (WAN). The modules filter and condition electronic data signals that enter and leave the network. Because of rapid advances in networking technology and component miniaturization, the product life cycle is relatively short, ranging from two to three years The total market for this specialized hybrid circuit has grown rapidly and is currently estimated to be about $250 million in combined worldwide sales for the three manufacturers and their several smaller competitors. Our selection of the interconnect data filters segment of the electronic component industry was based on the following four factors:
- All three companies view new product development as a key to their long-term success. As a result, the rapid rate of new product introduction in this industry provides sufficient data to analyze the relationship between new product development structures and time to market.
- The relatively small size of these companies enabled us to interact closely with designers, cross-functional teams, and senior decision makers and thus gain a deep understanding of the new product development process.
- The firms are comparable on several dimensions, such as company size, market growth, and customer base. Moreover, major customers dictate similar product specifications, pricing, and delivery for all three firms. We therefore did not have to control for such factors in determining the impact of the two product development structures on time to market.
- Despite the similarities of products and markets, these competitors have very different product development structures Two firms concentrate product development activities in a single location in the United States and the third firm splits product development among three different manufacturing facilities.
2.2 Product Development Process
Data interconnect filter modules and related components must be carefully matched to the requirements of the overall data communications network in which they reside. Therefore, companies manufacturing these devices design new products to meet the highly detailed electrical and mechanical specifications dictated by major customers (large computer manufacturers, and LAN and WAN equipment suppliers), standard- setting committees such as IEEE, and key complementary component manufacturers such as the large producers of integrated circuits. Each of the three firms in our study has a long history and familiarity with the key sources of design input and has established efficient methods to capture the pertinent information and specifications required for new product concepts.Once a new product concept is chosen, a product design team begins consideration of alternative design tradeoffs in light of the required specifications. The design team may include individuals from manufacturing, quality, and marketing, as well as materials suppliers. The design team initiates an iterative process of developing or improving a product design, building samples of that design, and testing the samples for specification compliance. This process continues until the design team is satisfied that the product design optimally meets the market's requirements and specifications and is manufacturable at the target cost. If the team concludes that the product development effort is not economically feasible, the project is abandoned.
The prototype design is then transferred to the plant(s) that will manufacture the product. At this point, process engineers at the plant level begin developing the process design, acquiring tooling and process equipment, assembling tools and fixturing, and training production operators in the required process steps. When process problems are resolved, production of the new product commences and is ramped to volume levels. In our study, we mark the end of this second phase of the development process by the shipment of the two-thousandth unit. Production managers in the three firms believe that by the time this cumulative volume is reached, the process difficulties are resolved and regular supply to customers can be ensured. We define then the time to market as the elapsed time from the date when the new product concept is finalized to the date volume production is achieved.
2.3 Data Collection and Variable Definitions
Data were collected from each of the three companies as part of a major field research effort during the summer and fall of 1993. For companies 1 and 3, product and operational data were obtained for the first calendar quarter of 1988 through the second calendar quarter of 1993. For Company 2, data were available only for the first quarter of 1990 through the second quarter of 1993. The new products were developed by design teams at various sites in the Far East, Europe, and North America.After extensive discussions with designers, cross-functional team members,
and project managers, we obtained critical information on the progress of
new product development at the three companies. The necessary information
was obtained from company records in accounting, engineering, manufacturing,
and marketing. From an initial set of 419 products, we retained 220 new products
for which complete histories on time to prototype, time to volume production
and engineering expenditures were available. We collected information at the
product level on the following four key variables:
1. Number of customers (CUST) - The number of individual customers
that expressed intent to buy the product and provided input on its design.
2. Engineering expenditures (ENGG) - The expenditures for engineering
related to new product development and process design (excluding engineering
expenditures to maintain ongoing production). In some cases, records were
available to assign expenses to individual products. In others, managers determined
engineering costs by individual products. The expenditures represent the investment
made by each company for product development effort.
3. Number of concurrent products (CPROD) - The average number of products
being developed concurrently during time to market.
4. Time to Market (MTIME) - The time taken from the time the new product
concept was finalized to achieving volume production.
2.4 Model
Figure 2 is a graphic representation of our proposed conceptual model. The dependent variable is time to market (MTIME). The independent variables are number of customers (CUST), engineering expenditures (ENGG), and number of concurrent products (CPROD).
The model can be represented by the following equations.
For a given product, the number of competing products is assumed to be exogenous. We expect that increasing the number of customer interaction and the number of products competing for the development resources would adversely impact the time to market. Conversely, increasing engineering expenditures would shorten the development cycle. Since the dependent variable is strictly positive, a regression based approach is inappropriate since the estimates will be biased and inconsistent. Moreover, predictions cannot be constrained to be positive and negative time values may be obtained. When the dependent variable is a measure of the time from one state to another, as in our data from start to volume production the instantaneous probability of transition may vary over time. The hazard function model captures such "duration dependence." It has been widely employed when the underlying dependent measure is time [13,14]. We provide below a brief description of the model.
Let ti, denote the time to market for product i. We assume that ti is a random variable with a probability density function f(t) and cumulative density function F(t). The hazard function, which gives the instantaneous rate of product completion at time t conditional on the product not having been introduced to the market until that time, is defined as follows:
For our empirical analysis, we employ a proportional hazard formulation [
14] and parameterize the hazard function as follows:
whrere:
In equation (5), Ic is an indicator variable which takes a value 1 if the product i is introduced by the concentrated structure and O otherwise. Note that the above specification for the hazard function is very flexible and allows for non-monotonicity in duration dependence. If both the parameters a and b in equation (4) are negative, the hazard function is monotonically decreasing; if both the parameters a and b are positive, the hazard function is monotonically increasing; and, if a is positive while b is negative, the hazard rate first increases and then decreases over time.
Given a random sample of N product development projects, the log-likelihood function for the sample is of the form
The consistent and efficient estimates of the parameters are obtained by maximizing the above log-likelihood function.
3. RESULTS AND MANAGERIAL IMPLICATIONS
3.1 Results
We estimate two models. The first is a baseline model which does not incorporate the differences across the two structures with respect to the impact of the exogenous variables on the expected time to market. The second model recognizes these structural differences. The parameter estimates for the two models are presented in columns 1 and 2 of Table 2.- Table 2 here -
As seen in the table, the parameter estimates for the exogenous variables are statistically significant and in the expected direction. To test the hypothesis that there is a differential impact of the variables (CUST, CPROD and ENGG) across the two structures, we conduct a likelihood ratio test. The //MATH// test statistic of 940.12 with 3 degrees of freedom strongly support our hypothesis.
The raw data for response time in the two structures without control for the number of products and customers across the two structures are shown in Figure 3. The concentrated structure has an average time to market of 4.27 quarters. In contrast, the distributed structure has a time to market of 5.41 quarters or about 30% longer than that of the concentrated structure.
However, this difference reflects the fact that response time depends on the actual number of customers for a product and number of concurrent products in each structure. Company 3, in comparison to the other two, has a higher average number of customers (33.4 vs. 18.4) and concurrent products (37.8 vs. 21.8) and therefore may be operating in an overloaded state relative to the companies with concentrated structures. If we control for this difference by comparing performance at the overall sample averages for customers and concurrent products, we obtain the times to market shown in Figure 4.
With this control, the time-to-market advantage reverses and favors the distributed structure (4.43 versus 4.71 quarters). This finding provides evidence of the advantage of close coordination between product designers and process engineers when customer interaction is efficiently managed [6]. Using the customer and product averages of the concentrated structure companies (rather than the overall sample average) shows an even greater advantage for the distributed structure (3.88 vs. 4.21 quarters). In fact, when we looked at specific products when the distributed structure had lower number of customers, we found a shorter time to market for the products developed during these periods.
Figure 5 shows the effect on total development time as a function of the number of customers (all other parameters held equal at sample average). The distributed structure
provides shorter development time with up to an average of 25 customers, then its relative performance degrades and the concentrated structure thereafter dominates. The response function is nonlinear and the degradation worsens with increasing overload. The 25- customer level is within the low end of the range of the distributed structure and the high end of the range of the concentrated structure. Hence, if Company 3 could establish mechanisms to more efficiently manage the interactions between customers and designers, it would outperform its concentrated structure rivals in time to market. It could then fully exploit gains from its product/process coordination as well as customer proximity.
3.2 Limitations
Our analysis addresses certain important issues in new product development, but the results must be interpreted with caution. Although we focus on a specific product type to minimize potential bias due to differences across products, such variation may be present within the family of products used in our study. Specifically, the level of complexity of a given product may directly affect time to market. Direct estimation of the impact of such product characteristics on time to market would yield better managerial insights. The additional data requirements for such a study would be substantial, particularly in view of the firms' reluctance to share confidential information on new product specifications. We view the analysis presented here as a first step in understanding the influence of product development structures on key elements of time to market.The unique qualities of this industry provide a very controlled environment. Applicability beyond this industry must be carefully considered. For instance, the key point of price and quality similarity among competitors can not be extended to the systems level in the computer and LAN/WAN equipment markets. There may be applicability in other types of industrial products such as components for the auto industry and other durable goods where similar price and quality control is exerted by large customers. We hope to pursue this in future research efforts.
Our analysis is based on 220 new products that had reached volume production. Occasionally, new product development efforts are terminated at intermediate stages of the prototyping process. To that extent, the sample is biased in favor of products that were successfully prototyped. We are unable to analyze incomplete development efforts because firms do not keep critical information on products that are dropped.
In our analysis, we do not have any measures of incremental newness of the products over the previous generations. Griffin [7] shows that newness of the product increases the overall product development time. Whether the resulting misspecification biases our estimates or not remains as an empirical issue. By and large, the competing firms developing the product deal with the same degree of newness, and to that extent the comparisons across structures will remain unaffected. A related issue is the extent to which re-use of existing products facilitates new product development [4]. Lack of data prevents us from examining such issues.
Finally, time to market is only one of a host of measures of successful new product introductions. In particular, cost, profit margin, market share, and other measures are also key criteria for evaluating new product introductions. The data requirements and framework for analyzing such measures are beyond the scope of this study.
4. CONCLUSIONS
In the industry we examined, the global customer needs were uniform. An interesting
issue for future research is the degree to which a distributed new product
development effort is necessary if customer needs vary significantly. Our
study is an initial effort to understand the complex issues related to new
product development strategies and time to market. As more detailed data on
new product development become available, we hope to undertake a more comprehensive
study of those issues. Finally, time to market is critical only if it leads
to gains in market share or a overall increase in profit. We hope to study
these issues in future research efforts.
The study was sponsored financially by the Sloan Foundation through the Computer Industry Study being conducted by Stanford University. The help and support provided there by William Miller and Garth Saloner are most appreciated. Additional funding was provided by the Carnegie-Bosch Institute at Carnegie Mellon where the assistance of Bruce McKern is gratefully acknowledged. The authors thank the senior executives, design engineers, and manufacturing personnel of the subject companies for their cooperation in providing access to confidential data and sharing valuable insights. They also thank Tom Hustad and two anonymous reviewers for their helpful comments and suggestions on the earlier version of this article. Finally, the authors wish to acknowledge the data collection efforts of William Kugler, Venketesh Nagar, and David Williamson.
1 Volume production denotes cumulative production of at least 2000 units. The measure is an industry norm.
REFERENCES
[1] Bartlett, C.A. and S. Ghoshal (1989), Managing Across Borders: The Transitional
Solution, Harvard Business School Press, Boston.
[2] Clark, K.B. and T. Fujimoto (1991), Product Development Performance: Strategy,
Organization and Management in the World Auto Industry, Harvard Business School
Press, Boston.
[3] Datar, S., C. Jordan, S. Kekre, S. Rajiv, and K Srinivasan (1994), "New
Product Development Structures and Time to Market," working paper, Carnegie
Mellon University.
[4] Eppinger, S.D. and K.T. Ulrich (1995), Product Design and Development.
New York: McGraw Hill, Inc.
[5] Ghoshal, S. and N. Nohria (1993), "Horses for Courses: Organizational
Forms for Multinational Corporations," Sloan Management Review, Winter.
[6] Griffin, A. and J.R. Hauser (1993), "The Voice of the Customer," Marketing
Science, 12, 1 -27.
[7] Griffen, A. (1993), "Metrics for Measuring Product Development Cycle Time,"
Journal of Product lnnovation Management, 10:2, 112- 125 (March).
[8] Hamel, G. and C.K. Prahalad (1991), "Corporate Imagination and Expeditionary
Marketing," Harvard Business Review, July-August.
[9] Hauser, J.R. and D. Clausing (1988), "The House of Quality," Harvard Business
Review, May-June.
[10] Hayes, R.H., S.C. Wheelwright and K.B. Clark (1988), Dynamic Manufacturing:
Creating the Learning Organization, The Free Press, New York. Chapters 10
and 11.
[11] Iansiti, M. (1993), "Real World R&D; Jumping the Product Generation
Gap," Harvard Business Review, May June.
[12] Jordan, C.C. and S. Kekre (1993), "The International Knowledge Factory:
Competitive Structures for the Absorption of Market and Technical Information,"
working paper, Carnegie Mellon University.
[13] Kalbfleisch, J.D. and R.L. Prentice (1980), The Statistical Analysis
of Failure Time Data, John Wiley & Sons, New York.
[14] Lancaster, T. (1990), The Econometric Analysis of Transition Data, Cambridge
University Press, Cambridge, England.
[15] Nayak, P.R. and A.C. Chen (1993), "Listening to Customers," Prism (Arthur
D. Little), Second Quarter
[16] Porter, M.E. ed. (1986) Competition in Global Industries, Harvard Business
School Press, Boston.
[17] Sanderson, S. and M. Uzumeri (1993), "Managing Product Families: The
Case of the Sony Walkrnan," working paper, Rensselaer Polytechnic Institute.
[18] Stalk, G. Jr. (1988), "Time: The Next Source of Competitive Advantage,"
Harvard Business Review, July-August.
[19] Stalk, G. Jr. and T.M. Hout, (1990), Competing Against Time: How Time
Based Competition is Reshaping Global Markets, The Free Press, New York.
[20] Williams, JR. (1992), "How Sustainable Is Your Competitive Advantage,"
California Management Review, Vol. 43, No. 3.
