Formalization of the Front-end Phase of the Innovation Process –
Competitive Advantage or a Path to Downfall?
Jarno Poskela, [email protected]
Helsinki University of Technology, Finland
Abstract
The influence of front-end process formalization on front-end performance is currently intensively debated issue
in the new product development literature. The main line of arguments state that process formalization in
general kills creativity and leads to decreased innovativeness. However, studies that would have investigated the
effect of formalization of the front-end phase, the creative and chaotic early part of the innovation process, are
scare. Typically studies consider both front-end phase and development project phase simultaneously, thus
averaging the totally different characteristics of these two phases. This article tested the association between
front-end process formalization and perceived superiority of created product concepts. In addition, this article
tested the classical contingency hypotheses whether the task uncertainty moderates this relationship. The study
is based on exploratory factor analysis and multiple regression analysis that are used to investigate 133 frontend cases collected from Finnish industrial companies. The results indicated, opposite to the existing theory, that
front-end process formalization is associated with superior product concepts. In addition, market uncertainty
positively moderates this association, i.e. the more market uncertainty is present the more positive is the
association. Implications of results from theoretical and practical point of views are discussed.
Introduction
The foundation for successful new product development is created in the front-end phase, which refers to
the activities that take place before the formal development project phase (Koen et al., 2001). The overall
structure and the main characteristics of the future product are all decided in the front-end phase, which
then strongly affects subsequent new product development activities. Recent studies indicate that these
early front-end activities represent the most troublesome phase of the innovation process, and at the same
time one of the greatest opportunities to improve the overall innovation capability of a company (Reid and
de Brentani, 2004; Herstatt et al., 2004; Nobelius and Trygg, 2002; Kim and Wilemon, 2002; Cagan and
Vogel, 2002). The front-end phase nourishes the new product development project phase by producing new
incremental and radical product concepts. The front-end phase results in a well-defined product concept,
clear development requirements and a business plan aligned with the corporate strategy (Kim and
Wilemon, 2002). In addition, the front-end phase should result a decision on how the product concept will
be developed further. The decision could be to continue with an immediate development project or to put
the concept ‘on hold’ to wait for more suitable timing, or even to kill the initiative. However, despite the
recognized importance and great development potential of the front-end phase, e.g. compared to the
development project phase, there has still been relatively little research on the best practices related to the
front-end phase (Nobelius and Trygg, 2002; Kim and Wilemon, 2002; Koen et al., 2001). The theoretical
discussion is still hindered by general level models, vague terminology and unclear definitions (Zhang and
Doll, 2001; Koen et al., 2001).
The front-end phase has a very strategic nature since important strategic decisions related to e.g.
target markets, customer needs satisfaction, value propositions, expected product price and product costs,
the main functionalities of products, and the predominately used technologies are all made at this stage
(Bonner et al., 2002; Smith and Reinertsen, 1998; Wheelwright and Clark, 1992). These decisions
embodied in a product concept define and guide the subsequent development activities later in the
innovation process. An important activity in the front-end phase is to ensure that decisions and choices
serve the best interests of the company and fulfill its long-term strategic objectives. However, strategic
guidelines might be missing, misleading or too general to assure an efficient link between strategies and
operative level activities, thus making decisions uncertain and unsystematic. Product concepts can become
“moving targets” when there is no comprehensive strategy directing the innovation processes
(Wheelwright and Clark, 1992). Other familiar symptoms reflecting front-end failure are new product
initiatives that are cancelled half-way through because they do not match the company’s strategy, and
delayed top-priority new product initiatives that suffer from a lack of prioritization of assignments
(Englund and Graham, 1999; Khurana and Rosenthal, 1997). Furthermore, Khurana and Rosenthal (1997)
have analyzed these front-end failures and found that they emerge because senior managers do not
communicate their strategic level expectations, such as the product's core benefits, choice of market
segments, and pricing of products, to the development team. Strategic statements can also be too abstract
without giving any direction to front-end activities (Smith and Reinertsen, 1998). In general, firms seem
to want more explicit links between strategy and the new product development process (Hertenstein and
Platt, 2000). Management’s ability to influence strategic choices in product development is naturally
greatest at the beginning of the innovation process. However, the typical real involvement pattern shows
that management gets heavily involved in the initiative after the design phase has already been completed
when development problems have become visible and just when large financial commitment is actually
needed (Smith and Reinertsen, 1998; McGrath, 1996; Wheelwright and Clark, 1992). Unfortunately, the
ability to influence the outcome then without considerable and costly redesign effort is low. Management
should invest their time proactively to confirm that critical choices made in the front-end phase are
strategically feasible from the company’s point of view (Smith and Reinertsen, 1998; McGrath, 1996;
Wheelwright and Clark, 1992).
The above mentioned challenges related to front-end execution and management involvement
relate to the interesting question of how management should actually control the front-end phase of the
innovation process. The term ‘management’ refers to individuals such as R&D (research and development)
directors or technology directors, who are responsible (based on their organizational position) for assuring
that new product development activities fulfill strategic objectives and serve the long-term development
needs of the organization. The creative nature of the front-end phase makes it difficult to use a hard
command type of control, but still certain controllability is needed to secure the effective use of resources
and the achievement of the company’s long-term objectives. Thus the critical question is how to control the
front-end phase of the innovation process while simultaneously maintaining the innovativeness and
assuring the company’s long-term objective achievement. However, contingency theorists and many others
have acknowledged that the degree of task uncertainty influences the optimal way of organizing
management processes (see e.g. Donaldson, 2001; Tidd et al., 2001; Burns and Stalker, 1966). Thus
innovations including different degrees of task uncertainty, e.g. incremental or radical product innovations,
probably need different control approaches. This leads to another important question of how task
uncertainty in incremental and radical innovations influences the applicability of different control
mechanisms.
The goal of finding appropriate means to balance control and creativity is not a new issue. There
are several studies that have raised this important question in management literature in general (see e.g.
Marginson, 2002; Simons, 1995) and in the innovation and new product development (NPD) context (see
e.g. Bonner et al., 2002; Tatikonda and Rosenthal, 2000; Park, 1998; Brown and Eisenhardt, 1997;
McGrath 1996). One challenge of interpreting sometimes conflicting results of existing management
control research in the NPD context, is due to the fact that these studies have investigated NPD projects as
a whole, without considering the characteristics of different phases of projects, e.g. differences between the
front-end phase and the development project phase. As several studies have shown, the nature of these
phases is totally different in terms of task characteristics and people involved (Kirsch, 2004; Koen et al.,
2001; Nixon, 1998; Zien and Buckler, 1997). The front-end phase shows characteristics of high uncertainty
and ambiguity, while the development project phase shows characteristics more of formality and certainty.
Reasoning goes further by arguing that because of the different nature of the phases, they should be
managed differently as well. This leads to the interesting notion of whether different control mechanisms
are generally applicable in different phases of the innovation process. Research on management control of
information system development projects give indications that types of management control and control
mechanisms change when the initiative proceed from the idea stage towards commercialization (Kirsch,
2004; Choudhury and Sabherwal, 2003). Results indicate that simple output based controls are preferred
over behavior control at the beginning of projects (Choudhury and Sabherwal, 2003). An other study
revealed that informal control modes dominated over formal methods in the requirements definition phase
of information system projects whereas formal methods were taken into use in the implementation phase
(Kirsch, 2004). It may be justifiable to wonder whether any type of control is appropriate i.e. has a positive
effect on performance in the highly uncertain and even chaotic front-end phase. Critical question from the
practical point of view is whether there is a certain limit (measured in terms of uncertainty) where different
types of control becomes ineffective. The importance of studying different phases of the innovation process
separately has been discussed by Davila (2000), Olson et al. (1995), and Kirsch (2004).
The author’s own notion is that industrial companies are currently intensively developing
systematic approaches to manage and control the front-end phase. Qualitative studies indicated that many
of those development interventions focused on creating some kind of stage-gate model (formalized process
model) for the front-end phase. A formal process model is a key mechanism to control front-end activities,
and is typically associated with creation of formal reporting channels, review and evaluation procedures,
and decision gates. Current theoretical understanding is in line with stated concerns of practitioners, who
were afraid of the possible influence a stage-gate process may have on innovativeness in the front-end
phase.
This article aims at helping to understand how management can control the front-end phase of
the innovation process in a product innovation context. Process formalization as one type of management
control mechanism is taken into closer investigation. Two research questions have been set for this paper:
1. How is front-end process formalization related to the front-end performance?
2. How does task uncertainty influence on the relationship between the process formalization and the
front-end performance.
Management control and front-end process formalization
Management control has been stated to be an important aspect of organizational design
(Eisenhardt, 1985), fundamental management activity (Jaworski, 1988), critical activity for organizational
success (Merchant, 1982), and also a central feature of all human organizations (Otley and Berry, 1980).
Merchant (1982) argues that control should especially be directed to strategically important areas in
organizations such as NPD. The traditional 1970s and 80s view of control emphasized managerial actions
as confirming that activities conform to existing strategic plans. The present understanding of management
control sees it as a function of divergent requirements between creativity and innovativeness, and intended
goal achievement (Simons, 1995). Simons discusses management control systems as “…the formal,
information-based routines and procedures managers use to maintain or alter patterns in organizational
activities” (Simons 1995, p. 5). This definition covers both top-down induced and bottom-up emerged
strategies. Simons emphasizes that the competitive pressure created by senior management is a catalyst for
innovation and adaptation. Thus traditional command type, top-down oriented control is no longer
sufficient. In addition to the top-down information flows and commands that inform lower level employees
about the organization’s intended strategies, there needs to be channels transferring information from the
bottom of the hierarchy to the top. Through these channels the top management receives information about
progress in achieving intended strategies and also information about threats and opportunities that may
contain seeds of new emergent strategies. (Simons, 1995)
The theoretical control framework of this paper is based on Hales (1993) who separates four
dimensions of control: 1) focus of control, 2) level of formality of control 3) level of interactiveness of
control, and 4) locus of authority of exercising control. The first dimension, focus of control, categorizes
management control by placing control practices in a chronological order based on the actual sequence
when the control is implemented. This leads to the following categories of management control: input,
process, output and value. Input control occurs before the controlled activity. Instructions, materials, and
the knowledge and skills of those carrying out the forthcoming work are the main objects of the control.
Process control, in turn, is exercised during the activity focusing on work processes and technical work
methods of the controlled employees. Output control takes place after the activity and focuses on outputs,
material, information or financial results. Finally, value control influences the activities all the time by
affecting the planning, implementation and evaluation of work activities. Value control is a kind of meta
control, which is based on the influence of beliefs and norms of the company. (Hales, 1993)
Management control can also be classified in formal and informal ways of implementing control
(the second dimension in Hales’ framework). Jaworski (1988 p. 26), who studied control in marketing
units, defines formal controls as “written, management-initiated mechanisms that influence the probability
that employees or groups will behave in ways that support the stated [marketing] objectives”. Informal
controls, conversely, are “unwritten, typically worker-initiated mechanisms that influence the behavior of
individuals or groups in [marketing] units”. Many of different control mechanisms can be applied either
informally or formally. Management control can also be applied either in interactive/personal or
bureaucratic/impersonal ways (the third dimension in Hales’ framework) (Hales, 1993; Bonner et al., 2002;
Simons, 1995; Fisher, 1995). Interactive control means that managers have personal contact with the
decision making activities of their subordinates (Simons, 1994). Hales emphasizes that personal control
manifests that control is exercised by one individual over others, whereas impersonal control is based on
rules and regulations (Hales, 1993). The locus of responsibility for implementing the control may also be
possessed by different parties within the organization (fourth dimension). The control may rest in the hands
of individuals (self-control), a group of colleagues (mutual, peer or clan control) or a body which is
separated from the work process itself (external control) (Hales, 1993). The latter case refers to traditional
top-down implemented control.
Organizational control has traditionally been based on the use of two means of control: output or
process (action or behavior control) (see e.g. Ouchi, 1979; Merchant, 1982; Eisenhardt, 1985; Jaworski,
1988; Hales, 1993; Simons, 1994; Ramaswami, 1996; Abernathy and Brownell, 1997; Bonner et al., 2002;
Marginson, 2002). The basic difference between these control types is that process control focuses on work
procedures and processes during the controlled activity, whereas output control focuses on the end results
of a certain activity after the event. Merchant (1982) uses the term “action control” to mean different ways
of controlling the actions that individuals in the organization are performing. According to Merchant, there
are three basic types of action control: 1) behavioral constraints e.g. segregation of duties prohibiting
improper activities; 2) action accountability including definitions of limits of appropriate behavior,
monitoring activities, and rewarding or punishing deviations from the acceptable limits, and; 3) pre-action
reviews in the form of direct supervision, formal planning reviews or expenditure approvals. In the case of
complete process control, management holds employees responsible for following the established process
guidelines and work instructions, and not responsible for the potential outcome of the specific activity
(Jaworski, 1988). Ouchi (1979) stated that behavioral control is appropriate in situations of high task
programmability and low outcome measurability, and the outcome control in the opposite situation. When
task programmability is perfect and outcome measurability is high, the organization has the option to use
either behavioral or outcome control. The organization then chooses the control mode which is the most
cost efficient. Behavioral control is typically preferred over output control if the means-ends relationships
are known, because of the real-time operating nature of behavioral controls which gives accurate control
information during the activity (Ouchi and Maguire, 1975). Eisenhardt (1985) states that an increase in task
programmability, the possibility of behavior measurement, the cost of outcome measurement, and outcome
uncertainty, lead to favor behavioral control. One critical precondition for the use of behavioral control is
that the employees under control must really know what kind of behavior is expected from them (Merchant,
1982).
Process control and especially process formalization in the front-end means specifying
procedures to be followed and monitoring that work activities are proceeding in accordance with the
defined procedures. Management aims at ensuring that those activities that are considered necessary and
critical for the success are thoroughly accomplished. In addition, management arranges review and decision
points during the processes and establishes reporting procedures in order to be kept informed about the
progress of front-end initiatives. The effect of process formalization on front-end performance in the
presence of high task uncertainty is hard to predict due to many conflicting findings. The organization
control literature states that critical pre-condition for process control and process formalization is that the
appropriate work process leading to the desired end results needs to be known (Ouchi, 1977 (Ouchi uses
the term knowledge of transformation process); Eisenhardt, 1985 (Eisenhardt uses the term task
programmability)). Thus routine, structured and independent tasks are suitable for instituting formal
process control. This is also the essence of classical contingency theory and the distinction between
mechanistic and organic structures (Burns and Stalker, 1966). The increase in task uncertainty should cause
reduction in formalization and an increase in decentralization (Donaldson, 2001). Lawrence and Lorsch
(1967) were among the first to link this causality into performance. They found that the situation (e.g. a
research lab) where high task uncertainty was associated with low formality and low centralization led to
higher performance.
The front-end phase, being an experimental and even chaotic endeavor, is not so fertile ground
for process control and its formalization based on the above arguments. However, the widely referred new
product development text books for practitioners give some indication that new product success may be, at
least partly, dependent on existence and efficiency of the defined, formal front-end process model (see e.g.
Cooper, 1998; Wheelwright and Clark, 1992). The literature provides several process models to decrease
fuzziness and increase systematic approach and manageability of the front-end phase (see e.g. Cagan and
Vogel, 2002; Nobelius and Trygg, 2002; Koen et al., 2001; Cooper, 1998; Khurana and Rosenthal, 1998;
McGrath, 1996). The Stage-Gate model is one of the most linear and formal process models presented to
manage the front-end phase. Copper (1998) has introduced a model for the front-end phase including three
phases (idea generation, preliminary investigation and business case preparation) and three decision gates.
An opposite process model, i.e. the non-linear and iterative process model, is a new concept development
model developed by Koen et al. (2001). The model consists of three key building blocks: a) five front-end
elements, b) the engine which is fuelled by leadership and innovation culture, and which nourishes and
gives power to the front-end elements, and c) external influencing factors such as organizational
capabilities, business strategy, and the enabling science. The front-end elements or activities included in the
model are opportunity identification, opportunity analysis, idea genesis, idea selection, and concept and
technology development. In addition to linearity, the level of formality can be used to categorize different
process models. Khurana and Rosenthal (1998) state that the formal approach includes implementing an
explicit and widely known process with clear decision making responsibilities and specific performance
measures. A more informal method is the culture-driven approach, which aims to assure that important
front-end issues, e.g., strategic vision, technical feasibility, customer focus, schedule, and coordination are
always on the minds of all key participants. Decision making structure in the form of decision gates or
review points is typically defined together with the front-end process model. Tatikonda and Rosenthal
(2000) point out that periodic reviews are important especially for senior management providing a time and
place for intervene and giving guidance regarding project decisions. The existence of specific review points
decrease also the probability that senior management involves hands-on, i.e. too deeply, in operative
decision making. The right timing and existence of adequate information to make decisions in these review
points are of importance (McGrath 1996). The process model is also associated with definition of reporting
hierarchy inside the organization. Simons (1995) discusses managers using monthly updates and exceptions
reports as diagnostic control mechanisms. These reports are used to confirm that no unpleasant surprises
emerge from the organization. Internal reporting is one of the basic functions that information systems are
designed to do in organizations.
Recent research has criticized the current management approaches because they adopt one single,
optimal model for the front-end without considering any contextual factors, e.g. differences between
incremental and radical innovations. For example, the study of Nobelius and Trygg (2002) showed that
front-end processes differ regarding performed activities and task sequences, as well as relative time
duration and perceived importance of specific tasks. The findings indicate that the definition of the frontend process model, which is applicable for all kinds of pre-project phases, is questionable. Buggie (2002)
has presented strong criticism against stage-gate types of models stating that they are not NPD models at
all, but more like general project management models which can be used only to control milestone
achievement. The most crucial fault of this kind of model is that its decision gates focus on searching for
‘fatal flaws’ of new initiatives, thus especially excluding many radical ideas. However, there is also some
evidence that a formal process in the front-end can lead to improved and faster decision making as well as
to more successful products (Koen et al., 2001; Montoya-Weiss and O’Driscoll, 2000; Khurana and
Rosenthal, 1998).
Naturally, the process formalization brings several advantages. Ability to focus, possibility for
replication and learning, and improved coordination and integration are typical advantages associated with
process formalization (Bonner et al., 2002; Tatikonda and Rosenthal, 2000; Hertenstein and Platt, 2000).
Process formalization provides a sense of structure and clear sequence of activities reducing uncertainty
regarding the work tasks. Defined processes provide both motivation and sense of accomplishments as well
as require employees continuously evaluate whether they are in the right track. In addition, formalization
helps to achieve more efficient coordination and cross-functional communication and may enhance a
feeling of collectiveness among the development group. (Tatikonda and Rosenthal, 2000) Hertenstein and
Platt (2000) state that not only do formal models and documentation enable the replication of process but
they also help management to monitor the process and to improve it when needed. Process formalization
enables both the management and employees to focus on the most critical development issues while
implementing the predefined processes. However, the above mentioned authors do not make distinction
between the front-end and development project phase in their studies. The existing studies have also
identified several disadvantages of process formalization such as decreased innovativeness, increased
corner cutting activities, negative attitudes among employees, excess bureaucracy, and decreased flexibility
(Bonner et al., 2002; Tatikonda and Rosenthal, 2000; Hertenstein and Platt, 2000; Amabile, 1998;
McGrath, 1996). Amabile (1998) states that granting a choice over applied work processes fosters
creativity by increasing employees’ sense of ownership and intrinsic motivation. Free choices regarding the
process allow employees to maximally utilize their substance expertise and creative thinking skills.
Ramaswami (1996) warns that excessive process formalization may actually lead to dysfunctional behavior
among employees. Excessive formalism may also result inefficiency, inflexibility and heavy bureaucracy
e.g. when required approvals are acquired for operative level decisions (McGrath, 1996). New product
development process formalization has negatively been related to project performance (Bonner et al., 2002;
Abernethy and Brownell, 1997) Bonner et al. (2002) found that formal process control was negatively
related to project performance. Process formalization led to delays, cost overruns, lower product
performance, and lower team performance in projects ranging from incremental improvements to radical
new products. Again, the above mentioned authors investigated a development project as a whole without
considering special characteristics of its phases.
However, since the existing research gives somewhat conflicting results of the applicability of
process formalization in terms of performance in a new product development project phase, it is believed
that the front-end phase including even more uncertainty is not suitable for process formalization. The
negative consequences of process formalization are more likely to overcome the potential advantages of
formalization. Thus the following hypotheses are created:
• H1: Front-end process formalization is negatively associated with a superiority of product
concept.
• H2: The more market uncertainty, the more negative the association between process
formalization and a superiority of product concept.
• H3: The more technology uncertainty, the more negative the association between process
formalization and a superiority of product concept.
Research method
The sample companies were derived from BlueBook database (TDC Hakemistot Oy Blue Book,
http://yrityshaku.bluebook.fi), which includes information of all the Finnish industrial companies. The
sample companies were derived from the database by using two selection criteria: 1) companies have more
than 50 employees, and 2) companies carry out product development activities. Different business units of
50 biggest Finnish companies (based on turnover figures in 2004) fulfilling the above criteria were also
included in the research. In total, 888 companies (company in this context refers also to different business
unit of 50 biggest Finnish companies) fulfilling these criteria were found from the database. The
questionnaire were sent to all these companies, i.e., to the whole population in December 2005. The
questionnaire was addressed to R&D Director, Research director, Technology director, CEO or R&D
responsible person in each company. These titles were considered as key informants with a purpose to find
a director/person who participates in controlling individual new product development initiatives in the
front-end phase from management’s point of view. The respondents were requested to select the last
completed front-end case and base their answers on that in order to avoid success bias.
The survey questionnaire was eight pages long and divided in two parts. The first part focused on
the background information of the company. The second part focused on the example front-end case itself,
which was a unit of analysis in this study. The questions covered different control mechanisms
(independent variables), front-end performance measures (dependent variables) and also some contextual
information regarding the front-end case. Before sending the questionnaire was tested both with academics
and practitioners as suggested by Fowler (2002). The mailing process included three separate contacts to
the company representatives. First contact was a mail consisting of a cover letter emphasizing the
importance of the survey, response instructions, the eight-page questionnaire, and a pre-paid return
envelope. Three weeks after first mailing, second contact was taken by an e-mail to non-respondents as
suggested by Dillman (2000). The final third contact was taken by a phone to the randomly selected 50
non-respondents.
Of these 888 companies, 137 returned the filled questionnaire, which leads to the response rate of
15.8 %. The response rate can be considered as acceptable in the light of the long eight-page questionnaire
and the fact that the questionnaire was targeted towards the director level where the time resources are
always scare. The final useable sample for statistical testing was 133. When a survey relies on the
responses of a single informant, special attention should be paid that the informant is knowledgeable in
survey domain (Campbell, 1955; John and Reve, 1982). The great majority of respondents (91.8 %) had
one of these expected positions to whom the questionnaire was sent. The respondent had 5,7 years
experience (range: 0-30) in their position in average and 12,8 years experience (range: 0-40) in the
organization in average.
Only 1.56 % of data of used measurement items were missing, which indicates that the returned
questionnaires were completed thoroughly. The missing values were visually inspected to find possible
patterns of missing data. However, not such patterns were found. Mean substitution was used to replace
missing values (Hair et al., 1998). The influence of mean substitution to final results was checked and
found to be non-existing. The response rate in this study was 15.8 %, which gives a reason to study a
possible response bias. One method to investigate the response bias is to compare early and late
respondents of the survey. Armstrong and Overton (1977) have suggested that late respondents, who
responded because of the increased stimulus, are relatively similar to non-respondents. Possible response
bias was analyzed by testing a difference in turnover, number of employees and R&D intensity (% of
turnover to R&D) between early (63 companies) and late (70 companies) respondents. No statistically
significant differences were found between early and late respondent groups. The results indicate that
response-bias is not a problem in this study and the sample can be considered to be representative of the
target population. Herman’s one-factor test was used to analyze common method variance (Podsakoff and
Organ, 1986). All the interested independent variables were entered in the factor analysis simultaneously.
This resulted in 6 independent factors as expected. In addition, first general factor accounted only 23.45 %
of the covariance of independent variables. This gives some indication that common method variance is not
a serious problem in this study.
Multitrait-multimehtod matrix analysis was done to assess convergent and discriminant validity
of measurement constructs (Campbell and Fiske, 1959). A good convergent validity exists if withinconstruct correlations are statistically significant. The inter-item correlations generally exceeded the
threshold value .30 (Hair et al. 1998) indicating a good convergent validity. A good discriminat validity
requires that there is a small number of cross-construct correlations that exceed within-construct
correlations. All the items with one exception loaded .30 or lower to other than a primary factor in the
factor analysis resulting in a good discrimnant validity.
Analysis methods
Two main statistical methods were used in this study. First, the exploratory factor analysis was applied to
test validity and undimensionality of the created measurement constructs (Hair et al., 1998). Exploratory
factor analysis was favored over confirmatory factor analysis, since the verified management control
measurement constructs applied in the front-end context are scare. Further, Cronbach’s inter-item
coefficient alpha was measured for each factor to evaluate the reliability of the measurement construct.
Second, a multiple regression analysis was used to test the created hypothesis (Hair et al., 1998). The
appropriateness of empirical data (such as a normality of residuals) was tested to investigate that multiple
linear regression analysis can be applied (Hair et al., 1998; Cohen and Cohen, 2003). Predictor value
centering was used to overcome problems of multicolinearity while investigating the moderating effects of
task uncertainty (Cohen and Cohen, 2003).
Measurement constructs
This study applies existing, validated measurement constructs as much as possible. However, there are not
so many empirical quantitative studies that would have investigated management control in the front-end
phase of the innovation process. Thus, the author needed to create new measurement constructs. Two
principles for creating new measurement constructs were applied. First, the new measurement construct
was based on modification of existing and validated measurement constructs from the other contexts, if the
close proxy was found. Second, when the new measurement construct was created from the scratch, it was
based on extensive literature analysis and tested with both academics and practitioners. The measurement
of the dependent variable and moderating variable “uncertainty” was based on the Likert scale from one to
five (1 = strongly disagree, 5 = strongly agree). Independent variables (other than “intrinsic task
motivation” and “influence of strategic vision” constructs, where the Likert scale was used) were measured
in the scale one to five asking respondents to judge the intensity to which extent different control
mechanisms were used in a particular case (1 = not at all, 5 = used in a great extent).
Process formalization measurement construct was created based on the extensive literature
review of different process control mechanisms used in new product development and front-end context.
The first measure concerned the use of a reporting system informing the management about the progress of
the front-end case. This kind of status reporting has been regarded as an important diagnostic control tool in
the literature (Simons, 1995; Cleland and King, 1975). The second item measured the extent to which the
front-end case was executed in accordance with the defined process model. The measure was derived from
discussion emphasizing the importance of specifying the overall structure and procedures in new product
development context (Bonner et al., 2002; Ulrich and Eppinger, 2001; Hertenstein and Platt, 2000;
Tatikonda and Rosenthal, 2000). The third measure focused on the existence of specific evaluation gates
during the front-end. These review points enable the management to consider the progress of the case and
to make decisions about appropriate direction as well as continuing the case (Davila, 2000; Tatikonda and
Rosenthal, 2000). Finally, the fourth item measured used direct supervision over the procedures used by the
front-end group. This measure was adopted and modified from Ramaswami (1996) but modified to the
context of this study. The Cronbach’s inter-item coefficient alpha for this construct is .79 indicating a good
reliability.
From the vast amount of different control mechanisms, the following six other control
mechanisms were adapted to this study: input control, output based rewarding, influence of strategic vision,
intrinsic task motivation of the development group, informal communication, and involvement in goal
setting. Since the primary focus in this study is on process formalization, these six constructs are
introduced as covariates in the regression analysis.
Much of the discussion of measuring product concept superiority (a dependent variable) is
adopted from Cooper (1994), who studied over 1000 new products and their development process with the
aim of finding drivers of successful product innovations. A product including unique attributes, superior
price/performance characteristics, and high customer satisfaction has greater chances for success in the
markets. Measures for this product concept superiority construct were collected and modified based on
variables used by Cooper (1994), Griffin and Page (1996) and Song and Montoya-Weiss (2001) who used
these measures product development project context, and especially by Herstatt et al. (2004) and
Kleinschmidt et al. (2005) who applied these measures in studying front-end performance. Product
superiority construct is consisted of five measures, two of them dealing product’s comparative position to
the competitors’ products, one concerning the potential competitive advantage created by the product, and
two measures related to the impact on customers. The variables were measured with a five point Likert
scale. Overall, this measure is found to have a reasonable reliability (alpha = .69).
Uncertainty was used both as a control variable and as a moderating variable in multiple linear
regression analysis. Classical contingency theory considers uncertainty of being one of the main factors
influencing optimal way of organizing work activities. There are two main factors defining uncertainty in
the product innovation context: applied technology and aimed target market (Tidd et al., 2001; Danneels
and Kleinschmidt, 2001). 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 (2002) emphasized that product
innovativeness (the uncertainty that product includes) must be evaluated from two different perspectives:
macro-level industry perspective and micro-level company perspective. First two measures both in market
uncertainty and technology uncertainty constructs reflects this notion. These measures were modified to fit
the context of this study from Danneels and Kleinschmidt (2001) who used these measures in the market
familiarity and technological familiarity measurement constructs. The third and fourth measures in both
constructs relate to the discussion of whether the new products can rely on firm’s existing technological
and marketing competencies or not. This is an important measure of uncertainty in this study since products
with a closer fit to existing competences of the firm tend to be more successful in average (Danneels,
2002). The third and fourth measures in market uncertainty construct, and the third measure in technology
uncertainty construct were modified from Danneels and Kleinschmidt (2001). The fourth measure in
technology uncertainty construct was created and found to be functioning adequately, based on the
discussion of Danneels and Kleinschmidt (2001). Two different factors with clear factor solution and high
loadings were found as expected. The Cronbach’s inter-item coefficient alpha for the market uncertainty
construct is .76 and the technology uncertainty construct .84. The variables were measured with a five point
Likert scale.
Several control variables were included in the regression model to take into account of the
potential effects of the firm, industry, and the case itself to the final results. Control variables for firm-level
effects included the size (logarithmic transformation of turnover), R&D intensity (logarithmic
transformation of percent of turnover invested in R&D) and Front-end intensity (existence of a separate
front-end group). Industry-level effects were considered by using industry sector as a dummy variable. In
addition, the objectives set for a front-end case were controlled (short term vs. long term).
Results
The results indicated that approximately one third of the investigated companies had defined front-end
process in quite detail (TABLE 1). Only 11.5 % of respondents indicated that there was no defined process
model for the front-end. However, almost half of the respondents (49.5 %) stated that the front-end process
is only superficially defined.
TABLE 1: FRONT-END PROCESS FORMALIZATION
Level of detailness of
front-end process
model definition
Not defined at all
Defined superficially
Defined quite detailly
Defined very detailly
Total
Number of
companies
15
65
44
7
131
Percent of nonmissing
11.5 %
49.6 %
33.1 %
5.3 %
100.0 %
Correlations and reliability statistics of the used measurement constructs are presented in TABLE 2. In
general, the used measurement constructs indicated good reliability. The highlighted area in TABLE 2
indicates the correlations between process formalization and other measurement constructs. Process
formalization and product concept superiority has a positive correlation (p < .01)
TABLE 2: CORRELATIONS AND RELIABILITY STATISTICS
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
Input control
Process formalization
Output-based rewarding
Influence of strategic vision
Intrinsic task motivation
Informal communication
Involvement in goal setting
Market uncertainty
Technology uncertainty
Product concept superiority
* p < .05; ** p < .01
Items
4
4
3
1
3
3
3
4
4
5
α
.79
.79
.76
.74
.91
.73
.76
.84
.69
1
1
.41**
.21*
.19*
.13
.31**
.13
.00
.15
.15
2
3
4
5
6
7
8
9
10
1
.26**
.10
.11
.26**
.19*
-.03
.04
.23**
1
.00
.25**
.12
.16
.17
.18*
.03
1
.20*
.15
.27**
.09
.03
.22*
1
.11
.18*
.05
.05
.13
1
.25**
.18*
.10
.05
1
.16
.02
.05
1
.36**
.12
1
.12
1
The hypotheses were tested by regressing the front-end performance variable on all the management
control variables, control variables and interaction terms. The first hypotheses (H1) predicted the
association between process formalization and product concept superiority (negative association). The
results in TABLE 3 show that process formalization is positively associated with product concept
superiority (Model 1) (beta =.203, t = 2.02, p = .046). Thus the hypothesis H1 needs to be rejected, but
actually the opposite hypothesis gets support. Model 2 and model 3 in TABLE 3 tested the moderating
effect of market uncertainty (H2) and technology uncertainty (H3) respectively. When the interaction term
(process formalization x market uncertainty) is inserted in the regression model (model 2), the F-value
changes significantly (F = 3.523, Sig. of F change = 0.01). The interaction term itself is also statistically
significant (beta =.290, t = 3.582, p = 0.01). The significant change of F-value indicates that market
uncertainty indeed moderates the association between process formalization and product concept
superiority (Hair et al., 1998). However, the hypothesis H2 needs to be rejected, since the association is
more positive, not more negative, as stated in the hypothesis. The model 3 tests the hypothesis H3, whether
the technology uncertainty moderates the association between process formalization and product concept
superiority. F change is not statistically significant (F = 2.480, Sig. of F change = 0.55) and thus the H3
needs to be rejected.
TABLE 3: REGRESSION COEFFICIENT ESTIMATES
Variables entered
Input control
Process formalization
Output-based rewarding
Influence of strategic vision
Intrinsic task motivation
Informal communication
Involvement in goal setting
Market uncertainty
Technology uncertainty
Model 1
Model 2
Model 3
-.07
.20**
-.09
.23**
.14
-.04
-.01
.11
.06
-.07
.17*
-.08
.21**
.14*
-.09
.00
.11
.11
-.07
.21**
-.01
.23**
.14
-.04
-.01
.12
.06
Process formalization x Market uncertainty
Process formalization x Technology
uncertainty
R2
Adjusted R2
F
Sig. of F change
.29***
.05
0.25
0.16
2.636
0.33
0.23
3.523
0.01
0.26
0.15
2.480
0.550
* p ≤ 0.10; ** p ≤ 0.05; *** p ≤ 0.01
Standard coefficient betas are shown
Dependent variable: product concept superiority
Control variables are not shown
Discussion
The fact that all existing theory-based, created hypotheses were rejected, and two opposite findings
emerged, gives indications that existing theory related to management control in the front-end phase is far
more than complete. The results indicate that process formalization in the front-end is positively associated
with superior product concepts. It seems that advantages created by process formalization overcome those
potential challenges created by formalization. That is, potential decrease in innovativeness or intrinsic
motivation discussed by Amabile (1998) is compensated by increased ability to make systematic and
coordinated decisions. It seems that process formalization is a mechanism to make the front-end phase
more systematic and manageable. The findings that Tatikonda and Rosenthal (2000) made in new product
development context in general, holds also in the front-end phase. Process formalization gives an overall
framework in accordance with the front-end activities are executed. This evidently reduces uncertainty and
ambiguity of front-end tasks. One critical advantage of process formalization is that it enables continuous
improvement of work processes (Hertenstein and Platt, 2000). Replication of the process over again reveals
shortcomings and inefficiencies that can be removed through process re-design. Executed front-end cases
provide benchmarks that enable learning and collection of best practice data base. This is what McGrath
(1996) meant by stating that without a defined process, each group makes the similar development work
differently resulting slower process and repeated mistakes. The results showing negative relationship
between process formalization and project performance (e.g. Bonner et al., 2002; Abernethy and Brownell,
1997) need to be critically evaluated in the light of present findings.
However, several formal process models exist and they are not alike. Different process models
can be presented as a continuum, one end being a linear stage-gate process (Cooper, 1998) and the other
end being a non-linear, iterative process model (Koen et al., 2001). Formality of process decreased while
approaching non-linear models. Even though the findings presented in this paper indicated that process
formalization is associated with superior product concepts, they fail to give instructions on how the frontend process should be formalized. It can be assumed that strict formal processes (such as a stage-gate) work
best for incremental innovations by optimizing efficiency (see e.g. Kim and Wilemon, 2002), while a less
structured iterative approaches leads to more innovative results (see e.g. Benner and Tushman, 2002).
However, as the results indicated, any type of model is better than no model. In general, a front-end process
model should be flexible and support the inherent differences and iterations of the front-end, reflecting the
ideas presented e.g. by Koen et al. (2001). Companies should develop one state-of-the-art front-end process
model which is flexible enough and customized for each development initiative. The front-end process
model should give a general frame of working, provide necessary tools, instructions and templates for
efficient work, enable appropriate control and coordination, and support effective and efficient front-end
experimentation.
The most interesting finding from theoretical point of view is probably the moderating effect of
market uncertainty on the association between process formalization and product concept superiority. The
results indicated that under high market uncertainty the positive relationship between process formalization
and product concept superiority is even stronger. This can be explained through improved coordination and
communication that process formalization enables. Ulrich and Eppinger (2001) emphasize that defined
model clarifies the roles of different functions and enables different functions to bring their competence and
knowledge to the development effort in a timely manner. In addition, the model typically specifies how and
when different parties should communicate with each other. Further, they emphasize that front-end phase is
the phase where the coordination of different expertise is the most essential. The empirical findings support
this. Collaboration between sales & marketing function and R&D function is of importance. The expertise
of sales and marketing function needs to be available to reduce uncertainties related to customer needs,
target markets and overall profitability of developed products. The fact that the similar positive moderating
effect was not found under technology uncertainty, further support this notion. When only technology
uncertainty is high (e.g. new technology is applied in existing product that is aimed at existing target
markets) R&D function itself is more capable of bringing a new product concept for the development
project phase. The need for coordination and collaboration is not so high. Improved coordination and crossfunctional communication enabled by process formalization may also enhance a feeling of collectiveness
among the development group (Tatikonda and Rosenthal 2000). Improved atmosphere and group feeling
naturally have positive influence on final results. In addition, process formalization brings two other
advantages. First, process formalization helps to assure that no critical activity (e.g. concept testing or price
sensitivity analysis in high market uncertainty situation) is passed. Clearly defined development process, if
appropriately designed, helps to assure that each step is thoroughly accomplished resulting high quality end
results (Ulrich and Eppinger 2001). Second, defined process model enables managers to focus on the most
critical development issues and to trust the guidance given by the model in trivial issues.
Contribution to practitioners can be summarized in three recommendations. First, practitioners
are advised to develop a formal process model for the front-end. This has a critical influence on front-end
success in general, and especially in the situation of high market uncertainty. However, practitioners should
be informed that there are other alternatives, beside the stage-gate model, that may provide better support
and controllability of front-end activities. Second, practitioners should build appropriate decision gates or
review points (e.g. amount and timing of gates, persons in a decision board, agenda of decision process,
information requirements for decision making) for the front-end. These gates enable the management to
steer the front-end initiatives without too deeply involve in operative level decision making. Steering
groups that can be typically found from the development project phase, are needed already when product
concepts are created. Last, practitioners are advised to build a reporting system, which informs
management of ongoing front-end initiatives. This kind of reporting system should ensure that potential
shortcomings or challenges in critical decisions are found early enough. In addition, the system should
enable to find those ideas that contain promising opportunities and seeds for strategic renewal.
There are some limitations that should be noticed while interpreting the findings and continuing
the research related to management control and the front-end phase. This study has three main limitations.
First, independent and dependent variables were answered by the same person. This may naturally have
influence on objectivity of evaluation of dependent variables. Second, this study evidently includes some
success bias since the respondents were asked to select the last completed front-end case as an example.
This obviously excludes those development cases that were cancelled already in the front-end phase. Third,
it can be always speculated whether the respondents were capable of separating the control mechanisms
used in the front-end phase and the subsequent development project phase.
Conclusions
The critical mission of front-end process formalization is to find an appropriate balance between
formal control mechanisms and activities fostering creativity. In addition, one process model should
provide support for different types of innovations having versatile requirements. These dichotomies need to
be solved to some extent by organization-specific choices. It can be concluded that both formalization and
flexibility are needed. The empirical findings indicated that any model is better than no model. Front-end
process formalization is associated with superior product concepts. In the situation of high market
uncertainty, this relationship is even stronger. Process formalization provides management an efficient
control mechanism to ensure that developed product concepts have superior value in competitive business
environment.
References
[1] Abernethy, M. A. & Brownell P. (1997). Management control systems in research and development
organizations: The Role of accounting, behaviour and personnel controls. Accounting, Organizations and
Society, Vol. 22, No. 3/4, 233-248.
[2] Amabile, T. M. (1998). How to kill creativity: Keep doing what you're doing. Or, if you want to spark
innovation, rethink how you motivate, reward, and assign work to people. Harvard Business Review
(September-October): 77- 87.
[3] Armstrong, J. S. & Overton, T. S. (1977). Estimating nonresponse bias in mail surveys. Journal of
Marketing Research, Vol. XIV, 396-402.
[4] Benner, M. J. & Tushman, M. (2002). Process management and technological innovation: a
longitudinal study of the photography and paint industries. Administrative Science Quarterly, 47, 676-706.
[5] Bonner, J.M., Ruekert, R. W. & Walker, Jr. O.C. (2002). Upper management control of new product
development projects and project performance. The Journal of Product Innovation Management, 19, 233245.
[6] Brown, S. L. & Eisenhardt, K. M. (1997). The art of continuous change: linking complexity theory and
time-paced evolution in relentlessly shifting organizations. Administrative Science Quarterly, Vol. 42, No.
1, 1-34.
[7] Buggie, F. D. (2002). Set the ”fuzzy front end” in concrete”, Research Technology Management, 45,
11-15.
[8] Burns, T. & Stalker, G. M. (1966). The management of innovation. Second Impression, Tavistock
Publications. 269 p.
[9] Cagan, J. and Vogel, G. M. (2002). Creating breakthrough products: innovation from product planning
to program approval. Prentice Hall.
[10] Campbell, D. T. (1955). The informant in quantitative research. The American Journal of Sociology,
Vol. 60, No. 4, 339-342.
[11] Campbell, D. T. & Fiske, D. W. (1959). Convergent and discriminant validation by the multitraitmethod matrix. Psychological Bulletin, 56, 2, 81-105.
[12] Choudhury, V. & Sabherwal, R. (2003). Portfolio of control in out-sourced software development
projects. Information System Research, 14, 3, pp. 291-314.
[13] Cleland, D. I. & King, W. R. (1975). System analysis and project management. McGraw-Hill Book
Company. 398 p.
[14] Cooper, R. G. (1998). Winning at new products: accelerating the process from idea to launch. 2nd
Edition, Addison Wesley Publishing Company. 358 p.
[15] Cooper, R. G. (1994). New products: The factors that drive success. International Marketing Review.
Vol. 11, No. 1, 60-76.
Note: Please take contact to the author for the full list of references.