4 HYPOTHESES TESTING METHODS
4.4 MEASUREMENT CONSTRUCT OPERATIONALIZATIONS
4.4.3 Control variables
Several control variables were included in the regression model to take into account the potential effects of the firm, industry, and the front end project itself in the final results. Control variables for firm-level effects included the size and R&D intensity of the company. Industry-level effects were considered using the industry sector as a dummy variable including three categories: piece goods industry, process industry, and other industry. Control variables for front end project level effects included the original objectives set for the project, definition of the front end process, and the uncertainty included in the project.
Firm size
Several studies have shown that the size of the firm can affect the final outcome of the process as well as how the activities are generally organized (controlled) in the company609. Larger companies, for example, rely less on personal control and more on control through bureaucratic structures (rules and procedures)610. Murphy and Kumar have found that smaller firms are more successful (meet or exceed market projections) than large firms because the products are typically designed for the more specific needs of a small target group or built directly for a customer under a defined contract611. Turnover (in 2004) was used as a variable controlling the size effect in this study. Because of its strong negative skewness toward small sales figures, a logarithmic transformation of turnover was used.
608
Ramaswami 1996 609
Donaldson 2001, Murphy and Kumar 1996, Ouchi 1977 610
Donaldson 2001, Jaworski 1988 611
R&D intensity
Murphy and Kumar found that the intensity of the firm’s R&D efforts predicted activities across the front end, for example R&D-intense firms stressed the importance of creativity and the utilization of internal employees in the idea generation process612. The size of the firm is naturally related to the number of different R&D projects undertaken613. Thus instead of the number of different R&D projects, logarithmic transformation of the percentage of turnover invested in R&D was used to control R&D intensity.
Industry sector
Industry sector was measured by requesting that respondents indicate the industry sector in which the company is operating (open-ended question). The classification of three dummy variables was done by the author post-hoc. From the management control point of view, it was considered appropriate to have two broader categories, piece goods industry and the process-based production industry, with specific characteristics that may influence the final performance of front end projects. Industries such IT/ICT technology and medical/biomedical were classified under the label “other industry”.
Objectives of front end project
The nature of defined objectives may have an effect on the final performance of the front end project614. The objectives of the development project were controlled by using two categories (a dummy variable). Respondents were requested to choose (which one of the statements describes the project objectives better) whether the objective of the project was to improve long-term profitability or short-term cash flows. The objectives of the development project are closely related to the other project-level control variable ‘uncertainty’.
612 Ibid. 613 Ibid. 614
Definition of front end process
The existence to which the front end process was defined was used as a control variable. Respondents were requested to choose from four categories: not defined at all, defined superficially, defined in some detail, and defined in great detail. Definition of the front end process may indicate its perceived importance and a general maturity level of front end execution.
Uncertainty included in the development project
Uncertainty was used both as a control variable and as a moderating variable in multiple linear regression analyses. A wide body of knowledge exists to measure uncertainty in different business contexts. The measurement items were modified to fit the context of this study from Danneels and Kleinschmidt615, Garcia and Calantone616, and Danneels617 that all measured uncertainty in the product innovation context. The items were measured with a five-point Likert scale (1 = strongly disagree…5 = strongly agree).
There are two main issues defining uncertainty in the product innovation context: applied technology and the target market618. The more new technology the product includes or the more unfamiliar the target market is, the more uncertainty the development task includes. Thus the uncertainty measurement covered both market and technology dimensions. Garcia and Calantone emphasized that product innovativeness (the uncertainty the product includes) must be evaluated from two different perspectives: the macro-level industry perspective and the micro-level company perspective619. The first two items reflects this notion both in market uncertainty and technology uncertainty variables. Considering this distinction, these items were modified to fit the context of this study from Danneels and Kleinschmidt, who used these measurement items in the market familiarity and technological familiarity measurement constructs620. The third and fourth items in both constructs relate to the discussion of whether the new products can rely on the firm’s existing technological and marketing competencies or not. This is an important measure of
615
Danneels and Kleinschmidt 2001 616
Garcia and Calantone 2002 617
Danneels 2002 618
Tidd et al. 2001, Danneels and Kleinschmidt 2001, Lynn and Akgun 1998 619
Garcia and Calantone 2002 620
uncertainty in this study since products with a closer fit with the existing competences of the firm tend to be more successful on average621. The third and fourth items in the market uncertainty construct and the third item in the technology uncertainty construct were modified from Danneels and Kleinschmidt622. The fourth item in the technology uncertainty construct was created and found to be functioning adequately, based on the discussion of Danneels and Kleinschmidt623.
Table 11 illustrates factor loadings for these two measurement constructs. Two different factors with a clear factor solution and high loadings were found as expected. Bartlett’s statistic is significant, MSA is .69, and the factor solution explains 63% of total variance. Cronbach’s inter-item coefficient alpha for the market uncertainty construct is .76 and the technology uncertainty construct .84.
Table 11. Measurement items and factor loadings for market uncertainty and technology uncertainty constructs.
Measurement items Factor 1 Factor 2
Market uncertainty
The planned target markets for the product concept were new to our organization. ,15 ,84
The planned target markets for the product concept were also new to other companies in
the industry of our organization. ,11 ,75
Our organization’s existing market research capabilities were not adequate for the
gathering of market information needed for the product concept. ,09 ,72 The market research/gathering of market information was done by using new methods
that were not previously used in our organization. ,17 ,68
Technology uncertainty
The applied technology in the product concept was new to our organization. ,90 ,06 The applied technology in the product concept was also new to other companies in the
industry of our organization. ,84 ,09
Our organization’s existing R&D capabilities were not adequate for developing the
product concept. ,65 ,23
Technology development and technology verification of the product concept was done
using new methods that were not previously used in our organization. ,82 ,19 Principal component analysis with Varimax rotation
N = 133, MSA = .69, Total variance explained = 63%
621
Danneels 2002 622
Danneels and Kleinschmidt 2001 623