Financial Resource Allocation in the Front End of Innovation

(1)

1

Financial Resource Allocation in the Front End of

Innovation

Master Thesis in Business Administration Specialization: Strategic Innovation Management

University of Groningen Faculty of Economics and Business First Supervisor: Dr. R.A. van der Eijk Second Supervisor: Dr. J.D. van der Bij

20-6-2016 Mark van den Berg

S2590107

Van der Doesstraat 3a 9726GX Groningen m.van.den.berg.19@student.rug.nl

Public version

Abstract

This study describes and tests a model regarding the impact of financial resource allocation in front-end activities within new product development (NPD) projects on financial performance. Taking a single industry and single firm perspective, data was collected from one of the largest consumer electronics firms in the world. Confronting the Resource Advantage Theory with the Constrain Theory, we hypothesize that relative more invested resources in front-end activities positively affects performance while controlling for internal and external conditions. By using data from 58 product groups and 11 business units over the last five years, it is found that investments in front-end activities indeed favorably and indepfront-endently impact financial performance. Although weakly significant, the results also indicate that the degree of newness of a NPD project positively moderates this relationship. This is the first study in this context that adapts a product group and business unit level perspective and empirically proves the, often underestimated, importance and influence of front-end activities on NPD. Theoretically, this study contributes to the development of the front-end of innovation literature and more specifically, the Resource Advantage Theory. This study is also interesting to managers, as they can use the findings to make more effective resource allocation decisions within NPD projects and more specifically, in the front-end of innovation. Keywords: New Product Development, Front End of Innovation, Resource Allocation, Resource Advantage Theory

(2)

2

Preface

(3)

3

Executive Summary

Product life cycles are decreasing in almost all high-technology goods industries, forcing companies to shorten their New Product Development (NPD) cycle and increase their innovation activities in terms of time and quality (Sommer, Hedegaard, Dukovska-Popovska, & Steger-Jensen, 2015; Labahn, Ali & Krapfel, 1996; Cooper, 1990). Remarkably, advancements in the literature on NPD do not seem to cause a decrease in failure rates within innovation projects (Cooper, 2013). In an effort to explain this paradox, a growing number of scholars point to the front end of innovation (FEI) as an underestimated but crucial factor for NPD performance (Markham, 2013). Contributing to this literature stream, this study attempts to link the FEI with NPD performance by examining the influence of resource allocation decisions within the FEI.

Findings of previous studies suggested, in line with the Resource Advantage Theory, that allocating relatively more resources in NPD results in better NPD performance (Daniel, Lohrke, Fornaciari, & et al., 2004). This paper analyses whether this logic also holds true within a FEI resource allocation context. In addition to the latter, the influence of the degree of newness (i.e. innovativeness) of an NPD project is also considered and examined. Hypotheses are tested by performing multiple statistical analyses on multi-year data from 58 product groups and 11 business units of a large multi-national consumer electronics firm.

The two most important findings of the study are: (1), within our model and sample, investing relatively more financial resources in the FEI predicts higher commercial sales performance and (2), the degree of newness has a positive moderating effect on this relation. The latter implies that NPD projects with a higher degree of newness benefit more from relatively more invested financial resources than NPD projects with a lower degree of newness.

(4)

4

1. Introduction

"Innovate or die" is a statement that is often heard in the modern day business world. With the ever increasing globalization and competition, product life cycles are decreasing in almost all high-technology goods industries, rendering products obsolete sooner and more often (Sommer, Hedegaard, Dukovska-Popovska, & Steger-Jensen, 2015). As a result of this intensified competition, the pace of innovation has never been higher and companies are therefore forced to shorten their product development cycle, speed up their innovation activities and move newer and better products to markets faster (LaBahn, Ali, & Krapfel, 1996). Managing the development of new products has become a top priority in many companies as competitors rush to commercialize technologies and satisfy fast changing customer needs (Mullins & Sutherland, 1998). In order to effectively and timely develop new products and to meet changing customer needs, companies are increasingly adopting formal New Product Development (NPD) processes to guide their product innovation projects (Cooper, 1982). Brown & Eisenhardt (1995) defined NPD as a process for success, survival and renewal. Although many studies focused on exploring best practices, even today NPD failure rates remain surprisingly high at roughly 65% for established firms (Adams, Bressant, & Phelps, 2006). It seems that managers keep struggling with structuring and managing these processes (De Medeiros, Ribeiro, & Cortimiglia, 2014). Simply adapting best practices may be not enough to stay competitive in the current age of innovation and information (Dervitsiotis, 2010).

1.1 Problem definition

(7)

7 time and financial resources must be devoted to the activities that precede the development of the product” (Cooper, 1988 p.250). The findings of Cooper (1988) and Markham (2013) are related to the Resource Advantage Theory of Hunt & Morgan (1996). This theory essentially claims that having sufficient resources (i.e. resource slack) increases success in the whole NPD process. However, there is no general consensus in the literature up until now as to whether having resource slack is indeed beneficial for NPD in general, let alone for the and NPD performance. A contrasting literature stream developed around the Constrain Theory (Ward 1995; Katila & Shane, 2005), which claims that having less (i.e. no resource slack) resources has a positive effect on creativity and out-the-box thinking which in term positively influences FEI and NPD performance. A possible explanation for the lack of consensus on the effects of having sufficient or not sufficient resources, is that research on FEI and resource allocation has taken different perspectives and definitions with regards to FEI, resource allocation and especially measures of success and performance. Beyond resource allocation, there are numerous studies that suggest that the innovativeness of a NPD project and its subsequent product is an important driver of NPD outcomes (Carbonell, Rodriguez Escudero, & Munuera Aleman, 2004). As characteristics of a NPD project with a high or low innovativeness differs greatly (Cooper, 1990), one might expect that resources allocation within these contexts may play different rolls and has differing effects on NPD outcomes. As Cooper (1988), (Brentani & Reid, 2012) and Markham (2013) all argued that the FEI is an underestimated, but very important and for success decisive phase in the NPD process, it is interesting to gain knowledge on how resource allocation affects this phase. As the seeds of NPD success are probably sown in the first few steps of the NPD process (Cooper 1988), the question remains how these seeds should be sown, or, more specifically, how to allocate resources in the FEI.

1.2 Purpose and significance of study

(8)

8 more empirically tested knowledge on the FEI (Markham, 2013; Akbar & Tzokas, 2013). With this knowledge, managers will have more grip on the process of allocating resources within NPD projects. Thirdly, the theoretical contribution of the study at hand, lays in the confrontation of the aforementioned Resource advantage and Constrain Theory. By combing different perspectives offered by these two theories, this study attempts to develop a more holistic perspective on resource allocation, the FEI and NPD performance. The specific contribution lies in a reconceptualization of resource allocation practices in FEI that has not been emphasized by empirical research to date. If NPD performance affects firm performance, FEI performance affects NPD performance, and resource allocation affects FEI performance, then increasing knowledge on how to effectively allocate resources could be the first step in increasing firm performance.

1.3 Research question

This study aims to provide empirical evidence to demonstrate the specific nature of the relationship between resource allocation in the FEI and its subsequent effects on NPD performance. The emphasis lays on validating a direct relation between factors that underlie FEI performance and NPD performance by answering the following research question: “What is the relationship between

resource allocation in the front end of innovation and subsequent NPD performance”?

To answer the main research question, the following sub-questions are formulated: - To what extent does pre-development influence the outcomes of a NPD project?

- How does new product innovativeness influences the relationship between pre-development investments and outcomes of a NPD project?

1.4 Research scope and domain

(9)

9

1.5 Research outline

This study continues with an in-depth review of relevant theories and concepts to construct a theoretical framework. Based on the developed framework, the hypothesis are developed and described. Subsequently, the results of the analysis will be presented and explained before they are compared with the insights of the expert panel. Next, in the discussion section, the findings are compared to relevant literature. Finally, concluding remarks and implications are given which are the basis for future research directions.

1.6 Summary chapter

(10)

10

2. Theoretical background

The literature review gives an overview of existing scientific literature and defines the applied theoretical constructs. The main theories and constructs used in this study are summarized in table 1.

Table 1

Theoretical concepts and explanations

Concept Explanation

New Product Development All the actives executed, resources allocated and decisions made to generate ideas, develop, test and launch the product (Cooper, 1994; Krishnan & Ulrich 2001).

Stage Gate Model A project management technique in which a NPD process is divided in stages to provide structure in moving the new product process from idea to launch (Cooper, 2006; Sommer et al., 2015).

Front End of Innovation The pre-development stages and activities of a NPD project (Wilemon, 2002; Markham, 2013).

Resource Advantage Theory Dynamic process theory which is used to explain differences in productivity and efficiency (Hunt & Morgan, 1996).

Resource Based View A model that sees resources as key to superior firm performance and a basis for competitive advantage (Barney, 1991; Henard & McFadyen, 2012). Constrain Theory Resource theory which predicts higher efficiency and effectively when having

constraining resources (Ward, 1994; Katila & Shane, 2005).

Degree of Newness Measure to determine the newness or innovativeness of a NPD project, process or product relative to the firm and/or market (Langerak & Hultink, 2006; Carbonell et al., 2004).

2.1 High tech innovations

The research context is set in the high technological consumer goods industry. The high-tech consumer goods industry is characterized by rapid development cycles and short product life cycles (Sommer et al., 2015). Remarkably, the number of innovation projects within firms in this industry is usually higher than the capacity to bring them all successfully to the market (Cooper & Edgett, 2012). This chronic lack of capacity results in complex resource allocation issues that managers and decision makers have to deal with (Wang, et al., 2002; Bolumole, Calantone, Di Benedetto, & et al., 2014).

2.2 New Product Development

(11)

11 is described as the transformation of an idea into a tangible or intangible product available for sale. Story, Smith and Saker (2001) point out that the particular terminology depends on the domain in which it is used and attribute this to the interdisciplinary nature of NPD. According to Cooper (1979), those in the management and marketing domain use the term NPD, but those in the R&D domain refer to 'innovation'1_{. Regarding the relevance of this topic, an increasing amount of scholars and industry}

experts are beginning to realize that having a successful NPD process is the single most important factor to be a winner in the global market (De Brentani, Kleinschmidt & Salomo, 2010).

Contributions to NPD literature started with the seminal studies by Booz, Allen & Hamilton (1982) and further developed in the 1980s (Cooper, 1982; Cooper 1988). It is interesting to note that "The level of

importance attributed to new product development is not matched by the level of success" (Cooper,

1990 p.44). Crawford (1979) and Booz, et al. (1982) are reporting NPD failure rates between 30% and 50%. More recently, Cooper & Kleinschmidt (1995) and Cooper (2013) found that the success rates of innovation projects did not improve in the past decades and remain evenly high. A remarkable conclusion, especially considering the high costs, wasted resources and loss of potential revenue that firms face when an NPD project fails. In his literature review, Eisenhardt (1995) identified three general shortcomings in the literature field on NPD which are of interests in the current study. Firstly, many contributions to the literature are yet to be empirically tested. Secondly, only a few studies focus on project, product group or business level as compared to the more frequently applied firm level perspective. The third shortcoming, also recognized by Markham (2013), is the lack of focus on specific phases within the whole NPD process.

Despite the high failure rates, there is widespread acknowledgement on the fact that NPD is a crucial activity for corporate survival (Cooper & Kleinschmidt, 2007). Eleborating on the latter and taking a wider perspective, NPD makes a significant contribution to the economic health and prosperity of societies and countries (Dervitsiotis, 2012). This acknowledgement has led many researchers to seek the ultimate ingredients for NPD success resulting in a comprehensive and complex body of literature. Where scholars are looking for success factors for NPD, managers also recognize the high failure rates of innovation projects and the apparent importance of NPD (Cooper & Edgett, 2012). To deal with the vast complexities of NPD and to reduce failure rates, many organizations have developed systematic Stage-Gate processes to structure and standardize their NPD process (Cooper, 1994).

1_{While acknowledging that the terms ‘New Product Development (project)’ and ‘Innovation (project)‘ are}

(12)

12

2.3 Stage-Gate Models

The first generation stage-gate model (SGM) was developed by Robert G. Cooper in the 1980s. The importance of the development of SGM is reflected in the fact that currently 88% of U.S. business employ SGM to manage, direct and control their product innovation efforts (APQC benchmark, 2010). Emphasizing on the relevance of SGM, there is widespread notion in the literature on SGM adaptation by organizations (Griffin, 1997; Koen, 2003; Adams, 2004; Mills, 2007; Grolund, Ronneberg & Frishammer, 2010). In the current study, the definition for Stage-Gate processes by Cooper (2006 p.214) is followed: “A Stage-Gate process is a conceptual and operational map for moving new product projects from idea to launch and beyond—a blueprint for managing the new product development process to improve effectiveness and efficiency”.

If many organizations adopt formal, standardized and structured development processes, but the failure rate of NPD remains high, what evidence suggests that using SGM’s is effective? Does this approach affects performance at all? In an attempt to answer this question, Booz et al. (1982) found that it does to a certain extent: firms that applied SGM’s performed, on average, better than firms with no such structure. More recent research by Cooper (2006) reached similar conclusions. Firms without SGM have, generally seen, even higher failure rates and a less

successful overall NPD process. According to Sommer et al. (2015), the major advantage of using SGM is to provide the quality focus in a firms’ new product program. Overall, the SGM methodology is built upon structure and standardization (Cooper, 1990). This entails difficulties for managing projects in their early stages of their life cycle (e.g. development phase). This is due to the fact that pre-development activities are all about flexibility and creativity (Griffin & Langerak, 2014). These are factors that transcends the classic SGM methodology which can cause managers to loose grip in the front end of innovation which in term could result in failures. A comparison and analysis of the

Stage-Gate based NPD at the focal

firm

(13)

13 differences and similarities between the focal firms’ Stage-Gate approach and the famous standard Stage-Gate model, developed over the years, by Cooper can be found in appendix C.

2.4 Front End of Innovation

Elaborating on the latter, there is no general consensus on the definition of the front end of innovation (FEI) in the literature yet (Wowak, Craighead, Ketchen, & et al, 2015). Additionally, there is no general consensus on what the exact activities include which together make up the FEI. Scholars are divided over the reputation and context of the FEI. Some authors call it the 'fuzzy' front end of innovation, meaning a mysterious and unmanageable stage (Reinertsen, 1999). Other authors noted that activities as idea genesis, idea selection and feasibility analysis make it manageable per definition and therefore refuse the use of the term ‘fuzzy’ (Urban & Hauser, 2004; Koen et al., 2001). Definitions of the FEI are all grounded on the process of bringing an idea to a more developed form; a business case on which a go/no go decision can be made for further development. In the current study, the most widely used definition of Kim & Wilemon (2002) is followed, defining the FEI as “predevelopment stages consisting of idea generation, product definition, and project evaluation” (p. 269). When an innovation project completes the FEI phase and receives approval, it moves into the Back End of Innovation (BEI). The BEI is defined as the execution-oriented formal process after the official approval has been given to continue to product development and subsequently commercialization, including testing, production start-up and market launch (Cooper, 1990). Linking the findings of Wowak et al. (2015) and Kim & Wilemon (2002) with the SGM methodology, we can divide the NPD process in two main phases, the Front End of Innovation and the Back End of Innovation (Cooper, 1990; Menor, Tatikonda & Sampson, 2002; Kim & Wilemon, 2002). Figure 1 depicts the graphical representation of this distinction.

Figure 1

Stage Gate Model of a New Product Development process

Idea Gate 1 Preliminary Gate 2 Gate 3 Gate 4 Gate 5

Assesment

Business

Case Development Testing

Product Launch

In the FEI, the main goal is to analyze, create and iterate on many alternatives that can potentially result in breakthrough opportunities (Markham, 2013). The FEI is the first step in NPD process in which opportunities are identified and concepts are developed before entering the BEI (Kim & Wilemon, 2002; Kock, Heising & Gemunden, 2014). In the idea generation stage, managers need to screen ideas with the objective to narrow ideas down (Cooper, 1990). Concerning preliminary assessment, the

(14)

14 attractiveness of remaining ideas are examined and an analysis of market conditions is performed (Papastathopoulou & Avlontinis, 2001). According to Cooper (1990) the result of this stage, and thus the FEI, could be a product concept, business case, startup plan and/or product specifications for insertion in the BEI.

Previous research on the FEI focused on observations and reports of human behavior in the FEI (Das, 2002; Smith & Reinertsen, 1991), on conceptual models of FEI (Brentani & Reid, 2012) and on how customers can be involved in the FEI and improve the overall NPD process (Merton, 2013). However, despite the growing number of studies on FEI in recent years, the outcomes do not provide a complete guide to effective FEI management (Akbar, 2013). Furthermore, it should also be noted that previous front-end studies did not fully account for today’s developments in context of open innovation and next generation SGM’s (Frishammar & Ylinenpaa, 2007).

The main reason for the increasing interest regarding FEI, can be found in the one single aspect scholars do agree on: the importance of FEI and its activities for the NPD process and more specifically NPD outcomes (Markham, 2013). The recognition of the FEI as a crucial and decisive phase emerged from NPD research literature on opportunity recognition and best practices (Floren and Fishammer, 2012). Wagner (2012) noted, in line with Cooper (1990), that the FEI determines largely the potential for success or failure in a NPD project. Zhang & Doll (2001, p. 95) stated on the importance of FEI: “most projects do not fail at the end; they fail at the beginning”. Authors give several reasons for the importance of FEI on NPD. Regarding the management of the FEI, Cooper (1993), Christiano, Liker, & White (2000) and Dwyer & Mellor (1991) argue that it is the most difficult stage to manage, due to its often unstructured nature and offers the largest potential for improvements. Since the FEI is the phase which requires and allows for creativity, it systematically differs in required management and organizational capabilities relative to the BEI (Akbar, 2013).

2.5 Resource allocation in NPD

(15)

15 their committed resources, thereby establishing a resource problem which results in NPD projects suffering from lack of resources on all functional areas. As few as 30% of the firms in the technological industry have enough R&D resources committed to their R&D projects and 76% of these firms have an overall poor balance between available resources and the number of ongoing NPD projects (Cooper, 2006). Additionally, 10% of these firms allocate sufficient resources to NPD projects to ensure they are executed at a quality fashion (Cooper, 2006). A tendency is that managers and decision makers favor incremental projects when confronted with serious resource constraints (Koen et al., 2002). When resources are scarce, managers take few changes and favor the sure bets which are typically smaller and of low degree of newness (Koen et al., 2002.

2.6 Resource allocation in the Front End of Innovation

Especially in the FEI, managers are often reluctant to invest large amounts of resources in projects which are still in this phase (Koen et al, 2002). This is arguably a result of the lack of information, perspective and potential that is available at that point in time. Managers and decision makers favor investing in proven technology and projects that are in latter stages of the NPD process and therefore have a certain guarantee for returns on investments (Cooper, 2006). This line of reasoning by managers is further strengthened by the fact that, in general, each stage is more expensive to complete in both monetary resources as man-hours than the preceding one (Cooper, 1990). At the same time, by passing stages, an innovation project typically also becomes easier to manage as more information becomes available, leading to a decrease in uncertainty which allows for better risk management (Sommer et al., 2015). Considering these findings, one might argue that this may be reason for the fact that in the last 18 years, innovation portfolios in high tech industries have drifted from balanced to extremely unbalanced with too many small, incremental projects and too few breakthrough projects (Cooper, 2013).

(16)

16 activities that this phase is central to new product success. Cooper (1988) concluded on FEI activities, that the investments of more dollars and more man-hours in this phase is strongly connected to success. This points to the notion that the seeds for success appear to be sown in the FEI. In more recent literature, similar conclusions have been reached. Markham (2013, p.90) found that, when taking the number of ideas that successfully transfer to formal development as measure of success, “devoting sufficient resources increase the percentage of ideas that move successfully from the front end into formal development".

(17)

17 Goldenberg et al. (2001) noted that resource-constrained innovation projects can often lead to products considered as highly innovative in the market. In contrast with the resource-driven mindset, the Constrain theory limits complacency and limits the tendency of NPD practitioners taking the path of least resistance, thereby potentially neglecting the search for novel solutions for the problem at hand (Ward, 1994). This is especially relevant in the pre-development phase of an innovation project, since in this phase requires most creativity and novelty (Wowak, et al., 2015). The main logic of this theory in an NPD context, is that having resource constrains is beneficial for the innovation process: ‘less is more’. Additionally, Moreau & Dahl (2005) found, in line with Katila & Shane (2005), that this effect is especially the case for an entrepreneurial setting, in which resource constrains force entrepreneurs to work efficient and make use of entrepreneurial ingenuity which is beneficial for performance.

2.7 Newness of a NPD project

(18)

18 implicates unpredictability and uncertainty which can lead to an increase in costs and required resources (Green, Gravin, & Aiman-Smith, 1995). Authors agree on the fact that having radical NPD project (i.e. projects with a high degree of newness) is detrimental for firm performance. Radical NPD projects and resulting new products are in some cases capable of causing a paradigm shift and can significantly alter the technological trajectory of product categories which can lead to first mover advantages and stronger competitive positions (Dosi, 1982). Although radical innovative projects are required because they often yield higher payoffs compared to incremental projects, these radical projects also entail greater risk (Cooper, 1993). This implies that both types require different management approaches (De Brentani, 2001; Song and Monotya-weis, 2001).

However, despite the fact that several authors argue that the degree of newness is a decisive factor for success, a recent meta-analyses only reports an average correlation between degree of newness and new product success of 0.07 (Szymanski, Kroff, & Troy, 2007). Overall, empirical research has not yet yielded a clear unambiguous picture of the effects of innovativeness on new product performance. There are many NPD studies that consider the degree of newness as a contingency but its effects are not fully understood (Martinsuo & Poskela, 2011). For managers it would be beneficial to know if, for instance, projects with a high level of innovativeness indeed require relatively more resources than projects with lower levels of innovativeness and more specifically, which contingent factor influence this process. Especially because these projects are responsible for a major part of the profitability of a company, but also have lower survival rates than incremental projects. (Cooper, 1994; Koen et al., 2014).

2.8 Summary chapter

(19)

(20)

20

3. Hypotheses development

In this section the theories, concepts and constructs that were reviewed are used to derive hypotheses in order to answer the research question. The conceptual model is graphically represented in figure 2.

3.1 Resource allocation

As Cooper (1990) and Zhang & Doll (2001) argued, the FEI is a critical and decisive phase in NPD. Put differently, the seeds for success are sown in the first phase of the NPD process (Cooper (1988). If the FEI has a significant effect on NPD, then a stronger front end performance should result in stronger new product performance (Smith & Reinertsen 1992). Following Cooper (1988) and the RA theory, one might conclude that investing relatively more resources in the FEI, will increase FEI performance and ultimately NPD performance. As mentioned in section 2.6, resource constrain theorists predict a negative influence of having resource slack on performance. However, the resource advantage perspective on NPD remains dominant in the literature (Daniel et al., 2004). To frame the current study and make data collection feasible, we focus solely on financial (i.e. monetary) resources and financial measures of NPD and product performance. To further expand knowledge and verification of the dominant logic of the RA theory and by answering the call for resource specific and NPD phase specific empirical research, we aim to research the expected positive relation between invested financial resources in the FEI and subsequent financial new product performance and thus hypothesize: Hypotheses 1: The relative amount of financial resources invested in the pre-development stage of an

NPD project is positively associated with financial new product performance.

3.2 Degree of newness

(21)

21 making process on allocating financial resources to the FEI. Overall, it is expected that, in the case of a NPD project with a high degree of newness, the need for sufficient (i.e. slack) resources is higher and influences the relationship between invested financial resources in the FEI and subsequent financial new product performance. We expect a moderating relationship of the level of newness on the relation between financial resource allocation and financial performance and thus hypothesize:

Hypothesis 2: The degree of newness of an NPD project positively moderates the relationship between

the relative amount of invested financial resources in pre-development and financial new product performance.

Although the degree of newness is been theorized to also have direct links to performance, we deliberately construct as moderator facture because this has a more managerial meaning trough linking both constructs with different conditions of innovativeness (Lee & O’Conner, 2003). The hypotheses are graphically represented in the conceptual model as shown in figure 2:

Figure 2 Conceptual model

3.3 Summary chapter

Resource allocation within NPD and the FEI are under-researched themes in the innovation literature. Therefore, we aim to explore the relation between committing resources in the FEI and the subsequent effects of these decisions on financial new product performance. As firms push to be highly innovative as compared to competitors, the number of radical innovation projects firms’ start is increasing. To study the effects of innovativeness, we aim to explore its expected moderating effect on the relationship between investing financial resource in the FEI and financial performance outcomes.

Relative amount of invested financial resources

in Pre-Development

Financial new product Performance H2 +

Degree of Newness

(22)

22

4. Research Methodology

Since there are several research gaps identified and there is a significant body of literature available to derive hypotheses, this study follows a theory testing approach (van Aken, Berends & van der Bij, 2012). In this section the research strategy, the process of data collection, measurement development and methods of data analyses are described.

4.1 Literature review

High quality peer-reviewed journals and papers were investigated to describe the current phenomenon of interest. Findings from these journals and papers led to the identification of the theoretical gaps and the development of adequate hypotheses described in section 2 and 3. Mainly articles in the field of NPD and innovation management were reviewed and retrieved from both the Business Source Premier Database and Google Scholar.

4.2 Sample

This study adapted a retrospective analysis of a sample of product groups and business units from a multi-national firm mainly active in consumer electronics and healthcare. The focal firm develops and launches products in different product-market categories worldwide. The accounting and reporting structure of the focal firm consists of three main divisions with each division having a number of business units (BU’s) and each BU having a number of product groups (PG’s). Each division is a branch of the overarching group corporation and operates autonomously to a certain extent. The divisions are directed by general managers, who are accountable to the top group management of the firm. Because these divisions vary in accounting, reporting and data storage processes, we ensured both consistency in the data and feasibility of the data collection process, by solely focusing on the consumer electronic division. The consumer electronic industry has adequate dynamics for this study as it is characterized by short product life cycles and new products are often introduced in the market and quickly followed-up by next generations (Coman & Ronen, 2007). Hence, as mentioned by interviewees, practitioners and experts at the focal firm, innovation is crucial within this division in order to stay competitive in global markets.

(23)

23 product level not being consistently available. A PG is often characterized by a joint production technology and its different products are sold in the same type of markets. The PG level is often reported as the smallest unit within large firms to which sales responsibility can be delegated (Yoon & Lilien, 1985). To generate additional valuable insights, BU level data was also included in the analysis. The single-industry and single-firm approach was adapted to rule out intra-industry effects that may have detracted from previous studies on NPD (Cooper & Kleinschmidt, 1994). As we ran analyzes, we treated each PG investment and its subsequent performance indicator as an individual case resulting in a total sample of 229 cases. Because sales data from PG’s in 2016 were not yet available at the time of this study, the cases from 2015 had to be removed from the sample which reduced the sample size to 172. Sample statistics are presented in table 2.

4.3 Data collection

Since there was no existing dataset available with suitable data for this study, a new dataset was created. Quantitative data was collected by contacting multiple senior executives that were responsible for finance, accounting, marketing and marketing intelligence at the focal division. Amongst other things, data consisted of research and development, sales, marketing and marketing intelligence data which is recorded in multiple enterprise resource planning, project management, accounting and reporting systems. As is the case in this research, collecting objective data is preferable since subjective data may give room for respondent biases and imperfect information reflecting human decision-making (Kleinschmidt & Cooper, 1991). PG’s in the sample had to be representative for the firm and consist of diverse innovation projects with diverse budgets, development cycle times, perceptual risk and newness levels. These requirements were used in order to construct a dataset with enough variance and measureable parameters and consists of cases from 58 product groups within 11 business units in each of the following years: 2012, 2013, 2014 and 2015.

Table 2 Sample statistics

(..)

4.4 Measures

(24)

24 interchangeable and collective developments, we chose to search for one universal used project type as proxy to determine the amount of pre-development expenditures. The expenses of projects that consisted of pre-development activities, needed for developing a rough idea, building the business-case and act as decision gate for further product development were used as pre-development expenditure measure. These pre-development activities are named ‘Advanced Development’ projects within the focal firm, (…).

4.4.1 Dependent variable

New Product Performance – Measuring new product performance is widely discussed in the literature

but no general consensus is reached yet (Daniel et al., 2004). It is difficult to define new product performance, since there are multiple aspects to consider (Montoya-Weiss & Cantone 1994). In the literature, the general division is made between operational and market outcomes (Tatikonda & Montoya-Weis, 2001). Since a perceptual assessment of performance may understate or overstate the actual performance, we used objective sales data (Henard & Szymanski, 2001). Since the innovation cycle time or budget adherence (internal indicators) are not of interest in this study, we chose to take an external perspective and focus solely on market outcomes in terms of sales on PG level as measure. In the consumer electronics market, the cycle time is relative high (Mullins & Sutherland, 1998). This implies that sales are often generated by products that are recently introduced (Ettlie & Subramaniam, 2004). However, as Blindenbach-driessen and Van Den Ende ( 2010) state, for project-based innovation structures it is difficult to determine the exact profit and revenues that are generated by the developed product. As the average total development time of a product within the focal firm product is around 1 year and detailed data on newly introduced products was not available, we made the assumption that R&D expenditures in year X0 contributed the most in sales in year X1.This method is adapted from

Brush, Bromiley, & Hendrickx (2000) and is measured as actual sales increase or decrease per PG in year X1 relative to year X0. This variable is named Sales Change.

4.4.2. Independent variable

Invested resources in development – Exclusively knowing the total investments in the

(25)

25 development and the total amount of invested financial resources within the end to end NPD process in year X0 and is named Pre-Dev Ratio.

4.4.3. Moderating variables

Degree of Newness – This study explicitly considers the degree of newness not as independent variable

that directly impacts sales performance, but as variable that moderates the relationship between Pre-Dev Ratio and subsequent financial performance outcomes. As our analysis is mainly on PG level, this entails that the degree of newness is measured on PG level and measures the amount of resources that are invested in the various categories for degree of newness. Kester, Hultink, & Griffin (2014) pointed out that new product innovativeness is most frequently used as measure of the degree of newness of a product. Booz et al. (1982) proposed six new product typologies that are often used to categorize degrees of innovativeness along two dimensions. They categorize new products along the relative newness of the dimensions market and firm. Decades later, Cooper & Kleinschmid (1995) adapted this classification to a simpler three point low, moderate and high degree innovativeness. The focal firm developed its own framework in which innovation projects are classified in different newness categories. The three classifications mainly used in the consumer electronic division are (…), we chose to assume that these classifications can, to a certain extent, be aligned with the typology of Kleinschmidt & Cooper (1991). The PG’s were categorized on an ordinal scale according to in which category of newness most (>50%) financial investments were made.

4.4.4 Control variables

To gain insights and clarify the relationships of interest, several control factors were added to the model to control for other extraneous effects on sales performance. Previous studies pointed towards several internal and environmental factors that can influence new product and thus sales performance (Lee & O'Connor, 2003; Montoya-Weiss, 1994). As the consumer electronic division of the focal firm can be viewed as autonomous entity to a certain extent, we followed Constantopoulos, Spanos, Prastacos, & et al. (2015) by also including several division level control variables. This was also due to the fact that not all control variable data was consistently available on PG level and BU level.

Division Size–According to preceding studies, the size of an organization can affect innovative and

operational performance through the fact that larger organizations typically have more resources and a larger workforce to meet goals and generate sales and revenue (Camisón-Zornoza et al., 2004; Li, 2013). Since the PG’s are all part of a relative autonomous division, we follow Li (2013) and include the division size by determining the number of employees within the divisions for each year.

R&D intensity –R&D intensity is widely used as control variable in innovation and NPD studies (Hitt,

(26)

26 several types of innovative output such as patents, new product introductions but also to sales (Hitt et al., 1997; Grandstrand & Oskarsson, 1994). Following Hitt, Hosikosson, Ireland & et al. (1991) and Tsai (2001), R&D intensity on division level was measured as the ratio of total R&D expenditures to the division’s total sales in year X0.

Marketing expenses –In previous studies, marketing expenses are proven to be related to sales

performance (Palmatier, Gopalakrishna, & Houston, 2006). Advertisement and promotion expenses could only be collected on BU level at the focal firm and are included on this level as control variable. According to several focal firm experts the absolute amount of marketing expenses. As total sales act as indicator of the strategic importance and size of a BU, we followed Kotler (1965) and constructed this variable as ratio between total advertisements and promotion expenses over total sales revenue. Furthermore, marketing literature suggests that, in contrast with R&D investments, marketing investments have the most effect on sales revenue on short term (Narver & Slater, 1990). To account for this short term effect (e.g. specific product placement and promotion at the time of market introduction), the marketing expenses for a BU in year X1 are included as control variable.

Market growth – An important control variable for sales performance is the growth or decline of the

market in terms of potential buyers (Jaworski, 1993). In order to control for this effect, we included the market growth as control variable adapted from Bharadwaj, Clark & Kulviwat (2005). Data for this variable was not consistently available on PG level. Therefore this variable was constructed by averaging the available market growth figures per year of PG’s within one single BU. Since market growth has the biggest effect on sales performance in the same year as sales performance is measured (Ronkainen, 1985), this control variable was included as measured in year X1. Hence, we must be

cautious with inferring conclusion form this specific variable since it is averaged out and thus not a perfect representation of market growth for all individual PG’s.

Business unit category – In order to control for differences in the characteristics of NPD processes and

specific market characteristics of certain BU’s (Tsai, 2001), we include BU category as control variable in our model. This is done by categorizing all the PG’s and their respective BU’s in two main groups: (..). By including a dummy coded variable in our model, we control for specific difference that may exist in these two main categories.

4.5 Methods of analyzes

(27)

27 investments and sales performance, the cases with zero R&D expenditures were removed from the sample. Zero investments could indicate that the firm is no longer investing in certain PG’s or BU’s, or it could have other unknown reasons. As this can bias the outcomes, these cases were removed reducing the sample to 148. To test the hypotheses, multiple hierarchical regression analyzes were performed. By assessing the fit of the model (i.e. variance explained) and the relative contribution of the controls and explanatory variable, hypothesis 1 can be examined. To test hypothesis 2, the mean-centered interaction effect as product term was included in the model to examine the moderating effect.

Several validity tests were performed prior to testing the hypotheses to secure the assumptions that have to be met for performing a regression analyzes and thus the validity of the results (Field, 2009). If explanatory variables have high multicollinearity, it entails that the explanatory variables have a high correlation which reduces the reliability of the results. A variance inflation factor (VIF) test with control variables included was conducted to test for multicollinearity. For each of the variables the VIF values were lower than or equal to 1.345. The average VIF for the explanatory variables was 0.98. Staying within the limit of 5, we can conclude that multicollinearity is not a problem in this sample (Sine, Mitsuhashi, & Kirsch, 2006). To check if the assumption of a linear relation between the independent and dependent variable is satisfied, the scatter plot was visually inspected. Although this gave no reason to doubt this assumption, this condition was further checked by including a quadratic term of the independent variable in the regression model. This yielded no significant change in the R-square or F-value change, implying that the assumption of linearity is satisfied. The variable Sales Change and the explanatory variables were checked for possible outliers by calculating Z-scores (standardized values). According to Tabachnick & Fidell (2001), a standardized value exceeding 3.29 can be identified as outlier. The cases which exceeded this value were removed from the dataset. As linear models assume a normal distribution of the dependent variable, the distribution of the variable Sales Change was tested. Sales Change was initially non-normally distributed (skewness Z-value of 51.80 and kurtosis Z-value of 291.22) and exceeded by far the acceptable range of -1.96 < x < 1.96 (Aiken & West, 1991). The variable also showed heterogeneous variance through a Shapiro-Wilk test (significant at p < 0.01) implying, again, a not-normal distribution and thus not compatible with the assumptions of linear regression. To counteract this, the dependent variable was transformed into a log variable to further ensure a normal distribution (Li, 2013). A constant was added to make sure negative sales changes were also transformed (Li, 2013).

(28)

28 Shapiro-Wilk test for normality showed non-significant (0,365, P > 0.05) for the dependent variable, thereby implying a normal distribution. Since normal distributions are based on continuous data, our ordinal moderator cannot be tested for normal distribution. For the independent variable Pre-dev Ratio, the Shapiro-Wilk test showed significant (P <0.01), thereby implying a non-normal distribution of these variables. The effects of possible heteroscedasticity were tested by visually inspecting the scatter plot of the regression’ standardized residual against the regression’ standardized predicted value. The residuals are approximately the same size for all values of X and shows no tendency in the error terms, implying that heteroscedasticity does not seem to be a problem (Garcia-Granero, 2002). Despite the Shapiro-Wilk test showed significant for Pre-Dev Ratio and thereby implying a non-normal distribution, the Q-Q plot suggest an approximately normal distribution. We can therefore assume that the data in the sample is approximately normally distributed and is thus compatible with regression analyses assumptions. Hence, the results must be interpreted with caution since not all data in the sample is perfectly normally distributed.

4.6 Triangulation

To increase construct validity and reliability of the results, we chose to adapt a triangulation approach (van Aken et al., 2012). This approach was chosen in order to combine two research strategies in studying the focal phenomenon. One strategy being the statistical analysis of the collected quantitative data. The other is collecting qualitative data and insights trough a focal firm expert panel. By analyzing not only quantitative data but also qualitative data, we aim to find additional insights which can yield complementary results (van Aken et al., 2012).

(29)

29

4.7 Summary chapter

(30)

30 Table 3

Overview of variables

Variable Measure operationalization Scale

Dependent Sales Change Percentage of actual sales change in year X1 relative to year

X0 on PG level (Brush, Bromiley, & Hendrickx, 2000).

Interval

Independent Pre-Dev Ratio Ratio between invested financial resources and total R&D invested resources in year X0 on PG level (Griliches, 1992).

Interval

Moderator Degree of Newness Coded in categories ranging 1-3 on PG level (Focal firm, 2016; Kleinschmidt and Cooper, 1991).

Ordinal

Control R&D intensity Ratio between total divisional revenue and divisional R&D expenses in year X0 (Hitt et al., 1991; Tsai, 2001).

Interval

Workforce size Number of employees in the division in year X0 (Li, 2013). Interval

Ratio AP expenses –total sales

Ratio between total advertisements & promotion expenses and total sales revenue per BU in year X1 (Kotler, 1965)

Interval

Market Growth Percentage of market growth or decline in terms of potential buyers per BU in year X1 (Rokainen, 1985; Bharadwaj et al.,

2005)

Interval

BU Category Main business unit category in which the product group or business unit belongs (Tsai, 2001).

Binary

Figure 3

Theory testing process*

* Adapted from Van Aken et al. (2012)

- Current phenomenon with little emperical evidence for academic explanations and predictions

- Development of research question

- Identification of constructs and important variables - Development of hypothesis

- Development of conceptual model - Large scale data collection - Statistical data analysis

- Qualiative validation and verification of the quantiative results trough expert panel and company experts

- Results - Conclusions

(31)

31

5. Results

The results of the main analyzes conducted on PG level to test the hypothesis are presented in section 5.1 through 5.3. In section 5.4, additional analyses on BU level are presented. In section 5.5, an alternative performance measure is presented in an attempt to gain additional insights.

5.1 Descriptives and Correlations

The descriptive statistics of the PG level dataset is presented in table 4. The count frequencies and distribution of the variable Degree of Newness (DN) are presented in Table 5. The descriptive statistics indicate that (…).

Table 4

Descriptive statistics product group level

(…) Table 5

Degree of Newness count frequencies (…)

The Pearson’s correlation matrix is presented in Table 6. It shows that the independent variable Pre-Dev Ratio is significantly correlated (P < 0.05) with the dependent variable Sales Change. Also, the degree of newness is (P < 0.05) correlated with Sales Change. Furthermore, workforce also seems to be weakly significant (P < 0.10) but negatively correlated with Sales Change. This is unexpected and rather counter intuitive since it suggest that an increase in workforce will lead to a decrease in sales performance. Also R&D Intensity is weakly significantly P < 0.10) correlated with Workforce.

Table 6

Pearson’s Correlation Matrix Product Group Level

(32)

32 6. R&D Intensity 0.048 0.113 0.070 -0.078 -0.471* 1

7. Market Growth -0.053 -0.143 0.004 0.154 0.105 -0.158 1

(33)

33

5.2 Hypothesis testing

Hypothesis 1 - Prior to hypothesis testing, the assumptions for regression models were tested which is

(34)

34 Table 7

Regression Analysis Hypothesis 1

Sales Change X1 Model 1 β Sales Change X1 Model 2 β Constant 6.589 6.577

Control AP-Sales Ratio X1 0.044 0.150

Workforce X0 -0.307 -0.316

R&D Intensity X0 -0.023 -0.104

Market Growth X1 -0.045 -0.004

BU Category X0 -0.051 0.652

Independent Variable Pre-Dev Ratio X0 0.308**

Model Summary F 1.611* 2.975**

R2 0.077 0.157

Adj. R2 0.029 0.104

N 103 103

* are significant at .10 level, ** at .05, and *** at .01 levels.

Hypothesis 2 - In the third model, displayed in table 8, the moderating effect as hypothesized is

(35)

35 Table 8

Regression Analyses Hypothesis 2

Sales Change X1 Model 1 β Sales Change X1 Model 2 β Sales Change X1 Model 3 β Constant 6.589 6.577 6.622

Control AP-Sales Ratio X1. 0.044 0.150 0.086

Workforce X0. -0.307 -0.316 -0.323

R&D Intensity X0 -0.023 -0.104 -0.138

Market Growth X1 -0.045 -0.004 -0.010

BU Category -0.051 0.652 0.111

Independent Variable Pre-Dev Ratio X0 0.308** 0.264**

Interaction effect Degree of Newness * Pre-Dev Ratio 0.189* Model Summary F 1.968* 3.505*** 3.134*** R2 0.075 0.153 0.188 Adj. R2 0.037 0.109 0.128 N 103 103 103

(36)

36

5.3 Post hoc analyses

As we found a weakly significant positively moderating effect of the degree of newness on the relationship between Pre-Dev Ratio and Sales Change, it would be interesting to gain further insights on what specific degree of newness, is most strongly responsible for strengthening the relationship as proposed in hypothesis 2.

Figure 4 Interaction effect

(37)

37 significant differences between the moderate and high degree of newness (P > 0.05). It is also interesting to note that the standard deviation of the moderate degree of newness group is the highest of all the groups. This suggests that in our sample, PG’s in which the majority of the financial resources are allocated to projects with a moderate degree of newness, have the greatest variety in sales performance both negative and positive.

5.4 Business level analyzes

In addition to the main analyzes preformed on PG level, there were also several analyses performed on business level. As the sample size (N= 25) is rather small, the results should be interpreted with caution. The results of this analyzes on aggregated level can potentially enrich the results nonetheless. A new sample was created from 11 business units (BU’s) with data of the years 2012, 2013, 2014 and 2015 in the consumer electronic division. As with the PG level analysis, each BU-year combination was treated as separate case. The cases of 2015 were removed from the sample due to the fact that sales data of 2016 was not yet available at the time of this study. Cases with zero R&D expenditures were also removed from the sample. This reduced our initial sample size from 44 cases to 31 cases. The variables were checked for possible outliers by calculating Z-scores and removing values that exceeded 3.29 (Tabachnick & Fidell, 2001). This reduced the sample size to 25 cases. Because the dependent variable Sales Change was initially not normally distributed, a log transformation was performed and a constant was added (Li, 2013). This yielded in a dependent variable that falls within the acceptable measures of skewness and Kurtoses, thereby implying a normal distribution. In contrast to the PG sample, the independent variable in the BU sample was also normally distributed. When the same method for labeling degree of newness was used as on PG level, it was found that all the cases fell in the low degree of newness category. Put differently, the majority of financial resources was spend on projects with a low degree of newness. This can be explained through a smoothening effect that occurs because the BU level investments are an average of the underlying PG level investments. Due to the smoothening effect, the following BU analyzes focuses solely on finding additional support for hypothesis 1.

(38)

38 between Pre-dev Ratio and Sales Change, suggesting a positive relationship. To research if there is a significant relationship between Pre-Dev Ratio and Sales Change on BU level, a hierarchical regression is performed. A relation is expected to be found since the variables on this level supposedly have a mathematical relationship with the PG level variables.

Table 9

Regression model business unit level

Sales Change X1 Model 1 Beta Sales Change X1 Model 2 Beta Sales Change X1 Model 3 Beta Constant 5.289 5.377 5.697

Control AP-Sales Ratio X1 0.191 0.235 0.398**

Workforce X0 -0.208 -0.237 -0.308

R&D Intensity X0 -0.057 0.017 -0.060

Market Growth X1 0.561*** 0.500***

BU Category X1 -0.034 0.054 0.049

Independent Variable Pre-Dev Ratio X0 0.274 0.379*

Model Summary F 3.396** 3.568** 2.475*

R2 0.472 0.543 0.331

Adj. R2 0.333 0.391 0.197

N 25 25 25

* are significant at .10 level, ** at .05, and *** at .01 levels

(39)

39 strong prediction power of Market Growth can also be seen by the drop in the R-square and F-value in model 3. The Durbin-Watson statistic of model 2 and 3 falls within critical values, implying that first order linear auto-correlation in the data is not a problem (Field, 2012). The model thus partially supports the earlier on PG level found positive significant relationship between Pre-Dev Ratio and Sales Change on BU level.

5.5 Alternative performance measure

Additionally, for additional BU performance analysis, an unconventional and rare data collection method was used in order to measure product performance in an alternative way. Customer satisfaction was adapted as alternative measure to examine the effects of resource allocation in the FEI. Two global retail websites (www.amazon.com and www.focalfirm.com) and two Dutch retail websites (www.bol.com, www.coolblue.com) were searched manually to retrieve customer reviews on products of the focal firm. All former mentioned internet web shops use a five-star system to let customers, who already bought and used the product, rate their satisfaction. This can be compared to a 5 point Likert scale (Mudambi & Schuff, 2010). A total of 38.092 reviews were collected from a total of 339 individual products that were introduced to the market in the last five years. These 339 products belonged to 31 individual PG’s. For each product, reviews were collected from as much individual websites as possible to increase reliability, yielding a weighted average per product per PG per year. This was averaged in a BU-year average which were threaded as separate cases. The conditions drafted to determine the validity of the cases are presented in table 15 (appendix A). This yielded in a final sample of 24 cases. To increase variance in the dependent variable it was transformed into a squared term (Li, 2013). For each case, the Pre-Dev Ratio was measured in year X0 and the change in customer

reviews in year X1 relative to year X0 assuming that most products developed in year X0 are launched

in year X1. To test for a normal distribution, a Shapiro-Wilk test was performed and showed

insignificant. In table 16 and 17 (appendix A) descriptive statistics and Pearson’s correlation coefficients can be found.

(40)

40 Dev Ratio and subsequent change in customer product satisfaction in terms of product rating. The low sample size could be a reason for the lack of prediction power and lack of significance.

5.6 Summary chapter

Table 10 presents an overview of the results by showing the hypotheses and if support is found for these hypothesis.

Table 10

Overview of hypothesis and their support

Hypothesis Content Support

H1 The relative amount of financial resources invested in the pre-development stage of an NPD project is positively associated with financial new product performance.

Support on PG level Partial support on BU level

H2 The degree of newness of an NPD project positively moderates the relationship between the relative amount of invested financial resources in pre-development and financial new product performance.

(41)

41

6. Expert panel

To increase the validity and understanding of the results, we used a focal firm expert panel to gather insights and information useful for interpreting the results and to place the findings within the focal firm context. In two sessions, we presented the results to six focal firm experts who all have been employed in various functions within the focal firm. During the sessions, we first presented the results. Next, the purpose of the expert panel was explained: 1. To validate the results. 2. Interpret the results in line with the experiences of the participants. 3. Identify related issues. Details on the process of the expert panel are described in section 4.7. Summarized findings of the expert panel sessions can be found in appendix B.

6.1 Main findings

Regarding the main findings, these were all validated and confirmed by the focal firm experts. They unanimously agreed on the fact that front end activities are hugely important for success in NPD and stated that (…).

6.2 Over-investing

Interestingly, regarding the found significant relationship between pre-development investments and sales growth (hypothesis 1), the participants were wondering if there exists something as “over-investing”. Relating to this, participants were also wondering "if there is an ideal ratio of

pre-development investments”. Most participants agreed that the latter is probably not the case, as the

differences between individual NPD projects are immense. The participants concluded with the notion that “this phase [the FEI] might be the most important stage in the whole NPD process”.

6.3 Degree of newness

The participants also agreed on the partial conformation of hypothesis 2. They stated that, in accordance with literature, that “more radical innovation projects need relative more resources in

terms of management attention, time and monetary funds”. Regarding the pillar classification system,

one participant argued that (…).

6.4 Marketing expenses

(42)

42 To conclude, the panel interprets the main results as follows: “investing sufficient financial resources

(43)

43

7. Discussion

Although the literature on NPD, FEI and resource allocation within NPD has seen significant development in recent years, most research is done on firm level (Daniel et al., 2004). To the best of our knowledge, this one of the first studies on resource allocation within the FEI (i.e. pre-development), which adapted a single firm and single division approach. We dived deeper than firm level, towards business unit and product group level with the main purpose of investigating the relationship between financial resource allocation in pre-development and financial performance of products that originate from NPD. After confronting the Resource Advantage theory (Hunt & Morgan, 1996) with the Constrain theory (Ward, 1994; Katila & Shane, 2005), we chose the RA theory as a baseline for the current study. The NPD process was carefully reviewed and linked to the context of Stage Gate Models and resource allocation. The findings of this study yield multiple novel and interesting insights, which will be described and put into context with previous literature on FEI and NPD in the remainder of this section.

7.1 Resource allocation in the Front End of Innovation

As hypothesized, a positive relationship was found between pre-development investments and sales performance. This result was significant, also when controlling for marketing expenses, R&D intensity, workforce size, market growth and business unit category. This implies that, when more investments are made in the FEI relative to the total investments in the NPD process, the subsequent sales performance in the following year increases. This is in line with the RA theory and argues therefore contrary to the Constrain theory and the ‘less is more’ school of thought. The latter theory puts forward the conservative argument in which not resource slack, but resource scarcity is a driver for innovation and success (Katila & Shane, 2005). It could be the case, in line with the reasoning of Morau & Dahl (2005) that the resource constrain logic is only relevant for an entrepreneurial setting which is not the case in the current study. Also, in the Constrain theory, the focus lays on creativity outcomes whereas in the Resource Advantage theory the focus lays on innovation performance outcomes (Katila & Shane, 2005). In the current context, the dependent variable and focus is on innovation performance and not creativity. Although creativity is beneficial for innovation performance, this indirect relationship between the Constrain theory and innovation performance makes the Constrain theory less applicable in this context.

Financial Resource Allocation in the Front End of Innovation