Description
Many firms that compete based on the development of new and innovative products have
begun to adopt concurrent new product development (NPD) processes in which product
design phases occur in a non-linear and iterative manner. While concurrent NPD processes
increase flexibility and reduce time-to-market as compared to traditional sequential processes,
concurrency increases task uncertainty since the product design process begins
before all important product features and specifications have been established. Such
changes can result in costly redesign and rework. Prior research suggests target costing,
where product design teams are assigned specific cost goals, is an effective method of controlling
costs in sequential NPD. Even so, it is unclear whether target costing will improve
cost reduction performance when combined with a concurrent NPD process due to
increased task uncertainty. We examine experimentally the ability of product design
groups to achieve specific or general cost reduction goals under simulated sequential or
concurrent NPD. We predict and find that the nature of the NPD process moderates the
effect of specific cost reduction goals on actual cost reduction performance.
The effect of cost goal speci?city and new product development
process on cost reduction performance
Mohan Gopalakrishnan
a
, Theresa Libby
b,?
, Janet A. Samuels
c
, Dan Swenson
a
a
W.P. Carey School of Business, Arizona State University, United States
b
School of Accounting and Finance, University of Waterloo, Waterloo, ON N2L 3G1, Canada
c
Thunderbird School of Global Management, Arizona State University, United States
a r t i c l e i n f o
Article history:
Available online 12 February 2015
a b s t r a c t
Many ?rms that compete based on the development of new and innovative products have
begun to adopt concurrent new product development (NPD) processes in which product
design phases occur in a non-linear and iterative manner. While concurrent NPD processes
increase ?exibility and reduce time-to-market as compared to traditional sequential pro-
cesses, concurrency increases task uncertainty since the product design process begins
before all important product features and speci?cations have been established. Such
changes can result in costly redesign and rework. Prior research suggests target costing,
where product design teams are assigned speci?c cost goals, is an effective method of con-
trolling costs in sequential NPD. Even so, it is unclear whether target costing will improve
cost reduction performance when combined with a concurrent NPD process due to
increased task uncertainty. We examine experimentally the ability of product design
groups to achieve speci?c or general cost reduction goals under simulated sequential or
concurrent NPD. We predict and ?nd that the nature of the NPD process moderates the
effect of speci?c cost reduction goals on actual cost reduction performance. While speci?c
cost goals result in higher reductions in product cost than general cost goals under a
sequential NPD process, speci?c goals are no better than general goals in motivating design
groups to reduce product cost under a concurrent NPD process; thus, we demonstrate
boundary conditions on the usefulness of target costing as a cost control method.
Ó 2015 Elsevier Ltd. All rights reserved.
Introduction
New product development (NPD) processes comprise
several phases that typically include planning, concept
design, product design and testing, and production start-
up (Davila, 2000). These phases have traditionally been
performed sequentially and in lock-step (Kalyaraman &
Krishnan, 1997; Valle & Vazquez-Bustelo, 2009). Decisions
about product features and speci?cations are identi?ed
and ‘‘frozen’’ before the actual design process begins
(Hertenstein & Platt, 2000). In contrast, under concurrent
NPD, design phases occur simultaneously and in a non-lin-
ear manner. Product speci?cations may unexpectedly
change due to upstream decisions about product features
that continue to occur even though downstream product
design activity has already begun (Loch & Terwiesch,
1998; Mitchell & Nault, 2007). Thus, task uncertainty,
de?ned by the number of exceptions and degree of impro-
visation required to complete internal tasks (Perrow,
1970), is higher under concurrent than under traditional
sequential NPD (Mitchell & Nault, 2007).
An important and relatively unexplored issue is how
?rms control NPD costs when task uncertainty is high.http://dx.doi.org/10.1016/j.aos.2015.01.003
0361-3682/Ó 2015 Elsevier Ltd. All rights reserved.
?
Corresponding author. Tel.: +1 519 888 4567x31088.
E-mail addresses: [email protected](M. Gopalakrishnan),
[email protected] (T. Libby), [email protected] (J.A. Samuels),
[email protected] (D. Swenson).
Accounting, Organizations and Society 42 (2015) 1–11
Contents lists available at ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
Prior research suggests target costing is an effective cost
management tool ?rms use during the NPD process
(Booker, Drake, & Heitger, 2007; Cooper, 2002; Davila &
Wouters, 2004; Dekker & Smidt, 2003; Everaert &
Bruggeman, 2002). Target costs are speci?c cost goals cal-
culated by subtracting a target pro?t from the product’s
market-driven sales price (Ax, Greve, & Nilsson, 2008). Set-
ting and working toward target costs can provide signi?-
cant cost savings in sequential NPD processes (Anderson
& Sedatole, 1998; Cooper, 2002; Cooper & Slagmulder,
1999).
1
Concurrent processes, on the other hand, are inher-
ently more uncertain than traditional sequential processes
(Mitchell & Nault, 2007) and thus, we examine whether
assigning speci?c cost goals (as would be the case under tar-
get costing) will also be effective in controlling costs under
concurrent NPD. This issue is important given the recent
widespread adoption of concurrent NPD processes in prac-
tice (Mitchell & Nault, 2007; Valle & Vazquez-Bustelo, 2009).
Hirst (1987) develops a theoretical proposition that task
uncertainty will limit the effectiveness of speci?c cost
goals (such as target costs) in directing effort and perfor-
mance, although he does not test this proposition empiri-
cally. Essentially, Hirst (1987) argues that as task
knowledge becomes less complete, task uncertainty
increases and individuals are less able to identify the most
effective ways to direct their effort towards improved per-
formance when presented with speci?c goals. This effect is
much less severe (or even nonexistent) under general (‘‘do
your best’’) goals. Therefore, based on the theory devel-
oped in Hirst (1987), we hypothesize that the nature of
the design process (sequential or concurrent) will moder-
ate the effectiveness of speci?c cost goals in motivating
increased cost reduction by new product design groups.
In particular, due to higher task uncertainty under a con-
current NPD process, speci?c cost reduction goals will be
less effective in reducing product cost under concurrent
relative to sequential NPD.
In our experiment, 186 participants are assigned to
three-person product design groups and are required to
redesign a small truck to meet new product speci?cations.
The group’s objective is to lower the truck’s cost while
achieving the stated functional/technical speci?cations.
2
Participants are students enrolled in MBA or other post-
graduate executive education programs. To operationalize
variations in speci?city of cost goals, we assign a speci?c
goal (‘‘achieve a ?nal product cost of $16,500’’) to half of
the design groups and a general goal (‘‘reduce costs as much
as you can’’) to the others. We operationalize the sequential
and concurrent design processes by varying the timing of
information delivered to design groups regarding new prod-
uct speci?cations. We simulate a more certain sequential
process by informing design groups of all product speci?ca-
tions before they begin to redesign the product. To simulate
a relatively less certain concurrent process, the design
groups receive the same information in total as under the
sequential process, but one third of the product speci?ca-
tions are provided at each of three different intervals during
the work period and the design groups do not know when or
if this new information will be received.
We ?nd that the nature of the design process (concur-
rent or sequential NPD) moderates the relation between
cost goal speci?city and cost reduction performance.
Groups assigned a speci?c cost goal reduced costs more
than those assigned a general cost goal in the sequential
NPD (low uncertainty) condition while cost reduction per-
formance is no better under speci?c cost goals than under
general (‘‘do your best’’) cost goals in the concurrent NPD
(high uncertainty) condition.
This paper makes several contributions to both research
and practice. First, we answer the call by Davila and
Wouters (2004) for additional research to better under-
stand how cost management tools like target costing can
help or hinder innovation given recent changes in the man-
ufacturing environment and the management of the NPD
process. Second, our ?ndings suggest cost goal speci?city
and the ?rm’s approach to the NPD process may jointly
in?uence design teams’ ability to reduce the cost of new
and redesigned products. These results are important given
recent popularity of concurrent NPD (Mitchell & Nault,
2007; Valle & Vazquez-Bustelo, 2009) in many organiza-
tions looking to reduce time-to-market for new products.
As more ?rms adopt concurrent processes, it will be impor-
tant to recognize that high task uncertainty may be a
boundary condition on the effectiveness of target costing;
hence developing alternate cost management tools that
are effective in the uncertain concurrent NPD environment
should be a priority for these organizations (Davila &
Wouters, 2004).
The remainder of this paper is organized as follows. The
next section reviews the literature and develops our
hypothesis. We then describe the research design and
results of hypothesis tests and ?nish with a discussion
and suggestions for further research.
Literature review and hypothesis development
Target costing
Target costing is de?ned as a pro?t planning and cost
management system used to support the development of
new and redesigned products (Kee & Matherly, 2013). Cost
targets are derived by deducting required pro?t margins
from market prices and typically serve to control costs dur-
ing the design stage of NPD (Ansari, Bell, & Okano, 2007). In
practice, once a product’s cost target is established, a team
works to design a product that satis?es customer require-
ments at no higher than the target cost (Cooper &
Slagmulder, 1999). In a target costing environment, design
teams are instructed to not just ‘‘design a good compo-
nent,’’ but to instead ‘‘design the best component for a
given amount of money’’ (Mihm, 2010, 1334).
Conditions under which target costing is adopted and
reasons for its use have been examined extensively using
1
While most prior research fails to distinguish sequential and concur-
rent NPD processes, Exhibit 2 in the review article of Hertenstein and Platt
(2000) provides an example of a traditional sequential NPD process map
studied in most of the prior accounting literature.
2
The experimental task is modeled after a target costing exercise used
internally by the Boeing Company (also described in detail in Everaert &
Swenson, 2014). The truck’s initial product cost was set at $20,000 for all
design groups.
2 M. Gopalakrishnan et al. / Accounting, Organizations and Society 42 (2015) 1–11
survey and ?eld data. Results indicate that target costing is
adopted most frequently by ?rms operating in uncertain
external environments (Dekker & Smidt, 2003; Tani et al.,
1994), although Ax et al. (2008) ?nd the relation between
intensity of competition and adoption of target costing to
be indirect through perceived environmental uncertainty
in their sample ?rms. Most respondents to prior surveys
on the adoption of target costing indicate it is perceived
as a useful cost control technique that can result in
improved product quality, more cost effective product
designs and timely new product introductions (Dekker &
Smidt, 2003; Ellram, 2000; Tani et al., 1994). Conditions
under which target costing is more or less effective in
reducing product cost have also been examined. For exam-
ple, Ibusuki and Kaminski (2007), in their ?eld study of an
automotive manufacturer’s NPD process, found that target
costing was most effective when paired with value engi-
neering methodologies while Filomena, Neto, and Duffey
(2009) examine a ?eld site in Brazil where managers
believed target costing should extend beyond careful
tracking of product-development costs to also include set-
ting a target cost for the NPD process itself. No prior stud-
ies of which we are aware have examined the usefulness of
target costing under concurrent NPD.
In the target costing literature, Everaert and Bruggeman
(2002) experimentally examine the effect of speci?c, dif?-
cult cost goals and time pressure on cost reduction in NPD.
While they do not specify the nature of the NPD process
(i.e., sequential or concurrent), the description of their
experimental task implies very little uncertainty in the
NPD process. Their results indicate speci?c dif?cult cost
reduction goals result in greater cost reduction than gen-
eral (‘‘do your best’’) goals, although this effect is dimin-
ished when time pressure is high. Everaert, Boër, and
Bruggeman (2000) use a similar experimental design to
examine the effects of cost goals and time pressure for a
‘‘derivative’’ new product, de?ned as requiring an incre-
mental change to an existing product, and for a ‘‘next gen-
eration’’ new product, de?ned as involving a more radical
change requiring signi?cant creativity from design engi-
neers. Results indicate signi?cantly higher cost reduction
in the presence of a speci?c cost goal for the derivative,
but not the next generation product implying that speci?c
cost goals improve cost reduction performance when task
complexity is low rather than high.
3
Even so, task uncer-
tainty has not previously been examined in this context.
Task uncertainty and NPD
In the current study, we focus on the product design
phase of NPD characterized by the development of the
physical product. In traditional sequential NPD, the prod-
uct design phase does not begin until all product speci?ca-
tions and features are con?rmed. Thus, sequential NPD is
described as linear and ‘‘lock-step’’ (Hertenstein & Platt,
2000). In concurrent NPD, the speci?cations and product
features coming out of the concept design phase are con-
sidered preliminary and may change even though the
product designers have begun to work on the next phase
in the NPD process (Krishnan & Ulrich, 2001). For this rea-
son, concurrent NPD has been described as non-linear and
iterative (Mitchell & Nault, 2007). Some managers prefer to
implement concurrent NPD processes because they can
reduce time-to-market for new products, although concur-
rency can also mean that product designers’ work becomes
inherently more uncertain (Mitchell & Nault, 2007).
4
Task uncertainty in concurrent NPD is driven by vari-
ability and lack of analyzability of the work that is per-
formed (Perrow, 1970). Variability refers to the degree of
routineness of the work performed while analyzability
refers to the degree to which exceptions to routine work
require workers to improvise a solution (often referred to
as degree of task knowledge). Concurrent NPD is character-
ized by more variability and lack of routine in the product
designer’s work than would be experienced under sequen-
tial NPD. In addition, concurrent NPD processes are inher-
ently less analyzable since, by design, they include more
‘‘exceptions’’ to work routines requiring designers to
improvise solutions more often than under sequential
NPD. Thus we characterize the product design phase car-
ried out under concurrent NPD as higher on task uncer-
tainty than the product design phase that is carried out
under sequential NPD processes.
Task uncertainty, goal speci?city and cost reduction
performance
According to Locke and Latham (2002), goals lead to
improved performance by directing attention and effort
toward goal-relevant activities. In addition, goals play an
energizing role with speci?c, dif?cult goals motivating
higher effort than easy goals or ‘‘do your best’’ goals. Goals
also affect persistence with speci?c, dif?cult goals motivat-
ing greater persistence at a task than easy or general (‘‘do
your best’’) goals. Finally, goals affect motivation indirectly
by leading to discovery and use of task-relevant
knowledge.
In a theoretical analysis based on goal-setting theory,
Hirst (1987) considers the effect of task uncertainty on
the link between budget goal dif?culty and performance.
5
Hirst (1987) argues that speci?c, dif?cult budget goals focus
attention on goal interpretation, strategy search and selec-
tion. In effect, goal speci?city makes behavior selective and
this can improve performance if the selective behavior
reduces effort expended on irrelevant activities. However,
as task knowledge becomes less complete, task uncertainty
increases and individuals are less likely to easily identify
the activities required to improve performance. Hirst
(1987) proposes therefore, that task uncertainty will moder-
ate the relation between goal speci?city and performance;
speci?c, dif?cult goals will lead to improved performance
when task uncertainty is low, but not when task uncertainty
is high.
3
Results for time pressure and the interaction of time pressure and cost
goal are not reported by the authors for either the derivative or next
generation products.
4
In fact, Mitchell and Nault (2007, 375) call uncertainty ‘‘the engineering
design problem most often examined in the literature on concurrency.’’
5
Budgets are a particular type of cost, pro?t or return-on-investment
goal against which performance can be evaluated (Kren, 1990).
M. Gopalakrishnan et al. / Accounting, Organizations and Society 42 (2015) 1–11 3
Empirical evidence based on Hirst (1987) is scarce.
More recent empirical papers citing Hirst (1987) tend to
study the effects of environmental uncertainty on the
adoption of goals rather than the effect of task uncertainty
on the effectiveness of goals (Hartmann, 2000, 2007). Thus,
we examine theory and empirical evidence beyond Hirst
(1987) of a relation among task uncertainty, performance
measures/targets and actual performance. For example,
Chenhall (2003) provides a review of the survey-based lit-
erature from the early 1980s until 2002 using a contin-
gency framework to study issues of management control.
Based on his review of theory and empirical results,
Chenhall (2003) forwards the proposition, similar to that
of Hirst (1987), that as task uncertainty increases, controls
become less formal including less reliance on planning and
on pre-established accounting performance measures.
One study reviewed by Chenhall (2003) of particular
importance to our work is Davila (2000) that uses four case
studies of NPD processes and prior literature to develop a
set of testable hypotheses concerning conditions under
which management control systems would be more and
less useful in NPD. Davila (2000) hypothesizes and ?nds
that management control systems are used less intensively
in NPD when technological uncertainty is high. Similar
results are found by Abernethy and Brownell (1997) for
R&D organizations. While neither Davila (2000) nor
Abernethy and Brownell (1997) address target costing or
cost goal speci?city directly, Dekker, Groot, and Schoute
(2012) examine the effect of task uncertainty on the spec-
i?city of divisional performance targets. They ?nd that
?rms with lower task uncertainty are more likely to use
speci?c performance targets.
In the NPD context, product design groups can be
assigned speci?c or general cost goals (Cooper, 2002;
Tornberg, Jamsen, & Paranko, 2002) to motivate them to
reduce product costs. Groups that operate under a sequen-
tial NPD process receive ?nalized design information at the
outset; therefore, they are more easily able to focus their
efforts on achieving a speci?c cost target. Under a concur-
rent NPD process, there is much uncertainty and variability
in the timing and type of product speci?cation information
that is available to design groups once the design process
has begun (Loch & Terwiesch, 1998; Roemer, Ahmadi, &
Wang, 2000). Furthermore, groups must periodically redi-
rect their cost reduction efforts to incorporate new speci?-
cations as they are received. In essence, the concurrent
NPD process, when paired with a speci?c cost reduction
target, can actually place additional cognitive demands
on group members which can reduce the speci?c cost
goal’s potential effectiveness in motivating effort to reduce
costs compared to those groups assigned a general cost
goal (Byron, Khazanchi, & Nazarian, 2010). These argu-
ments imply assigning speci?c cost reduction goals will
improve performance more under a sequential than a con-
current NPD process leading to the following hypothesis:
Hypothesis. The nature of the product design process will
moderate the relation between cost goal speci?city and cost
reduction performance; groups assigned a speci?c cost goal
will reduce product cost more than groups assigned a general
cost goal under a sequential, but not a concurrent NPDprocess.
Experimental design and method
The experiment employs a 2 (speci?c/general cost reduc-
tion goal) Â 2 (sequential/concurrent NPD process)
between-subjects design. Participants include 186 students
enrolled in executive education and MBA programs at large
universities inthe US, Europe andIndia.
6
Participants are ran-
domly assigned to an experimental condition and then to one
of 62 three-person groups. The objective of the group was to
redesign a truck made of LEGO™ parts in order to meet new
product speci?cations while reducing cost.
7
The trucks pro-
duced by six groups did not meet the required product speci?-
cations, stability or safety tests and were disquali?ed.
8
In
addition, one group did not answer the post-experimental
questionnaire including our ‘‘understanding of the cost work-
shop’’ covariate (described in detail below). Therefore, 55
groupsareincludedintheANCOVAusedtotest our hypothesis.
Participants completed the task as part of a classroom
exercise. They were told they would be working on a group
activity to learn more about product design and cost man-
agement techniques. Completion of the truck redesign task
was followed by debrie?ng by the instructor concerning the
learning objectives of the activity. Participation in the activ-
ity was a required element of the course and results of the
exercise (e.g., which groups were able to reduce costs the
most) were revealed to the participants once complete.
Participants’ mean age was 28.9 years (std. dev. = 5.63)
and their full-time work experience averaged 6.8 years
(std. dev. = 6.2 years). Seventy percent of participants were
male. Nineteen percent of participants reported working in
accounting/?nance, 28 percent were engineers and the
remaining participants reported their functional area as
‘‘other.’’
Task overview
Groups engaged in an exercise to redesign a LEGO™
model truck subject to certain design speci?cations, quality
requirements and cost constraints (see Appendix A). The
truck’s cost was determined by summing its direct mate-
rial, direct labor, and overhead costs. Material and labor
costs were assigned based on the size and quantity of
LEGO™ parts used. Overhead costs were assigned based
on the number of different types of parts used in the truck’s
design (see the product costing worksheet in Appendix B).
To develop the experimental task, two of the authors
participated in internal training workshops on target cost-
ing conducted at The Boeing Company. Similar to our
experiment, Boeing’s workshop included a ‘‘hands-on’’
target costing exercise using an advertising display wagon
as the product. We also received feedback on the degree of
6
Program of enrollment was not a signi?cant covariate in our hypothesis
tests described below.
7
While many ?rms that implement target costing employ cross-
functional design teams (Swenson, Ansari, Bell, & Kim, 2003), we chose
not to attempt to simulate a cross functional team in the lab to limit
potential noise in our results.
8
Speci?cally, the trucks redesigned by three groups failed to meet one of
the design speci?cations, one group failed the minor crash test, and two
groups failed the lift test. Failures were distributed across experimental
conditions.
4 M. Gopalakrishnan et al. / Accounting, Organizations and Society 42 (2015) 1–11
realism in our experimental task from members of the
Consortium for Advanced Management—International’s
(CAM-I) Target Costing Group. We then conducted several
pilot tests to re?ne and standardize the experimental task.
The experiment lasted two hours and 45 min and, for
consistency, administrators followed a script throughout.
For the initial instructional phase, all participants received
a set of written materials and viewed a presentation by the
administrator describing the truck’s building blocks and
components. Next, each participant received a set of parts
with written instructions about how to build the truck
according to its original design. Participants then spent
20 min learning how to build the original truck.
For the next step, the administrator explained how to
calculate the truck’s product cost using the product costing
worksheet. Participants used this worksheet to calculate
the product cost for the truck they had just built. At this
point in the exercise, all participants had handled the truck
parts, built the truck according to its original design, and
used the product costing worksheet to calculate its cost.
Participants were then randomly assigned to a three-per-
son design group.
Groups next received information describing their task
along with a speci?c or a general cost goal. Each group
member completed a pre-experimental questionnaire that
assessed their understanding of the goal, their self-ef?cacy
and commitment to achieving the goal. Each group then
received two completed trucks in the original design (one
to be redesigned and the other as a reference of the original
design) and a standardized set of additional LEGO™ parts
to use when redesigning their truck. The groups also
received several blank product costing worksheets.
All groups had 60 minutes to complete their truck rede-
signs. Each workstation was partitioned off to prevent par-
ticipants from observing other groups’ work. When time
ran out, groups were told to stop working and each partic-
ipant completed a post-experimental questionnaire with
demographic and process-related questions. Following the
experiment, all trucks were examined by the administrator
to ensure they met quality and design speci?cations.
9
Independent variables
Speci?city of the cost goal
The ?rst between-subjects variable manipulated the
cost goal at two levels, a speci?c cost goal and a general
cost goal. All groups started with a truck that had a product
cost of $20,000. The written instructions for the speci?c
cost goal condition stated: ‘‘Your group’s objective is to
redesign your truck and achieve a cost goal of $16,500.’’ The
written instructions for the general cost goal condition
stated: ‘‘Your objective is to redesign your truck and reduce
costs as much as you can.’’
The speci?c cost goal was developed after reviewing the
goal setting literature and conducting several pilot tests.
While the literature reveals that the presence of speci?c,
dif?cult goals improves performance (Locke & Latham,
1990), it also emphasizes the importance of setting achiev-
able goals (Webb, 2004). Similar to Everaert and
Bruggeman (2002), we set the speci?c cost goal at a level
of cost reduction that was achieved 40% of the time by
the groups with no speci?c cost goal in our pilot tests of
the experiment.
Sequential and concurrent NPD process
The second between-subjects variable manipulated task
uncertainty via a sequential and a concurrent NPD process
by varying the point at which revised product speci?ca-
tions were provided to the design groups. The revised spec-
i?cations were divided into three segments and were
presented using both video clips and handouts. Details
are provided in Appendix C.
In the sequential condition, all revised speci?cations
were received by the design groups before the product
redesign process began while in the concurrent condition,
groups received the ?rst piece of information at the begin-
ning of the exercise, the second piece 20 min into the exer-
cise, and the third piece 40 min into the exercise. Thus,
concurrent design groups did not know if or when new
information would be received once the product redesign
process began. All groups were also informed of the
amount of time remaining until the end of the work period
at 20-min intervals. Groups in the sequential condition
learned how much time was remaining only while groups
in the concurrent condition also received the new pieces of
information at each 20-min interval.
Dependent variable
Our dependent variable is the dollar value of cost reduc-
tion achieved by each group measured as the difference
between the truck’s original design cost ($20,000) and
the cost of the redesigned truck. The truck’s original and
redesigned costs were calculated using an activity-based
costing (ABC) approach.
10
Our study included the number
of different types of parts utilized as an activity-based cost
driver. We did not tell the groups how to lower costs or
redesign their truck; however, they could reduce the truck’s
cost by eliminating unnecessary parts, by using less expen-
sive parts, and by reducing part variety.
11
The amount by
which cost was reduced was recalculated and veri?ed by
the administrator at the end of the session.
9
The practitioner literature expresses concern that quality standards
could be relaxed as companies pursue aggressive cost goals (Cooper &
Slagmulder, 1999; Kato, 1993). Therefore we used pilot tests to determine
how and if participants’ would cut corners while they redesigned their
trucks. We then controlled this potential behavior by developing Appendix
A to clarify the truck’s design speci?cations and other quality requirements
including stability and safety tests.
10
Prior research has suggested that the method of cost reporting can
in?uence behavior (Buchheit, 2004; Dearman & Shields, 2005). Thus, we
utilize a simple activity-based costing method in all conditions to minimize
potential cost reporting format confounds. While accounting/?nance
participants may have better understood the costing method, functional
background when entered as a covariate in our analysis was statistically
insigni?cant (p-value > 0.70).
11
Although using part variety as a cost driver added complexity to the
product costing exercise, it also made the experiment more robust by
having the participants consider an often overlooked cost driver that can
signi?cantly affect product costs.
M. Gopalakrishnan et al. / Accounting, Organizations and Society 42 (2015) 1–11 5
Results
Manipulation checks
To test participants’ understanding of the nature of
their cost goals, each participant responded to the question
‘‘What was your group’s objective?’’ The question was
open-ended and two of the researchers read through the
responses and coded them as follows: (1) cost reduction
was not mentioned at all; (2) general reduction of costs
was mentioned or (3) the speci?c cost goal was mentioned.
Seventy-two of the 81 participants in the general cost goal
condition completed the questionnaire and of these, four
failed to write an objective which mentioned cost reduc-
tion. Seventy-seven of the 87 participants in the speci?c
cost goal condition completed the questionnaire. Twelve
of these participants mentioned cost reduction, but did
not state the speci?c cost reduction goal in their response
while only one did not mention a cost goal at all. Given our
manipulation check question was qualitative and open
ended and no groups had more than one participant that
failed to mention the correct cost reduction goal, we retain
all groups in the sample for purpose of hypothesis testing
in the results section below. Even so, for completeness,
we also disclose results of our hypothesis tests for the
reduced sample in the tests of hypotheses section below.
Covariate
On our post-experimental questionnaire, we asked par-
ticipants to indicate the degree to which they understood
how to use the product costing worksheet on a scale of 1
(not at all) to 5 (completely). We then calculated the group
mean degree of understanding by averaging responses to
this question provided by each of the three members of
each group. The mean group response was 4.71 (std.
dev. = 0.28) indicating groups on average understood the
product costing worksheet very well. In addition, there
were no signi?cant differences in group means depending
on experimental condition. We do, however, ?nd a signi?-
cant positive correlation between responses to this ques-
tion and our dependent variable of cost reduction
(r = 0.31, p < 0.05). This indicates that groups that were
comfortable with and understood the product costing
sheet were better able to understand what was driving
their costs and achieve more effective cost reduction.
Based on this result, we include ‘‘understanding of the
worksheet’’ as a covariate in our hypothesis tests reported
below.
12
Descriptive statistics
Descriptivestatistics concerningdollars of cost reduction
by experimental condition are presented in Table 1. All
means areadjustedfor theeffect of thecovariate, that is, par-
ticipants’ understanding of the cost worksheet. In the
sequential condition, the adjusted mean dollars of cost
reduction per truck is $3786 (std. dev. = $814) when there
is a speci?c cost goal and $2819 (std. dev. = $855) when
there is a general cost goal. In the concurrent condition,
adjusted mean dollars of cost reduction is higher under a
general cost goal (mean = $3737; std. dev. = $778) than
under a speci?c cost goal (mean = $3507; std. dev. = $483).
13
Tests of hypothesis
We hypothesized that the nature of the NPD process
would moderate the relation between goal cost speci?city
and cost reduction performance such that groups assigned
a speci?c cost goal would reduce costs more than groups
assigned a general cost goal under the sequential, but not
the concurrent NPD process. We perform an analysis of
covariance (ANCOVA) to test this prediction. The results
of our ANCOVA are presented in Table 2 (Panel A).
The signi?cant interaction between cost goal speci?city
and NPD process (F = 10.04, p < 0.01) indicates initial sup-
port for our contention that the nature of the NPD process
moderates the relation between cost goal speci?city and
cost reduction performance.
14
Next we perform an ANCOVA
within each of the sequential and concurrent conditions (see
Table 2, Panel B). Consistent with our hypotheses, results
indicate a signi?cant main effect of cost goal speci?city in
Table 1
Adjusted mean (std. dev.) dollars of cost reduction by condition (n = 55).
NPD Process Cost goal type Overall
Speci?c General
Sequential $3786 $2819 $3302
($814) ($855) ($959)
n = 15 n = 14 n = 29
Concurrent $3507 $3737 $3622
($483) ($730) ($615)
n = 14 n = 12 n = 26
Overall $3647 $3278 $3462
($691) ($905) ($ 816)
n = 29 n = 26 n = 55
Note: Means are adjusted for the effect of the covariate (evaluated at
mean = 4.67) measured as the mean rating for each group of participants’
understanding of the costing worksheet (1 = not at all and 5 = com-
pletely). We lose one group in our analysis as they did not answer the
question about understanding of the cost worksheet used as our covari-
ate. Dollars of cost reduction is measured as the difference in cost
between the original and the redesigned truck.
12
The covariate was not signi?cantly correlated with cost goal type
(p = 0.43) or design process (p = 0.91) and no other potential covariates
were identi?ed.
13
We note the standard deviation of cost reduction dollars varies greatly
among cells. Even so, the Levene test for homogeneity of variance was not
signi?cant (F = 1.46, p = 0.24). Second, to examine whether outliers in?u-
ence our results, we create simple pass/fail categories based on whether the
group achieved the $3500 cost reduction goal. In the sequential condition,
80% of groups assigned speci?c cost goals achieved the cost reduction goal
while only 21% of groups assigned general cost goals achieved the goal. In
the concurrent condition, 50% of groups assigned speci?c cost goals met the
goal while 69% of groups assigned general cost goals achieved the goal.
These proportions follow the same pattern as results of our main
hypothesis tests reported in the next section of the paper and differences
in proportion are signi?cant based on a chi-square test (p-value < 0.01).
Thus, outliers do not appear to have a major in?uence on our results.
14
When we exclude the 17 groups that include one group member who
failed to mention the appropriate cost goal on the manipulation check
question, the interaction between cost goal and NPD process is marginally
signi?cant (p = 0.07) and the covariate is no longer signi?cant (p = 0.29).
6 M. Gopalakrishnan et al. / Accounting, Organizations and Society 42 (2015) 1–11
the sequential (F = 11.48, p < 0.01), but not in the concurrent
NPD condition (F = 1.15, p = 0.29).
15
Our results provide sup-
port for our hypothesis.
Supplemental analysis
To supplement our main analyses, we examine various
process measures obtained as part of the post-experimen-
tal questionnaire to see if they are associated with differ-
ences in cost reduction performance. First, we analyzed
the number of times that group members self-reported
redesign attempts on the truck as a whole during the 60-
min exercise. This statistic could be a potential explanation
for greater cost reduction by groups in the sequential con-
dition. We did not, however, ?nd a signi?cant difference in
the number of redesign attempts depending on cost goal,
NPD process or the interaction between goal and NPD pro-
cess (all p > 0.24).
In the concurrent condition, groups are required to
modify their design several times to incorporate new prod-
uct speci?cations. Thus, it is possible that the nature of the
concurrent NPD process will naturally increase partici-
pants’ frustration with the process. Frustration has been
identi?ed as a stressor that can place additional cognitive
demands on design groups (Byron et al., 2010). To examine
whether frustration affects our participants’ cost reduction
performance, we asked participants to respond to the fol-
lowing question: ‘‘To what extent did the timing of when
you received your design change information affect your
truck building experience?’’ Participants responded on a
scale of 1 (caused no frustration) to 5 (caused a lot of
frustration). Results of an ANOVA with frustration as the
dependent variable and cost goal and NPD process as
independent variables indicates a signi?cant difference
depending on NPD process (p < 0.01) with those in the con-
current condition indicating signi?cantly more frustration
(mean = 2.64, std. dev. = 0.85) than those in the sequential
condition (mean = 1.79, std. dev. = 0.57). Neither the main
effect of cost goal speci?city (p = 0.94) nor the interaction
between cost goal speci?city and NPD process (p = 0.71)
were signi?cant predictors of frustration with the task.
Everaert and Bruggeman (2002) ?nd that time pressure
diminishes the bene?cial effects of speci?c goals on cost
reduction performance for incremental design changes. In
our study, all groups operate under similar time con-
straints and pilot testing indicates that the 60 min allowed
was viewed by most participants as suf?cient to complete
the redesign. Even so, it is possible that some participants
perceived that time pressure limited their ability to
achieve cost goals. Therefore, as part of our post-experi-
mental questionnaire we asked participants about the
extent to which they perceived time pressure limited their
effectiveness. Results of an ANOVA with perceived time
pressure as a dependent variable and cost goal type and
design process as independent variables indicated neither
of the independent variables nor the interaction between
them were signi?cant (all p > 0.24). In addition, we ran
the same ANCOVA reported in Table 2, but with the addi-
tion of perceived time pressure as a covariate. Results were
qualitatively similar to those presented in Table 2 although
the perceived time pressure covariate was not signi?cant
(p = 0.58). Thus, time pressure does not appear to be an
alternative explanation for our experimental results.
Finally, we examined whether participants perceive the
task to be more dif?cult depending on experimental condi-
tion by asking ‘‘How dif?cult was it to achieve your group’s
objective?’’ Responses were given on a scale of 1 (not at all
dif?cult) to 5 (extremely dif?cult). Results of an ANOVA
with perceived dif?culty as the dependent variable and
cost goal type and design process as independent variables
indicates perceived dif?culty is signi?cantly higher under
speci?c cost goals (mean = 3.17, std. dev. = 0.68) than
general cost goals (mean = 2.60, std. dev. = 0.50). Perceived
Table 2
Analysis of covariance of dollars of cost reduction by cost goal type and NPD process (n = 55).
Sources of variation df MS F p
Panel A: ANCOVA – full sample
Cost goal speci?city 1 1,856,335 3.83 0.06
NPD process 1 1,371,830 2.83 0.10
Cost goal  NPD process 1 4,864,176 10.04
Many firms that compete based on the development of new and innovative products have
begun to adopt concurrent new product development (NPD) processes in which product
design phases occur in a non-linear and iterative manner. While concurrent NPD processes
increase flexibility and reduce time-to-market as compared to traditional sequential processes,
concurrency increases task uncertainty since the product design process begins
before all important product features and specifications have been established. Such
changes can result in costly redesign and rework. Prior research suggests target costing,
where product design teams are assigned specific cost goals, is an effective method of controlling
costs in sequential NPD. Even so, it is unclear whether target costing will improve
cost reduction performance when combined with a concurrent NPD process due to
increased task uncertainty. We examine experimentally the ability of product design
groups to achieve specific or general cost reduction goals under simulated sequential or
concurrent NPD. We predict and find that the nature of the NPD process moderates the
effect of specific cost reduction goals on actual cost reduction performance.
The effect of cost goal speci?city and new product development
process on cost reduction performance
Mohan Gopalakrishnan
a
, Theresa Libby
b,?
, Janet A. Samuels
c
, Dan Swenson
a
a
W.P. Carey School of Business, Arizona State University, United States
b
School of Accounting and Finance, University of Waterloo, Waterloo, ON N2L 3G1, Canada
c
Thunderbird School of Global Management, Arizona State University, United States
a r t i c l e i n f o
Article history:
Available online 12 February 2015
a b s t r a c t
Many ?rms that compete based on the development of new and innovative products have
begun to adopt concurrent new product development (NPD) processes in which product
design phases occur in a non-linear and iterative manner. While concurrent NPD processes
increase ?exibility and reduce time-to-market as compared to traditional sequential pro-
cesses, concurrency increases task uncertainty since the product design process begins
before all important product features and speci?cations have been established. Such
changes can result in costly redesign and rework. Prior research suggests target costing,
where product design teams are assigned speci?c cost goals, is an effective method of con-
trolling costs in sequential NPD. Even so, it is unclear whether target costing will improve
cost reduction performance when combined with a concurrent NPD process due to
increased task uncertainty. We examine experimentally the ability of product design
groups to achieve speci?c or general cost reduction goals under simulated sequential or
concurrent NPD. We predict and ?nd that the nature of the NPD process moderates the
effect of speci?c cost reduction goals on actual cost reduction performance. While speci?c
cost goals result in higher reductions in product cost than general cost goals under a
sequential NPD process, speci?c goals are no better than general goals in motivating design
groups to reduce product cost under a concurrent NPD process; thus, we demonstrate
boundary conditions on the usefulness of target costing as a cost control method.
Ó 2015 Elsevier Ltd. All rights reserved.
Introduction
New product development (NPD) processes comprise
several phases that typically include planning, concept
design, product design and testing, and production start-
up (Davila, 2000). These phases have traditionally been
performed sequentially and in lock-step (Kalyaraman &
Krishnan, 1997; Valle & Vazquez-Bustelo, 2009). Decisions
about product features and speci?cations are identi?ed
and ‘‘frozen’’ before the actual design process begins
(Hertenstein & Platt, 2000). In contrast, under concurrent
NPD, design phases occur simultaneously and in a non-lin-
ear manner. Product speci?cations may unexpectedly
change due to upstream decisions about product features
that continue to occur even though downstream product
design activity has already begun (Loch & Terwiesch,
1998; Mitchell & Nault, 2007). Thus, task uncertainty,
de?ned by the number of exceptions and degree of impro-
visation required to complete internal tasks (Perrow,
1970), is higher under concurrent than under traditional
sequential NPD (Mitchell & Nault, 2007).
An important and relatively unexplored issue is how
?rms control NPD costs when task uncertainty is high.http://dx.doi.org/10.1016/j.aos.2015.01.003
0361-3682/Ó 2015 Elsevier Ltd. All rights reserved.
?
Corresponding author. Tel.: +1 519 888 4567x31088.
E-mail addresses: [email protected](M. Gopalakrishnan),
[email protected] (T. Libby), [email protected] (J.A. Samuels),
[email protected] (D. Swenson).
Accounting, Organizations and Society 42 (2015) 1–11
Contents lists available at ScienceDirect
Accounting, Organizations and Society
j our nal homepage: www. el sevi er. com/ l ocat e/ aos
Prior research suggests target costing is an effective cost
management tool ?rms use during the NPD process
(Booker, Drake, & Heitger, 2007; Cooper, 2002; Davila &
Wouters, 2004; Dekker & Smidt, 2003; Everaert &
Bruggeman, 2002). Target costs are speci?c cost goals cal-
culated by subtracting a target pro?t from the product’s
market-driven sales price (Ax, Greve, & Nilsson, 2008). Set-
ting and working toward target costs can provide signi?-
cant cost savings in sequential NPD processes (Anderson
& Sedatole, 1998; Cooper, 2002; Cooper & Slagmulder,
1999).
1
Concurrent processes, on the other hand, are inher-
ently more uncertain than traditional sequential processes
(Mitchell & Nault, 2007) and thus, we examine whether
assigning speci?c cost goals (as would be the case under tar-
get costing) will also be effective in controlling costs under
concurrent NPD. This issue is important given the recent
widespread adoption of concurrent NPD processes in prac-
tice (Mitchell & Nault, 2007; Valle & Vazquez-Bustelo, 2009).
Hirst (1987) develops a theoretical proposition that task
uncertainty will limit the effectiveness of speci?c cost
goals (such as target costs) in directing effort and perfor-
mance, although he does not test this proposition empiri-
cally. Essentially, Hirst (1987) argues that as task
knowledge becomes less complete, task uncertainty
increases and individuals are less able to identify the most
effective ways to direct their effort towards improved per-
formance when presented with speci?c goals. This effect is
much less severe (or even nonexistent) under general (‘‘do
your best’’) goals. Therefore, based on the theory devel-
oped in Hirst (1987), we hypothesize that the nature of
the design process (sequential or concurrent) will moder-
ate the effectiveness of speci?c cost goals in motivating
increased cost reduction by new product design groups.
In particular, due to higher task uncertainty under a con-
current NPD process, speci?c cost reduction goals will be
less effective in reducing product cost under concurrent
relative to sequential NPD.
In our experiment, 186 participants are assigned to
three-person product design groups and are required to
redesign a small truck to meet new product speci?cations.
The group’s objective is to lower the truck’s cost while
achieving the stated functional/technical speci?cations.
2
Participants are students enrolled in MBA or other post-
graduate executive education programs. To operationalize
variations in speci?city of cost goals, we assign a speci?c
goal (‘‘achieve a ?nal product cost of $16,500’’) to half of
the design groups and a general goal (‘‘reduce costs as much
as you can’’) to the others. We operationalize the sequential
and concurrent design processes by varying the timing of
information delivered to design groups regarding new prod-
uct speci?cations. We simulate a more certain sequential
process by informing design groups of all product speci?ca-
tions before they begin to redesign the product. To simulate
a relatively less certain concurrent process, the design
groups receive the same information in total as under the
sequential process, but one third of the product speci?ca-
tions are provided at each of three different intervals during
the work period and the design groups do not know when or
if this new information will be received.
We ?nd that the nature of the design process (concur-
rent or sequential NPD) moderates the relation between
cost goal speci?city and cost reduction performance.
Groups assigned a speci?c cost goal reduced costs more
than those assigned a general cost goal in the sequential
NPD (low uncertainty) condition while cost reduction per-
formance is no better under speci?c cost goals than under
general (‘‘do your best’’) cost goals in the concurrent NPD
(high uncertainty) condition.
This paper makes several contributions to both research
and practice. First, we answer the call by Davila and
Wouters (2004) for additional research to better under-
stand how cost management tools like target costing can
help or hinder innovation given recent changes in the man-
ufacturing environment and the management of the NPD
process. Second, our ?ndings suggest cost goal speci?city
and the ?rm’s approach to the NPD process may jointly
in?uence design teams’ ability to reduce the cost of new
and redesigned products. These results are important given
recent popularity of concurrent NPD (Mitchell & Nault,
2007; Valle & Vazquez-Bustelo, 2009) in many organiza-
tions looking to reduce time-to-market for new products.
As more ?rms adopt concurrent processes, it will be impor-
tant to recognize that high task uncertainty may be a
boundary condition on the effectiveness of target costing;
hence developing alternate cost management tools that
are effective in the uncertain concurrent NPD environment
should be a priority for these organizations (Davila &
Wouters, 2004).
The remainder of this paper is organized as follows. The
next section reviews the literature and develops our
hypothesis. We then describe the research design and
results of hypothesis tests and ?nish with a discussion
and suggestions for further research.
Literature review and hypothesis development
Target costing
Target costing is de?ned as a pro?t planning and cost
management system used to support the development of
new and redesigned products (Kee & Matherly, 2013). Cost
targets are derived by deducting required pro?t margins
from market prices and typically serve to control costs dur-
ing the design stage of NPD (Ansari, Bell, & Okano, 2007). In
practice, once a product’s cost target is established, a team
works to design a product that satis?es customer require-
ments at no higher than the target cost (Cooper &
Slagmulder, 1999). In a target costing environment, design
teams are instructed to not just ‘‘design a good compo-
nent,’’ but to instead ‘‘design the best component for a
given amount of money’’ (Mihm, 2010, 1334).
Conditions under which target costing is adopted and
reasons for its use have been examined extensively using
1
While most prior research fails to distinguish sequential and concur-
rent NPD processes, Exhibit 2 in the review article of Hertenstein and Platt
(2000) provides an example of a traditional sequential NPD process map
studied in most of the prior accounting literature.
2
The experimental task is modeled after a target costing exercise used
internally by the Boeing Company (also described in detail in Everaert &
Swenson, 2014). The truck’s initial product cost was set at $20,000 for all
design groups.
2 M. Gopalakrishnan et al. / Accounting, Organizations and Society 42 (2015) 1–11
survey and ?eld data. Results indicate that target costing is
adopted most frequently by ?rms operating in uncertain
external environments (Dekker & Smidt, 2003; Tani et al.,
1994), although Ax et al. (2008) ?nd the relation between
intensity of competition and adoption of target costing to
be indirect through perceived environmental uncertainty
in their sample ?rms. Most respondents to prior surveys
on the adoption of target costing indicate it is perceived
as a useful cost control technique that can result in
improved product quality, more cost effective product
designs and timely new product introductions (Dekker &
Smidt, 2003; Ellram, 2000; Tani et al., 1994). Conditions
under which target costing is more or less effective in
reducing product cost have also been examined. For exam-
ple, Ibusuki and Kaminski (2007), in their ?eld study of an
automotive manufacturer’s NPD process, found that target
costing was most effective when paired with value engi-
neering methodologies while Filomena, Neto, and Duffey
(2009) examine a ?eld site in Brazil where managers
believed target costing should extend beyond careful
tracking of product-development costs to also include set-
ting a target cost for the NPD process itself. No prior stud-
ies of which we are aware have examined the usefulness of
target costing under concurrent NPD.
In the target costing literature, Everaert and Bruggeman
(2002) experimentally examine the effect of speci?c, dif?-
cult cost goals and time pressure on cost reduction in NPD.
While they do not specify the nature of the NPD process
(i.e., sequential or concurrent), the description of their
experimental task implies very little uncertainty in the
NPD process. Their results indicate speci?c dif?cult cost
reduction goals result in greater cost reduction than gen-
eral (‘‘do your best’’) goals, although this effect is dimin-
ished when time pressure is high. Everaert, Boër, and
Bruggeman (2000) use a similar experimental design to
examine the effects of cost goals and time pressure for a
‘‘derivative’’ new product, de?ned as requiring an incre-
mental change to an existing product, and for a ‘‘next gen-
eration’’ new product, de?ned as involving a more radical
change requiring signi?cant creativity from design engi-
neers. Results indicate signi?cantly higher cost reduction
in the presence of a speci?c cost goal for the derivative,
but not the next generation product implying that speci?c
cost goals improve cost reduction performance when task
complexity is low rather than high.
3
Even so, task uncer-
tainty has not previously been examined in this context.
Task uncertainty and NPD
In the current study, we focus on the product design
phase of NPD characterized by the development of the
physical product. In traditional sequential NPD, the prod-
uct design phase does not begin until all product speci?ca-
tions and features are con?rmed. Thus, sequential NPD is
described as linear and ‘‘lock-step’’ (Hertenstein & Platt,
2000). In concurrent NPD, the speci?cations and product
features coming out of the concept design phase are con-
sidered preliminary and may change even though the
product designers have begun to work on the next phase
in the NPD process (Krishnan & Ulrich, 2001). For this rea-
son, concurrent NPD has been described as non-linear and
iterative (Mitchell & Nault, 2007). Some managers prefer to
implement concurrent NPD processes because they can
reduce time-to-market for new products, although concur-
rency can also mean that product designers’ work becomes
inherently more uncertain (Mitchell & Nault, 2007).
4
Task uncertainty in concurrent NPD is driven by vari-
ability and lack of analyzability of the work that is per-
formed (Perrow, 1970). Variability refers to the degree of
routineness of the work performed while analyzability
refers to the degree to which exceptions to routine work
require workers to improvise a solution (often referred to
as degree of task knowledge). Concurrent NPD is character-
ized by more variability and lack of routine in the product
designer’s work than would be experienced under sequen-
tial NPD. In addition, concurrent NPD processes are inher-
ently less analyzable since, by design, they include more
‘‘exceptions’’ to work routines requiring designers to
improvise solutions more often than under sequential
NPD. Thus we characterize the product design phase car-
ried out under concurrent NPD as higher on task uncer-
tainty than the product design phase that is carried out
under sequential NPD processes.
Task uncertainty, goal speci?city and cost reduction
performance
According to Locke and Latham (2002), goals lead to
improved performance by directing attention and effort
toward goal-relevant activities. In addition, goals play an
energizing role with speci?c, dif?cult goals motivating
higher effort than easy goals or ‘‘do your best’’ goals. Goals
also affect persistence with speci?c, dif?cult goals motivat-
ing greater persistence at a task than easy or general (‘‘do
your best’’) goals. Finally, goals affect motivation indirectly
by leading to discovery and use of task-relevant
knowledge.
In a theoretical analysis based on goal-setting theory,
Hirst (1987) considers the effect of task uncertainty on
the link between budget goal dif?culty and performance.
5
Hirst (1987) argues that speci?c, dif?cult budget goals focus
attention on goal interpretation, strategy search and selec-
tion. In effect, goal speci?city makes behavior selective and
this can improve performance if the selective behavior
reduces effort expended on irrelevant activities. However,
as task knowledge becomes less complete, task uncertainty
increases and individuals are less likely to easily identify
the activities required to improve performance. Hirst
(1987) proposes therefore, that task uncertainty will moder-
ate the relation between goal speci?city and performance;
speci?c, dif?cult goals will lead to improved performance
when task uncertainty is low, but not when task uncertainty
is high.
3
Results for time pressure and the interaction of time pressure and cost
goal are not reported by the authors for either the derivative or next
generation products.
4
In fact, Mitchell and Nault (2007, 375) call uncertainty ‘‘the engineering
design problem most often examined in the literature on concurrency.’’
5
Budgets are a particular type of cost, pro?t or return-on-investment
goal against which performance can be evaluated (Kren, 1990).
M. Gopalakrishnan et al. / Accounting, Organizations and Society 42 (2015) 1–11 3
Empirical evidence based on Hirst (1987) is scarce.
More recent empirical papers citing Hirst (1987) tend to
study the effects of environmental uncertainty on the
adoption of goals rather than the effect of task uncertainty
on the effectiveness of goals (Hartmann, 2000, 2007). Thus,
we examine theory and empirical evidence beyond Hirst
(1987) of a relation among task uncertainty, performance
measures/targets and actual performance. For example,
Chenhall (2003) provides a review of the survey-based lit-
erature from the early 1980s until 2002 using a contin-
gency framework to study issues of management control.
Based on his review of theory and empirical results,
Chenhall (2003) forwards the proposition, similar to that
of Hirst (1987), that as task uncertainty increases, controls
become less formal including less reliance on planning and
on pre-established accounting performance measures.
One study reviewed by Chenhall (2003) of particular
importance to our work is Davila (2000) that uses four case
studies of NPD processes and prior literature to develop a
set of testable hypotheses concerning conditions under
which management control systems would be more and
less useful in NPD. Davila (2000) hypothesizes and ?nds
that management control systems are used less intensively
in NPD when technological uncertainty is high. Similar
results are found by Abernethy and Brownell (1997) for
R&D organizations. While neither Davila (2000) nor
Abernethy and Brownell (1997) address target costing or
cost goal speci?city directly, Dekker, Groot, and Schoute
(2012) examine the effect of task uncertainty on the spec-
i?city of divisional performance targets. They ?nd that
?rms with lower task uncertainty are more likely to use
speci?c performance targets.
In the NPD context, product design groups can be
assigned speci?c or general cost goals (Cooper, 2002;
Tornberg, Jamsen, & Paranko, 2002) to motivate them to
reduce product costs. Groups that operate under a sequen-
tial NPD process receive ?nalized design information at the
outset; therefore, they are more easily able to focus their
efforts on achieving a speci?c cost target. Under a concur-
rent NPD process, there is much uncertainty and variability
in the timing and type of product speci?cation information
that is available to design groups once the design process
has begun (Loch & Terwiesch, 1998; Roemer, Ahmadi, &
Wang, 2000). Furthermore, groups must periodically redi-
rect their cost reduction efforts to incorporate new speci?-
cations as they are received. In essence, the concurrent
NPD process, when paired with a speci?c cost reduction
target, can actually place additional cognitive demands
on group members which can reduce the speci?c cost
goal’s potential effectiveness in motivating effort to reduce
costs compared to those groups assigned a general cost
goal (Byron, Khazanchi, & Nazarian, 2010). These argu-
ments imply assigning speci?c cost reduction goals will
improve performance more under a sequential than a con-
current NPD process leading to the following hypothesis:
Hypothesis. The nature of the product design process will
moderate the relation between cost goal speci?city and cost
reduction performance; groups assigned a speci?c cost goal
will reduce product cost more than groups assigned a general
cost goal under a sequential, but not a concurrent NPDprocess.
Experimental design and method
The experiment employs a 2 (speci?c/general cost reduc-
tion goal) Â 2 (sequential/concurrent NPD process)
between-subjects design. Participants include 186 students
enrolled in executive education and MBA programs at large
universities inthe US, Europe andIndia.
6
Participants are ran-
domly assigned to an experimental condition and then to one
of 62 three-person groups. The objective of the group was to
redesign a truck made of LEGO™ parts in order to meet new
product speci?cations while reducing cost.
7
The trucks pro-
duced by six groups did not meet the required product speci?-
cations, stability or safety tests and were disquali?ed.
8
In
addition, one group did not answer the post-experimental
questionnaire including our ‘‘understanding of the cost work-
shop’’ covariate (described in detail below). Therefore, 55
groupsareincludedintheANCOVAusedtotest our hypothesis.
Participants completed the task as part of a classroom
exercise. They were told they would be working on a group
activity to learn more about product design and cost man-
agement techniques. Completion of the truck redesign task
was followed by debrie?ng by the instructor concerning the
learning objectives of the activity. Participation in the activ-
ity was a required element of the course and results of the
exercise (e.g., which groups were able to reduce costs the
most) were revealed to the participants once complete.
Participants’ mean age was 28.9 years (std. dev. = 5.63)
and their full-time work experience averaged 6.8 years
(std. dev. = 6.2 years). Seventy percent of participants were
male. Nineteen percent of participants reported working in
accounting/?nance, 28 percent were engineers and the
remaining participants reported their functional area as
‘‘other.’’
Task overview
Groups engaged in an exercise to redesign a LEGO™
model truck subject to certain design speci?cations, quality
requirements and cost constraints (see Appendix A). The
truck’s cost was determined by summing its direct mate-
rial, direct labor, and overhead costs. Material and labor
costs were assigned based on the size and quantity of
LEGO™ parts used. Overhead costs were assigned based
on the number of different types of parts used in the truck’s
design (see the product costing worksheet in Appendix B).
To develop the experimental task, two of the authors
participated in internal training workshops on target cost-
ing conducted at The Boeing Company. Similar to our
experiment, Boeing’s workshop included a ‘‘hands-on’’
target costing exercise using an advertising display wagon
as the product. We also received feedback on the degree of
6
Program of enrollment was not a signi?cant covariate in our hypothesis
tests described below.
7
While many ?rms that implement target costing employ cross-
functional design teams (Swenson, Ansari, Bell, & Kim, 2003), we chose
not to attempt to simulate a cross functional team in the lab to limit
potential noise in our results.
8
Speci?cally, the trucks redesigned by three groups failed to meet one of
the design speci?cations, one group failed the minor crash test, and two
groups failed the lift test. Failures were distributed across experimental
conditions.
4 M. Gopalakrishnan et al. / Accounting, Organizations and Society 42 (2015) 1–11
realism in our experimental task from members of the
Consortium for Advanced Management—International’s
(CAM-I) Target Costing Group. We then conducted several
pilot tests to re?ne and standardize the experimental task.
The experiment lasted two hours and 45 min and, for
consistency, administrators followed a script throughout.
For the initial instructional phase, all participants received
a set of written materials and viewed a presentation by the
administrator describing the truck’s building blocks and
components. Next, each participant received a set of parts
with written instructions about how to build the truck
according to its original design. Participants then spent
20 min learning how to build the original truck.
For the next step, the administrator explained how to
calculate the truck’s product cost using the product costing
worksheet. Participants used this worksheet to calculate
the product cost for the truck they had just built. At this
point in the exercise, all participants had handled the truck
parts, built the truck according to its original design, and
used the product costing worksheet to calculate its cost.
Participants were then randomly assigned to a three-per-
son design group.
Groups next received information describing their task
along with a speci?c or a general cost goal. Each group
member completed a pre-experimental questionnaire that
assessed their understanding of the goal, their self-ef?cacy
and commitment to achieving the goal. Each group then
received two completed trucks in the original design (one
to be redesigned and the other as a reference of the original
design) and a standardized set of additional LEGO™ parts
to use when redesigning their truck. The groups also
received several blank product costing worksheets.
All groups had 60 minutes to complete their truck rede-
signs. Each workstation was partitioned off to prevent par-
ticipants from observing other groups’ work. When time
ran out, groups were told to stop working and each partic-
ipant completed a post-experimental questionnaire with
demographic and process-related questions. Following the
experiment, all trucks were examined by the administrator
to ensure they met quality and design speci?cations.
9
Independent variables
Speci?city of the cost goal
The ?rst between-subjects variable manipulated the
cost goal at two levels, a speci?c cost goal and a general
cost goal. All groups started with a truck that had a product
cost of $20,000. The written instructions for the speci?c
cost goal condition stated: ‘‘Your group’s objective is to
redesign your truck and achieve a cost goal of $16,500.’’ The
written instructions for the general cost goal condition
stated: ‘‘Your objective is to redesign your truck and reduce
costs as much as you can.’’
The speci?c cost goal was developed after reviewing the
goal setting literature and conducting several pilot tests.
While the literature reveals that the presence of speci?c,
dif?cult goals improves performance (Locke & Latham,
1990), it also emphasizes the importance of setting achiev-
able goals (Webb, 2004). Similar to Everaert and
Bruggeman (2002), we set the speci?c cost goal at a level
of cost reduction that was achieved 40% of the time by
the groups with no speci?c cost goal in our pilot tests of
the experiment.
Sequential and concurrent NPD process
The second between-subjects variable manipulated task
uncertainty via a sequential and a concurrent NPD process
by varying the point at which revised product speci?ca-
tions were provided to the design groups. The revised spec-
i?cations were divided into three segments and were
presented using both video clips and handouts. Details
are provided in Appendix C.
In the sequential condition, all revised speci?cations
were received by the design groups before the product
redesign process began while in the concurrent condition,
groups received the ?rst piece of information at the begin-
ning of the exercise, the second piece 20 min into the exer-
cise, and the third piece 40 min into the exercise. Thus,
concurrent design groups did not know if or when new
information would be received once the product redesign
process began. All groups were also informed of the
amount of time remaining until the end of the work period
at 20-min intervals. Groups in the sequential condition
learned how much time was remaining only while groups
in the concurrent condition also received the new pieces of
information at each 20-min interval.
Dependent variable
Our dependent variable is the dollar value of cost reduc-
tion achieved by each group measured as the difference
between the truck’s original design cost ($20,000) and
the cost of the redesigned truck. The truck’s original and
redesigned costs were calculated using an activity-based
costing (ABC) approach.
10
Our study included the number
of different types of parts utilized as an activity-based cost
driver. We did not tell the groups how to lower costs or
redesign their truck; however, they could reduce the truck’s
cost by eliminating unnecessary parts, by using less expen-
sive parts, and by reducing part variety.
11
The amount by
which cost was reduced was recalculated and veri?ed by
the administrator at the end of the session.
9
The practitioner literature expresses concern that quality standards
could be relaxed as companies pursue aggressive cost goals (Cooper &
Slagmulder, 1999; Kato, 1993). Therefore we used pilot tests to determine
how and if participants’ would cut corners while they redesigned their
trucks. We then controlled this potential behavior by developing Appendix
A to clarify the truck’s design speci?cations and other quality requirements
including stability and safety tests.
10
Prior research has suggested that the method of cost reporting can
in?uence behavior (Buchheit, 2004; Dearman & Shields, 2005). Thus, we
utilize a simple activity-based costing method in all conditions to minimize
potential cost reporting format confounds. While accounting/?nance
participants may have better understood the costing method, functional
background when entered as a covariate in our analysis was statistically
insigni?cant (p-value > 0.70).
11
Although using part variety as a cost driver added complexity to the
product costing exercise, it also made the experiment more robust by
having the participants consider an often overlooked cost driver that can
signi?cantly affect product costs.
M. Gopalakrishnan et al. / Accounting, Organizations and Society 42 (2015) 1–11 5
Results
Manipulation checks
To test participants’ understanding of the nature of
their cost goals, each participant responded to the question
‘‘What was your group’s objective?’’ The question was
open-ended and two of the researchers read through the
responses and coded them as follows: (1) cost reduction
was not mentioned at all; (2) general reduction of costs
was mentioned or (3) the speci?c cost goal was mentioned.
Seventy-two of the 81 participants in the general cost goal
condition completed the questionnaire and of these, four
failed to write an objective which mentioned cost reduc-
tion. Seventy-seven of the 87 participants in the speci?c
cost goal condition completed the questionnaire. Twelve
of these participants mentioned cost reduction, but did
not state the speci?c cost reduction goal in their response
while only one did not mention a cost goal at all. Given our
manipulation check question was qualitative and open
ended and no groups had more than one participant that
failed to mention the correct cost reduction goal, we retain
all groups in the sample for purpose of hypothesis testing
in the results section below. Even so, for completeness,
we also disclose results of our hypothesis tests for the
reduced sample in the tests of hypotheses section below.
Covariate
On our post-experimental questionnaire, we asked par-
ticipants to indicate the degree to which they understood
how to use the product costing worksheet on a scale of 1
(not at all) to 5 (completely). We then calculated the group
mean degree of understanding by averaging responses to
this question provided by each of the three members of
each group. The mean group response was 4.71 (std.
dev. = 0.28) indicating groups on average understood the
product costing worksheet very well. In addition, there
were no signi?cant differences in group means depending
on experimental condition. We do, however, ?nd a signi?-
cant positive correlation between responses to this ques-
tion and our dependent variable of cost reduction
(r = 0.31, p < 0.05). This indicates that groups that were
comfortable with and understood the product costing
sheet were better able to understand what was driving
their costs and achieve more effective cost reduction.
Based on this result, we include ‘‘understanding of the
worksheet’’ as a covariate in our hypothesis tests reported
below.
12
Descriptive statistics
Descriptivestatistics concerningdollars of cost reduction
by experimental condition are presented in Table 1. All
means areadjustedfor theeffect of thecovariate, that is, par-
ticipants’ understanding of the cost worksheet. In the
sequential condition, the adjusted mean dollars of cost
reduction per truck is $3786 (std. dev. = $814) when there
is a speci?c cost goal and $2819 (std. dev. = $855) when
there is a general cost goal. In the concurrent condition,
adjusted mean dollars of cost reduction is higher under a
general cost goal (mean = $3737; std. dev. = $778) than
under a speci?c cost goal (mean = $3507; std. dev. = $483).
13
Tests of hypothesis
We hypothesized that the nature of the NPD process
would moderate the relation between goal cost speci?city
and cost reduction performance such that groups assigned
a speci?c cost goal would reduce costs more than groups
assigned a general cost goal under the sequential, but not
the concurrent NPD process. We perform an analysis of
covariance (ANCOVA) to test this prediction. The results
of our ANCOVA are presented in Table 2 (Panel A).
The signi?cant interaction between cost goal speci?city
and NPD process (F = 10.04, p < 0.01) indicates initial sup-
port for our contention that the nature of the NPD process
moderates the relation between cost goal speci?city and
cost reduction performance.
14
Next we perform an ANCOVA
within each of the sequential and concurrent conditions (see
Table 2, Panel B). Consistent with our hypotheses, results
indicate a signi?cant main effect of cost goal speci?city in
Table 1
Adjusted mean (std. dev.) dollars of cost reduction by condition (n = 55).
NPD Process Cost goal type Overall
Speci?c General
Sequential $3786 $2819 $3302
($814) ($855) ($959)
n = 15 n = 14 n = 29
Concurrent $3507 $3737 $3622
($483) ($730) ($615)
n = 14 n = 12 n = 26
Overall $3647 $3278 $3462
($691) ($905) ($ 816)
n = 29 n = 26 n = 55
Note: Means are adjusted for the effect of the covariate (evaluated at
mean = 4.67) measured as the mean rating for each group of participants’
understanding of the costing worksheet (1 = not at all and 5 = com-
pletely). We lose one group in our analysis as they did not answer the
question about understanding of the cost worksheet used as our covari-
ate. Dollars of cost reduction is measured as the difference in cost
between the original and the redesigned truck.
12
The covariate was not signi?cantly correlated with cost goal type
(p = 0.43) or design process (p = 0.91) and no other potential covariates
were identi?ed.
13
We note the standard deviation of cost reduction dollars varies greatly
among cells. Even so, the Levene test for homogeneity of variance was not
signi?cant (F = 1.46, p = 0.24). Second, to examine whether outliers in?u-
ence our results, we create simple pass/fail categories based on whether the
group achieved the $3500 cost reduction goal. In the sequential condition,
80% of groups assigned speci?c cost goals achieved the cost reduction goal
while only 21% of groups assigned general cost goals achieved the goal. In
the concurrent condition, 50% of groups assigned speci?c cost goals met the
goal while 69% of groups assigned general cost goals achieved the goal.
These proportions follow the same pattern as results of our main
hypothesis tests reported in the next section of the paper and differences
in proportion are signi?cant based on a chi-square test (p-value < 0.01).
Thus, outliers do not appear to have a major in?uence on our results.
14
When we exclude the 17 groups that include one group member who
failed to mention the appropriate cost goal on the manipulation check
question, the interaction between cost goal and NPD process is marginally
signi?cant (p = 0.07) and the covariate is no longer signi?cant (p = 0.29).
6 M. Gopalakrishnan et al. / Accounting, Organizations and Society 42 (2015) 1–11
the sequential (F = 11.48, p < 0.01), but not in the concurrent
NPD condition (F = 1.15, p = 0.29).
15
Our results provide sup-
port for our hypothesis.
Supplemental analysis
To supplement our main analyses, we examine various
process measures obtained as part of the post-experimen-
tal questionnaire to see if they are associated with differ-
ences in cost reduction performance. First, we analyzed
the number of times that group members self-reported
redesign attempts on the truck as a whole during the 60-
min exercise. This statistic could be a potential explanation
for greater cost reduction by groups in the sequential con-
dition. We did not, however, ?nd a signi?cant difference in
the number of redesign attempts depending on cost goal,
NPD process or the interaction between goal and NPD pro-
cess (all p > 0.24).
In the concurrent condition, groups are required to
modify their design several times to incorporate new prod-
uct speci?cations. Thus, it is possible that the nature of the
concurrent NPD process will naturally increase partici-
pants’ frustration with the process. Frustration has been
identi?ed as a stressor that can place additional cognitive
demands on design groups (Byron et al., 2010). To examine
whether frustration affects our participants’ cost reduction
performance, we asked participants to respond to the fol-
lowing question: ‘‘To what extent did the timing of when
you received your design change information affect your
truck building experience?’’ Participants responded on a
scale of 1 (caused no frustration) to 5 (caused a lot of
frustration). Results of an ANOVA with frustration as the
dependent variable and cost goal and NPD process as
independent variables indicates a signi?cant difference
depending on NPD process (p < 0.01) with those in the con-
current condition indicating signi?cantly more frustration
(mean = 2.64, std. dev. = 0.85) than those in the sequential
condition (mean = 1.79, std. dev. = 0.57). Neither the main
effect of cost goal speci?city (p = 0.94) nor the interaction
between cost goal speci?city and NPD process (p = 0.71)
were signi?cant predictors of frustration with the task.
Everaert and Bruggeman (2002) ?nd that time pressure
diminishes the bene?cial effects of speci?c goals on cost
reduction performance for incremental design changes. In
our study, all groups operate under similar time con-
straints and pilot testing indicates that the 60 min allowed
was viewed by most participants as suf?cient to complete
the redesign. Even so, it is possible that some participants
perceived that time pressure limited their ability to
achieve cost goals. Therefore, as part of our post-experi-
mental questionnaire we asked participants about the
extent to which they perceived time pressure limited their
effectiveness. Results of an ANOVA with perceived time
pressure as a dependent variable and cost goal type and
design process as independent variables indicated neither
of the independent variables nor the interaction between
them were signi?cant (all p > 0.24). In addition, we ran
the same ANCOVA reported in Table 2, but with the addi-
tion of perceived time pressure as a covariate. Results were
qualitatively similar to those presented in Table 2 although
the perceived time pressure covariate was not signi?cant
(p = 0.58). Thus, time pressure does not appear to be an
alternative explanation for our experimental results.
Finally, we examined whether participants perceive the
task to be more dif?cult depending on experimental condi-
tion by asking ‘‘How dif?cult was it to achieve your group’s
objective?’’ Responses were given on a scale of 1 (not at all
dif?cult) to 5 (extremely dif?cult). Results of an ANOVA
with perceived dif?culty as the dependent variable and
cost goal type and design process as independent variables
indicates perceived dif?culty is signi?cantly higher under
speci?c cost goals (mean = 3.17, std. dev. = 0.68) than
general cost goals (mean = 2.60, std. dev. = 0.50). Perceived
Table 2
Analysis of covariance of dollars of cost reduction by cost goal type and NPD process (n = 55).
Sources of variation df MS F p
Panel A: ANCOVA – full sample
Cost goal speci?city 1 1,856,335 3.83 0.06
NPD process 1 1,371,830 2.83 0.10
Cost goal  NPD process 1 4,864,176 10.04