Available online at www.sciencedirect.

com

Accounting, Organizations and Society 34 (2009) 322–347

www.elsevier.com/locate/aos

Reasons for management control systems adoption:

Insights from product development systems choice
by early-stage entrepreneurial companies
Antonio Davila a,*, George Foster b, Mu Li c
a
IESE, University of Navarra, Spain
b
Graduate School of Business, Stanford University, United States
c
Applied Micro Devices, United States

Abstract

Recent theoretical and empirical work indicates that management control systems (MCS) are an important element
in enhancing innovation. We extend this research thrust examining the adoption of MCS in product development,
arguably one of the business processes where innovation plays a major role. Using a sample of 69 early-stage entre-
preneurial companies, data are collected from questionnaires and interviews with each of the CEO, financial officer,
and business development managers pertaining to product development MCS. We examine seven different systems:
project milestones, reports comparing actual progress to plan, budget for development projects, project selection pro-
cess, product portfolio roadmap, product concept testing process, and project team composition guidelines. We
address three distinct questions: (1) What are the reasons-for-adoption of these systems? The nature of our sample
allows us to trace back to the adoption point and develop a set of reasons-for-adoption from the analysis of the data.
While MCS fulfill certain roles as described in the literature, these reasons-for-adoption are distinct from these roles.
Results indicate that certain events lead managers to adopt these systems and address the challenges that they face.
They include contracting and legitimizing the process with external parties and internal reasons-for-adoption such as
managers’ background, learning by doing, need to focus the organization, or reaction to problems. (2) Are these rea-
sons-for-adoption associated with differences across companies in the time from their founding date until these sys-
tems are adopted (time-to-adoption)? Prior research has looked at the covariance of various organizational
variables with this timing; this study goes a step further by looking at the effect of different reasons-for-adoption
on this timing. Our evidence finds an association between these two variables. (3) Are these reasons-for-adoption rel-
evant to performance? We find that the reason-for-adoption is associated with the on-time dimension of product
development performance.
Ó 2008 Elsevier Ltd. All rights reserved.

*
Corresponding author. Tel.: +34 93 602 4183; fax: +34 93 253 43 43.
E-mail address: [email protected] (A. Davila).

0361-3682/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aos.2008.08.002

1
 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347 323

Introduction between control and performance in both radical

as well as incremental innovation projects in the
Formal management control systems (MCS) pharmaceutical industry.2 Ditillo (2004) describes
have traditionally been associated with mechanistic MCS as a key element in knowledge intensive
organizations (Burns & Stalker, 1961). These systems firms.3 Similarly, Chapman (1998) presents evi-
support the periodic execution of the same routines dence consistent with the relevance of these sys-
with little if any changes. Their relevance to the tems in uncertain environments.
innovation process – a process associated with Based on a sample of 69 technology-based
uncertainty, with unknown links between inputs early-stage companies, the paper examines the
and outputs, with exceptions, and with outputs that adoption of MCS within an organizational pro-
are often hard to evaluate – is less clear. Ouchi cess where innovation has a pivotal role: the
(1979) used a research department to illustrate clan product development process.4 The focus on
control where social norms substitute for formal product development led to sampling from tech-
management systems. Mintzberg’s separation of nology-based firms.5 Product development is a
planning and managing (Mintzberg, 1976) and key aspect in these firms. If MCS are important
Quinn’s logical incrementalism (Quinn, 1978) also to managing innovation, this sample of compa-
highlight the limitations of traditional MCS. A fun- nies will be (on average) ahead in their use.6
damentally different perspective is that these sys- Our objective of learning about the adoption of
tems may provide important discipline to help MCS led to adding the early-stage criteria. The
manage uncertainty. Recent theoretical develop- focus on the adoption stage (rather than the evo-
ments offer various concepts that support the need lution of existing MCS) suggested studying com-
for formal management control systems (MCS) in panies that are going through the transition
uncertain settings.1 For instance, the distinction from birth to early-stage when the MCS are
between coercive and enabling bureaucracies (Adler
& Borys, 1996) suggests that MCS may be instru- 2
Cardinal identifies input, behavior and output control where
mental to innovation. Gavetti and Levinthal input control (scientific diversity and professionalization) might
(2000) present a learning model where companies be interpreted as informal control while behavior and output
that jointly rely on planning and learning by doing control are formal MCS. She finds all three types of control being
are predicted to perform better in uncertain envi- associated with performance for incremental and radical innova-
tion projects.
ronments compared to alternative strategies. Thus, 3
Ditillo (2004) interprets MCS similarly to the way this paper
forward looking efforts typically associated with uses the term. He describes them as ‘‘the design as well as use of
MCS complement fast reaction to new information coordination mechanisms based on the standardization of either
to improve how organizations deal with uncer- input, action, or results” (p. 402).
4
tainty. Zollo and Winter (2002) argue that the Because we study MCS in the specific process of product
development, MCS speaks to the systems that are used within this
essence of dynamic capabilities is adaptive routines process. In particular we examine the following MCS: project
– including information-based routines. Simons milestones, reports comparing actual progress to plan, budget for
(1995) interactive systems concept can have an development projects, project selection process, product portfolio
explicit role in sparking innovation around strategic roadmap, product concept testing process, and project team
uncertainties. composition guidelines. We collect data on actual systems rather
than on particular theoretical constructs associated with the
For the most part, recent empirical evidence design and use of these seven systems.
also indicates that innovation processes may gain 5
We chose to study a specific realization of innovation through
from the presence of MCS. Abernathy and Brow- looking at product development process rather than the broader
nell (1999) use Simons’ model to examine the use concept of ‘‘organizational innovation” to increase the power of
of budgets ‘‘as a dialogue, learning and idea the research design by reducing the noise and potential con-
founding factors.
creation machine” during episodes of strategic 6
The objective of the paper is not to examine whether the
change. Cardinal (2001) reports an association presence of MCS is needed for a company to be innovative.
Rather, its objective is to understand why MCS are adopted in a
1
MCS are defined as formal management control systems particular process associated with innovation and whether distinct
following Simons’ definition: ‘‘formal, information-based rou- reasons for adoption are associated with the time it takes to
tines and procedures managers use to maintain or alter patterns formalize these systems and to on-time performance. To probe the
in organizational activities” (Simons, 1995, p. 5). Throughout the link between the presence of MCS and innovation, one would
paper, MCS is used to refer to this definition unless otherwise need to develop a measure of innovativeness and compute this
noted. measure on a year-by-year basis for each company in our sample.

2
324 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347

adopted for the first time. We conceptualize MCS argument was not researched in depth for several
in our field research as formal systems particular decades.7 Cardinal, Sitkin, and Long (2004)
to product development including: project mile- observe that much of the literature has ‘‘virtually
stones, budget for development projects, reports ignored the origins and evolution of organiza-
comparing actual progress to plan, project selec- tional control” (p. 411). Recent research has
tion process, product portfolio roadmap, product started to address the origins and evolution issue.
concept testing process and, project team compo- The two main streams of research are: (a) case
sition guidelines. studies that describe the rise (and fall in some
Accordingly, this paper is built at the intersection cases) of controls over the early years of startup
of two main research thrusts. (a) The study exam- companies (Cardinal et al., 2004) and (Granlund
ines the product development process – where inno- & Taipaleenmaki, 2005), and (b) large-sample
vation has a significant role. Building on existing based studies of the association between MCS
theory on the relevance of MCS to innovation evolution and organizational variables such as
processes, it provides new evidence to advance our age, size, company strategy, and the presence of
conceptual understanding of why MCS are an venture capital (Davila, 2005; Davila & Foster,
important aspect of innovation management. In 2005, 2007; Sandino, 2007). The present study
particular, it highlights the difference between the probes the reasons that led to a differential
reasons-for-adoption of MCS in innovation pro- MCS adoption for 69 technology-based early-
cesses and the objectives that these systems pursue. stage companies in the product development
(b) Our sample of entrepreneurial companies also management systems area. Product development
speaks to the literature on the emergence of MCS; systems are especially important for high-techno-
thus, the conceptual development around the adop- logy companies given the pivotal role that product
tion of MCS in product development processes can innovation plays in their growth.
be the basis for empirical work to understand the Our research design combines qualitative data
adoption of these systems in other organizational from over 200 interviews and quantitative data
processes in early-stage companies. The study gathered using questionnaires. It also triangulates
speaks to the adoption of MCS in a process tradi- the data using three different informants per com-
tionally associated with innovation. Because the pany. Access to qualitative descriptions of why
study is framed within early-stage entrepreneurial MCS were adopted allows us to develop a frame-
companies to capture the adoption event, it does work of reasons why such systems are adopted.
not speak to the evolution of MCS in established We identify two external based reasons (labeled
firms. legitimize and contract) and four internal reasons
We examine three related research questions: (1) (labeled manger’s background, need to focus8,
why MCS are first adopted in the product develop- chaos, and learning). One benefit of the structured
ment process? In particular, we present various qualitative analysis of interviews is the rich descrip-
reasons-for-adoption of MCS, (2) what is the rela- tion we can report to illustrate the six different rea-
tionship between these reasons-for-adoption and sons-for-adoption.
speed of adoption? and (3) what is the relationship The second and third research questions build
between reasons-for-adoption and a key product upon the framework of reasons-for-adoption
development performance measure (on-time developed in the first question. In particular, the
development)? Fig. 1 illustrates these research second research question examines how the six dif-
questions. ferent reasons for adoption influence the speed of
To inform these questions, the literature offers adoption of product development systems. Speed
arguments on why organizations adopt MCS.
These arguments have an early antecedent in 7
MCS in small and medium companies have been studied using
Greiner (1972). The first stage of his growth cross-sectional designs (thus not looking at MCS adoption)
model deals with the emergence of MCS. It (Amat, Carmona, & Roberts, 1994; Romano & Ratnatunga,
describes the emergence of MCS as a crisis of 1994). Also the evolution of existing MCS has been examined
leadership where: ‘‘increased number of employees (Oakes, Townley, & Cooper, 1998). Yet the adoption of these
systems has only received attention recently.
cannot be managed exclusively through informal 8
‘‘Need to focus” refers to the manager’s adoption of MCS to
communication (. . .) and new accounting proce- focus the company on executing the strategy when the company is
dures are needed for financial control.” Yet this failing to do so because of lack of systems.

3
 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347 325

Organizational Covariates Such As:

Size
Age
Strategy
VC Funding
Founder Replaced

First Research Question

Reasons-for-Adoption Second Research
External Reasons Question
Legitimize Time to Adoption of Organizational
Contract MCS Performance
Internal Reasons: Proactive
Manager Background Third Research
Need to Focus
Question Innovation Process
Internal Reasons: Reactive
To Chaos Performance
To Learning

MCS Roles

Legend
: Prior Work : Current Work

Fig. 1. Overview of research questions and related literature. First research question: what are the reasons-for-adoption of product
development MCS? Second research question: are these reasons-for-adoption associated with differences across companies in the time from
their founding date until these systems are adopted (time-to-adoption)? Third research question: are these reasons-for-adoption relevant to
performance?

of adoption is measured using time elapsed from development performance are likely to have signifi-
the start of the company to the time reported for cant variation. The sample selection is designed to
the adoption of the project milestones’ system.9 capture these companies’ transition phase into
We find that the reasons-for-adoption associated MCS; moreover, their product development pro-
with manager’s background, chaos, legitimize, cesses are still evolving – for instance, through the
need to focus, and contract are found to be signif- adoption of MCS – and thus expected to show dif-
icantly related with faster adoption vis-à-vis a sub- ferent performance levels. This fact allows us to
set of our sample of companies with higher study the relationship between reason-for-adoption
reliance on informal product development control and speed of adoption and performance.
systems. The paper brings new evidence to the growing lit-
The third research question probes whether the erature on the relevance of MCS to enhance the per-
same six different reasons for MCS adoption influ- formance of firms employing organic structures
ence a key product development performance mea- (Kalagnanam & Lindsay, 1999) and in particular
sure (on-time development). Our results indicate to innovation processes (Bisbe & Otley, 2004; Granl-
that when MCS are adopted because of the man- und & Taipaleenmaki, 2005). The main findings
ager’s background, product development perfor- include: (1) we identified six reasons for MCS
mance is enhanced. adoption. Two reasons-for-adoption are externally
Because our sample comprises early-stage com- related: legitimizing the company and contracting
panies, their level of MCS adoption and product with external parties. This evidence highlights the
role of external parties in shaping management sys-
9 tems internal to the firm (Pfeffer & Salancik, 1978);
We adopt this measure because ‘‘project milestones” is the
system adopted most often. We also find similar results for an influence that is typically associated with financial
‘‘reports comparing actual progress to plan” and ‘‘budgets for reporting. We also identify four internal reasons-for-
development projects” as the speed of adoption measure. adoption, two of them proactive: managers’ back-

4
326 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347

ground, focus the attention of the organization on Conceptual underpinnings and literature linkages
executing the strategy (need to focus), and two reac-
tive: react to problems, and code learning (associ- Fig. 1 provides an overview of our three research
ated with formalizing repetitive yet non-formalized questions and its links to the existing literature. This
processes). Our evidence also provides qualitative section explores those links to highlight our contri-
data consistent with MCS roles identified in the butions. The initial arguments in the literature
literature including stimulating dialogue and idea (‘Limitations of management control systems’) saw
creation; controlling execution through diagnostic MCS as constraining innovation and therefore
systems; and stabilizing an environment that, by incompatible with it. Over time, evidence started
the nature of the innovation process, is already rich to accumulate on the benefits of MCS on innovation
in opportunities. (2) Using these six new reasons for- (‘Theoretical justifications for and empirical evi-
adoption categories, the study then examines the dence on the role of MCS in innovative settings’)
impact of different reasons-for-adoption on the and theoretical arguments emerged to explain this
speed of adoption of MCS. We find that managers’ evidence. To address our research questions, ‘Rea-
background is associated with the fastest time- sons/events and roles of MCS in product develop-
to-adoption. (3) Finally, we find that MCS adopted ment’ reviews the different roles of MCS proposed
to code learning or because of the managers’ in the literature. These roles are contrasted against
background is associated with better on-time the reasons-for-adoption that we identify from our
development (an important dimension of product empirical evidence. ‘The emergence of MCS in
development performance). entrepreneurial companies’ presents recent empiri-
The next section (‘Conceptual underpinnings cal work on the emergence of MCS in entrepreneur-
and literature linkages’) of this paper reviews the ial companies. This work provides the framework to
literature on innovation and MCS. We highlight analyze our two additional research questions: the
how the literature has emphasized the different relationship between (i) reason-to-adopt and time-
roles that MCS may have; thus, what purpose do to-adoption (our second research question), and
MCS fulfill in organizations. Yet, these roles are (ii) reason-to-adopt and performance (our third
not necessarily associated with the reasons why research question).
MCS are adopted. These reasons are associated
with events that do not have a one-to-one relation- Limitations of management control systems
ship with the roles that these systems play. There-
fore, a particular event may trigger the adoption Management control systems can stifle innova-
of a system with a particular role in a certain envi- tion (Ouchi, 1979). If not designed to deal with
ronment and a system with a different role in uncertainty, they may constrain cross-functional
another environment. The reasons why systems interaction, limit communication to established
are adopted are distinct from the roles that the sys- patterns, penalize deviations, and diffuse leader-
tems ultimately play. This section also presents the ship. Mintzberg’s work (1976, 1979, 1994) high-
evolution of the literature from an initial notion lights the separation between planning and
that innovation and MCS were incompatible to managing as two separate processes, the first ruled
the new broader support for MCS having a positive by formal systems and the second more informal
role in promoting innovation. ‘Field research and decoupled from MCS. Quinn (1978) argues
design’ presents the research design. The study is that the restrictions imposed by organizations’ for-
based on a multi-case, multi-method research mal systems limit the innovation abilities to ‘‘logi-
design. We sent questionnaires and interviewed cal incrementalism.” Damanpour’s (1991) meta-
three managers in each of 69 high-growth technol- analysis of empirical work on organizational deter-
ogy entrepreneurial companies. This design allows minants of innovation reports a negative associa-
for triangulation of the data and combining quali- tion between innovation and formalization. These
tative and quantitative data to develop a frame- MCS are most useful when task analyzability is
work and use statistical generalization of the high and the number of exceptions is low (Perrow,
findings. The following three sections present the 1970) such as in low innovation settings. In high-
results to our three research questions. ‘Discussion’ innovation environments, they may diminish the
relates these findings to the existing literature and intrinsic motivation and freedom that innovation
discusses limitations and future research. requires (Amabile, 1998). Empirical studies have

5
 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347 327

confirmed these predictions. Abernethy and Lillis study setting. They describe how managers rely on
(1995) find that ‘‘spontaneous contact and ‘‘inte- an enabling use of these systems to cope with the
grative liaison devices” which allow regular, per- uncertainty of day-to-day operations.
sonal and intensive contact” are more prevalent The concept of adaptive routines (Weick, Sutc-
in flexible manufacturing firms while traditional liffe, & Obstfeld, 1999) describes routines as resilient
performance measurement systems are de-empha- because of their capacity to adapt to unexpected
sized. Abernethy and Brownell (1997) report higher events. This concept portrays routines as flexible
reliance on personnel control in research and to absorb novelty rather than rigidly to suppress
development departments. Rockness and Shields it. They offer organizational members a stable
(1988) echo these conclusions and find that the rel- framework to interpret and communicate when fac-
evance of budgets in R&D settings is highest for ing unexpected events. They ‘‘usefully constrain the
planning purposes and decreases monotonically direction of subsequent experiential search” (Gav-
for monitoring, evaluating, and rewarding. etti & Levinthal, 2000, p. 113). Reliability rather
While these arguments do not support the need than replicability identify routines in uncertain set-
for management control systems in innovative set- tings. Feldman and Rafaeli (2002) extend this argu-
tings, recent empirical evidence and theoretical ment to include routines as drivers of key patterns
arguments provide a different perspective. Innova- of communication among organizational members.
tion management appears to benefit from having a Miner, Bassoff, and Moorman (2001) describe the
balance between ‘‘tight” and ‘‘loose” controls to constant interaction between routine activities and
provide both the support and direction for innova- improvisation in product development. Routines
tion. The positive impact of these controls on inno- provide the background for improvisation to hap-
vation highlights the need to understand why these pen and learning to accumulate.
systems are adopted in processes where innovation These concepts highlight the positive effect that
has a relevant role. MCS may have on innovation. MCS are viewed as
flexible and dynamic frames adapting and evolving
Theoretical justifications for and empirical evidence to the unpredictability of innovation, but stable to
on the role of MCS in innovative settings frame cognitive models, communication patterns,
and actions.
Simons’ typology (Simons, 1995) identifies Evidence is accumulating of the positive effects
interactive systems as information-based routines from adopting MCS in uncertain settings. For
to identify knowledge required to address strate- instance, managing environmental uncertainty has
gic uncertainties. One attractive feature of the repeatedly been associated with MCS (Chenhall &
interactive systems concept is that it allows top Morris, 1986; Khandwalla, 1972; Simons, 1987).
management to guide the search stage of the Chenhall and Morris (1995) identify extensive use
innovation process, without falling into the cyber- of management accounting systems with superior
netic model. Recent empirical studies (Abernethy performance in companies following entrepreneur-
& Brownell, 1997; Bisbe & Otley, 2004) rely on ial as opposed to conservative strategies. Henri
interactive systems to examine MCS in uncertain (2006) finds that the dynamic tension that emerges
environments. as systems are used both in a diagnostic and interac-
While interactive systems speak to the front end tive way is associated with improved performance.
of the innovation process, the concept of enabling Chapman (1998) used four case studies to conclude
bureaucracy (Adler & Borys, 1996) addresses the ‘‘that accounting does have a beneficial role in
role of MCS throughout the stages of assimilation highly uncertain conditions” (p. 738). Howard-
and exploitation of knowledge. Enabling bureau- Grenville (2003) used an ethnographic approach in
cracy is designed to ‘‘enhance the users’ capabilities one high-technology company to document the rel-
and to leverage their skills and intelligence” (p. 68) evance of organizational routines to confront uncer-
rather than with ‘‘a fool-proofing and deskilling tain and complex situations.
rationale.” Thus, organizations assimilate and exploit Within product development, prescriptive recom-
the knowledge accumulated in the first stage through mendations to practitioners emphasize the impor-
flexible, transparent, user-friendly routines. Ahrens tance of MCS (McGrath, 1995). Several research
and Chapman (2004) apply the concept of enabling studies have found that planning and well-coordi-
bureaucracy to analyze the role of MCS in a field nated project execution are associated with product

6
328 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347

success (Cooper, 1995; Zirger & Maidique, 1990). adoption as their main research variable.10 They
These studies hint to a role of MCS although they find that faster adoption of these systems is associ-
fail to provide a theoretical justification for their ated with the presence of venture capital funding,
findings (Brown & Eisenhardt, 1995). Using a con- size, age of the firm, and the replacement of the
trol framework, Cardinal (2001) found that the founder as CEO. Sandino (2007) examines how
three types of control systems – input, behavior, time-to-adoption of MCS in retail firms depends
and output control – enhance radical innovation – on their strategy given a set of common systems
arguably the most uncertain type of innovation. being adopted. Cardinal et al. (2004) describe
Davila (2000) reports a positive association between how the control approach evolves from informal
the use of management accounting information to formal in a case study of a growth firm. These
and product development performance. Brown studies look at the emergence of MCS and they
and Eisenhardt (1997) describe successful product rely on time-to-adoption as their proxy to address
innovation as blending ‘‘limited structure around this question.
responsibilities and priorities with extensive com- Organizational characteristics such as age, strat-
munication and design freedom” so that ‘‘this com- egy, lifecycle stage, or size are covariates in vari-
bination is neither so structured that change cannot ous studies; yet the co-variation between time-to-
occur nor so unstructured that chaos ensues” (p. 1). adoption and organizational variables is mediated
This evidence highlights the importance of under- by events (reasons-for-adoption) that are absent in
standing MCS in product development. these studies. Simons (1995) provides arguments
indicating that a potential event leading to MCS
The emergence of MCS in entrepreneurial companies adoption is a breakdown in processes – such as
failure to meet deadlines or quality problems. In
The previous sub-sections argue for the this case, the reason-for-adoption is an organiza-
importance of MCS to the innovation process tional failure, an event that can be associated to
from theory and evidence perspectives. This different MCS roles – such as either a making
sub-section reviews another stream of literature goals explicit role or a coordination role. In other
relevant to our research question – the literature instances, MCS may be triggered by an event
on the emergence of MCS in entrepreneurial associated with one MCS role such as adopting
companies. MCS to legitimize the company vis-à-vis potential
Our research questions examine the adoption of customers. However, no explicit arguments have
MCS in a particular innovation process, product been provided to answer the question of why are
development. Thus, both literatures offer relevant MCS adopted (Cardinal et al., 2004). One objec-
frames of reference. While the innovation literature tive of this paper is to answer this question
is relevant to the need for MCS in product develop- grounded in the qualitative evidence that the data-
ment, the MCS emergence literature examines in base provides.
more detail variables associated with how fast these As described in the previous paragraph, prior
systems are adopted. These ideas are most relevant literature on the emergence of MCS has focused
to our second and third research questions relating on two main variables: time-to-adoption and orga-
time-to-adoption and performance to reasons-for- nizational performance. Time-to-adoption mea-
adoption. sures the speed to MCS adoption and thus is
Greiner (1972)’s growth model describes a first relevant to understand the phenomenon. Organiza-
transition (labeled crisis of control) where he tional performance addresses the issue of whether
argues for the adoption of MCS. At the end of MCS adoption is relevant. Davila and Foster
the first growth phase informal management no (2005, 2007) find that faster adoption of budgets
longer works and MCS are required. Moores and and of a broader measure of MCS adoption is
Yuen (2001) provide initial evidence on the emer- associated with faster company growth as mea-
gence of MCS. They find, as they expected, that sured by number of employees. This result suggests
management systems are adopted in the growth that faster adoption is relevant to organizational
stage of the firm. Davila (2005) in his study of performance. While this result is important, it does
the emergence of HR systems and Davila and Fos-
ter (2005, 2007) in their research of management 10
They define time-to-adoption in a similar way: time from the
accounting systems’ adoption focus on time-to- inception of the company to adoption of a particular system.

7
 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347 329

not address the question of whether different rea- important that whoever sets the goals also
sons-for-adoption are associated with faster adop- makes them clear to the organization and that
tion; it also does not address the question of these goals remain stable for a meaningful per-
whether these reasons are related to performance. iod of time” (p. 80). Uncertainty provokes a
While no theory exists to guide our priors, the constant shift of priorities that may undermine
empirical evidence provides an exploratory answer the innovation process. MCS typically state
to these questions.11,12 explicit goals. This increases their stability
and visibility, facilitates convergence in mean-
Reasons/events and roles of MCS in product ing across organizational actors, and provides
development the clarity that creativity is argued to require.
(2) Code learning from past. Lundberg (1995)
Existing knowledge on MCS focuses on the roles indicates that procedures help innovation by
that these systems fulfill; in other words, why these coding learning from past experience (Levitt
systems exist. Yet, understanding the reasons why & March, 1988). Coded routines facilitate
these systems are adopted in the first place is the diffusion across the organization and over
another important aspect of MCS research. This time of organizational capabilities (Nelson &
paper presents new evidence on the reasons MCS Winter, 1982).
are first adopted within the context of product (3) Help coordination. Lundberg (1995) also points
development. This is a different lens to view MCS out the importance of coordinating different
than the extensive literature on the roles that MCS innovation efforts across the organization.
can play. We show that in some cases the reason MCS decouple the efforts of organizational
for adoption is unrelated to a particular role. More- actors and reduce coordination costs through
over, the same reason for adoption can lead to the the explicit negotiation of local goals.
adoption of MCS with different roles across compa- (4) Plan the sequence of steps. Process planning
nies. To fully position the reasons-for-adoption that clarifies the sequence of steps to achieve cer-
emerge from the empirical study in this paper, the tain organizational goals and provides a blue-
following paragraphs describe the MCS roles that print for coordinating the innovation effort
the existing literature identifies: over time (Cohen et al., 1996).
(5) Promote accountability and facilitate control.
(1) Make goals explicit and stable. Amabile MCS facilitate control by exception (Simons,
(1998) indicates that innovation is enhanced 1995) where managerial attention is required
when people are granted freedom to achieve only if innovation results deviate from
goals that are clear and stable for a sufficiently expectations.
long period of time. She states ‘‘it is far more (6) Contract with external parties. External con-
stituencies, such as partners, may impose
11
While there are no theoretical models to guide our priors, MCS to enhance their monitoring within the
their future development is important because there is no clear firm. These intermediate milestones facilitate
directional prediction. For instance, it might be argued that MCS contracting with outside partners (Powell,
that are adopted because of managers’ prior experience will 1998). Pfeffer and Salancik (1978) highlight
emerge before MCS that code existing learning; because manag-
ers do not have to go through the learning process. However, the
the relevance of the external context in
opposite argument can hold: coding existing learning can lead to explaining how firms are organized.
faster emergence because the hiring of managers’ with prior MCS (7) Symbols to legitimize. Finally, new institution-
experience may only happen when the company codes and alism (Carruthers, 1995; Powell & DiMaggio,
realizes the limitations of the MCS that it has put in place. 1991) views cognitive processes as relevant to
12
The relationship between reasons for adoption and perfor-
mance also has no clear directional predictions. For instance, if
explain management systems. It identifies
the knowledge about how to design MCS is not in the company formal processes as symbolic to externally
(for instance, when systems are designed as a reaction to legitimize the innovation process of the orga-
problems), the design is likely to be worse than if the knowledge nization through an appearance of compe-
exists (for instance, when systems are designed based on the prior tency. Management systems do not fulfill as
knowledge of managers) and impact performance. Alternatively,
MCS developed as a reaction to problems may be better adapted
much a technical need as conforming to
than MCS developed based on an experience different from the external demands decoupled from technical
particular needs of the company. needs.

8
330 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347

Appendix A illustrates these different roles of transition point is likely to be a recent event in the
MCS using quotes from interviews with executives life of these firms and thus managers are expected
of our early-stage entrepreneurial company sample. to be able to better articulate the reasons-for-adop-
These quotes reinforce the importance of these dif- tion (as well as identify the time of adoption).
ferent roles. ‘Results for the adoption of manage-
ment control systems in product development’ Data sources
presents the reasons-for-adoption identified in the
empirical study and compares them to the MCS The three main information sources for each com-
roles described in the literature. pany are public data, three questionnaires, and three
interviews. We first collected as much information as
possible from public sources – such as company’s
Field research design
web pages and press releases from Lexis-Nexis. This
information was used to familiarize the research
To capture the richness needed to explore why
team with the characteristics and products of each
companies adopt MCS in the product development
company. Next, each company received three ques-
process, we adopt a cross-sectional, multi-method,
tionnaires – one for the CEO, another for the CFO,
multi-case field research design. The aim of the
and a third one for the business development man-
cross-sectional multi-case design is to gather a large
ager. The purpose of the questionnaire was to collect
enough variation to probe our research questions,
quantifiable information about the company and its
to capture the detail required to answer the ques-
processes – for instance, whether product develop-
tions, and to link contextual variables to the adop-
ment projects were early, on-time, or late as a mea-
tion of MCS. The multi-method design relies on
sure of product development performance or the
qualitative data to identify patterns of behavior
time of adoption of each of seven product develop-
and quantitative data to examine covariates that
ment management systems. The questionnaire was
may inform the research questions. Our data
sent in advance to have additional information to
include questionnaire and semi-structured inter-
prepare the interviews, to provide managers with
views on the adoption and role of MCS in young
guidance of what the research was about, and to have
technology companies. Capturing the quality,
the opportunity to clarify any questions regarding
depth, and richness to understand the experience
questionnaire’s answers. Appendix B reproduces
of the actors (Seidman, 1998) demands detailed
the relevant questions for this study.
descriptions of the phenomena (Kvale, 1996; Mar-
This use of questionnaires is somewhat atypical
shall & Rossman, 1995). The focus of the study on
in the literature. Rather than having the question-
product development as a relevant aspect of innova-
naire measuring the underlying variables using psy-
tion processes drives the decision to sample among a
chometric tools, it was used to collect factual data.
population of technology companies. We expect
Because the variables in the research had not been
product development to be a significant enough
previously used in an empirical study, interview
aspect of their strategy to have received manage-
data were more appropriate to identify them. The
ment attention. The sampling of a population of
objective was to infer these variables from the
young firms is intended to capture the point in time
descriptions provided in the interviews and examine
when formal systems, if any, are adopted. Managers
whether they matched the theoretical variables.
in young firms are more likely to identify the process
The final phase of the data collection included
that led to the adoption these systems were recently
separate semi-structured interviews with the CEO,
adopted compared to more mature firms where
CFO, and business development manager of each
MCS have been around for several decades.13 This
company. The objective of these interviews was to
13
The sampling of young firms has to do with the focus of the gain detail about the company, its history, its strat-
research on the adoption of MCS. Managers of younger firms are egy and the adoption, design, and use of MCS. Each
more likely to have lived through the adoption of these systems. interview lasted between forty-five and ninety min-
An alternative is to examine the adoption of accounting utes. Over 200 such interviews were conducted for
innovations (such as Beyond Budgeting, ABC, or Balanced
the sample examined in this paper. The interviews
Scorecard) in mature firms (Innes, Mitchell, & Sinclair, 2000) or
the evolution of MCS as a reaction to changes in the objective relied on a detailed protocol listing the questions
and power structure of external constituencies (Oakes et al., to be addressed. The protocol insured that the main
1998). topics of the research were systematically covered

9
 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347 331

during the conversation, but the semi-structured technology firms because of their potential rele-
nature of the interview gave the flexibility of follow vance as a growth industry. This sub-sample was
up questions to clarify the particular practices at extended using three additional databases particular
each company (Marshall & Rossman, 1995). The to the industry: Rich’s High-Tech Business Guide to
relevant protocol questions are reproduced in Silicon Valley and Northern California (2000/2002),
Appendix C. Interviews were conducted in person BioScan (Oct. 2001), and the US Business Browser
or by phone and at least two researchers were pres- (c. 2001).
ent in every interview. Interviews were taped and A letter addressed to the CEO was sent to every
then transcribed. The questionnaire, sent prior to firm in the sample. The letter described the purpose
the interview, facilitated focusing the interview of the research, the research process, and the bene-
around the key aspects of interest. fits of participating – a half-day conference where
participating companies were invited to a presenta-
Sample description tion of the managerial implications of the research
project and a written document of the findings.
The final sample includes 69 young, high technol- The letter was followed up with a phone call to
ogy companies. We construct the sampled popula- entice participation; a company was dropped from
tion using the following selection criteria: (1) high the sample if it had not accepted or declined partic-
technology, (2) less than 10 years old, (3) between ipation after five phone calls. The sample selection
50 and 150 employees, (4) independent, and (5) in process is detailed in Panel A, Table 1. Excluding
a limited geographic area.14 These criteria identified companies that were acquired, went out of business
companies where product development is likely to or are ineligible, the response rate is 20%.16 Compa-
be a foundation of their competitive advantage. nies acquired or that went out of business were sig-
They also identify companies more likely to have nificantly younger than the eligible sample but
recently and independently transitioned through comparable in terms of sales and number of
the stage of formalizing product development pro- employees. Within the eligible sample, we compared
cesses, rather than companies that have had systems companies that participated to those that did not, in
in place for a long time or systems imposed by a order to assess potential self-selection bias; we
parent company. We did not restrict firms to be found no significant differences in sales, number of
public or private, foreign owned, or venture funded; employees and age.17 The final sample includes 11
however the majority of firms were private, domes- biotechnology companies, 48 information technol-
tically owned, and venture funded. The population ogy companies, and 10 companies in other indus-
of firms was sourced from the CorpTech Internet tries; in addition 59 received venture capital. Panel
directory of technology companies. We accessed the B of Table 1 provides additional descriptive statis-
database in January and June 2002 and build our tics on the sample.
sample from the following industries (using Corp-
Tech industry classification): biotech (BIO), com- Data analysis
puter hardware (COM), manufacturing (MAN),
medical equipment (MED), pharmaceuticals The analysis of why firms adopt MCS for prod-
(PHA), photonics (PHO), computer software uct development was structured in two stages. In
(SOF), subassemblies (SUB), test & measuring the first stage, interview data were coded – to sum-
equipment (TAM), and telecommunications
(TEL).15 We also purposefully over-sampled bio- 16
The 20% is a conservative estimate. It assumes that all the
companies that did not respond (to the five telephone contacts)
14
The main reason for the geographic criterion was research (which means that they did not pick up the phone) were eligible.
funding (more than 50% of the interviews were done at the The percentage of companies that participated over the percent-
companies’ premises). While this decision reduces the potential age of companies that were contacted is 38%. The demands on
impact of omitted variables that may vary with geography, it company resources were much higher than a traditional mail
limits the generalizability of results. survey; we asked each company to fill three questionnaires and
15
We excluded from these lists any companies that were also have separate one-hour interviews with each of three different
listed as ‘‘Energy,” ‘‘Environment,” ‘‘Chemical,” ‘‘Defense”, managers.
17
‘‘Transportation” or ‘‘Non”. ‘‘Non” companies are not primarily We compared means and medians of sales, employees, and
high-tech companies. The other industries are excluded because age (variables available from the databases that we accessed) for
they face a different regulatory and/or institutional environment. both groups in our sample. We also use the non-parametric
We also excluded organizations cross-listed in these industries. Mann–Whitney test on these variables with identical conclusions.

10
332 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347

Table 1 following a structured process. The sample was

Early-stage entrepreneurial company sample: sample construc- divided into two groups. The three researchers inde-
tion and descriptive statistics on final sample
pendently identified the main topics covered in each
Panel A: sample construction interview for one of the groups. The result was the
Companies in the initial database 624
Minus
dissection and reorganization of the original tran-
Companies that went out of business 94 scripts into broad topics. Then, the researchers
Companies acquired 88 met to contrast the topics that each one identified,
Companies ineligible in some other waya 102 agree in a common set of terms to identify them,
Companies that did not respondb 158 and discussed any differences in interpretation of
Companies that declined participation 113
Final sample of companies 69
the transcripts. Next, the second group of interviews
was independently coded using the common termi-
Mean Std. Quartile Median Quartile nology. Finally, at the end of the process the team
dev. 1 (0.25) (0.50) 3 (0.75) met to contrast the results of this second coding
Panel B: sample descriptive statistics on final sample effort and discuss differences and new topics, if
Number of 1.73 0.80 1 2 2 any. The objective of sub-dividing the sample into
CEOs two groups is to contrast the model that emerges
Age (in 7.42 2.33 5 7 9
years)
from the first analysis using a hold-up sample. The
Employeesc 119 62.6 72 114 160 same process was iteratively used to analyze with
R&D 0.39 0.26 0.16 0.38 0.61 increasing detail each of the topics until a stable
intensity set of constructs were identified that explained the
(%)d phenomenon examined. The process evolved in an
Revenues 10,691 11,711 2468 7140 15,156
($‘000)e
iterative and non-linear fashion where the topics
Income 10,175 15,598 12,139 5400 18 and constructs where revised to better capture the
($‘000)e insights of the independent analyses. The end result
Funding 52,441 59,865 8963 39,300 72,750 is a set of typologies that describe different aspects
($‘000)f of the adoption of MCS in product development
a
These companies are too small, too old, subsidiaries of other (Marshall & Rossman, 1995).
companies, or with no significant product development activity. The second stage of the analysis combines the
b
These are companies that did not respond to the five tele-
quantitative data collected through the question-
phone contacts.
c
Number of employees is calculated at the peak of each com- naires with the qualitative data derived from the
pany’s size. analysis of the interviews. The product develop-
d
R&D employees are estimated as a percentage of total ment systems included in the questionnaire were
employees defined as the sum of R&D employees for each of the selected based on a review of the product develop-
years reported divided by the sum of total employees for those
ment literature (such as Ulrich and Eppinger
same years. Only companies that reported R&D employees are
included. (1995)). Those systems more frequently cited were
e
Revenues and income are for the last year of data available. included in the questionnaire. The unit of analysis
f
Funding is the total external funding for each company. is the company; therefore each observation com-
bines qualitative-interview data with quantitative-
questionnaire data. Interviews and questionnaires
marize, interpret, and classify the information. To were designed to collect different types of data
limit the potential bias inherent in the analysis of from the companies. The objective of analyzing
qualitative data, three researchers coded each one these two sources of data is to establish patterns
of the interviews. To systematically proceed through leading to a framework of reasons-for-adoption
the coding process, each researcher used the Nvivo (Strauss & Corbin, 1990). The final dataset com-
qualitative coding software. This software details bines survey data with the variables identified in
the analysis from the raw data to the theoretical the coding of the qualitative data to propose rela-
propositions, thus providing an auditable trail of tionships among these variables (King, Keohane,
the analysis. Because of the exploratory nature, each & Verba, 1994). The findings reported in the paper
researcher may potentially identify different con- are the end process of the analysis; however the
structs that explain the observed patterns. To iden- audit trail documentation allows tracking the con-
tify common constructs, the coding was done clusions to the raw data.

11
 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347 333

Table 2
Early stage entrepreneurial company sample: product development descriptive statistics
Companies adopting (%) Time-to-adoption (years since company founding)
Mean Quartile 1 (0.25) Median (0.50) Quartile 3 (0.75)
Panel A: descriptive statistics on product development systems
Project milestones 81 2.70 1 2 5
Reports comparing actual progress to plan 71 2.88 1 3 4
Budget for development projects 68 3.02 1 2 5
Project selection process 65 3.44 1 3 5
Product portfolio roadmap 61 3.36 1 3 5
Product concept testing process 45 2.74 1 2 5
Project team composition guidelines 39 3.00 1 2 5

Type of measures Respondents Updating frequency (per month)

Mean Quartile 1 (0.25) Median (0.50) Quartile 3 (0.75)
Panel B: descriptive statistics on product development measures
Time 62 2.81 1 4 4
Budget/financial 30 2.36 1 1 4
Product functionality 30 2.08 0.83 2 4
Customer 24 1.78 1 1 2
Quality 15 2.90 2 4 4
Panel A: Companies adopting each system are the percentage of companies that had adopted the system at the point of data collection.
Time-to-system is the mean number of years since founding to the adoption of the particular system for those companies that adopted each
system.
Panel B: The table reports the number of respondents that listed each type of measure among the three top measures for managing product
development and the frequency (times per month) that the measures are updated.

Results for the adoption of management control cept testing process. These systems are likely to be
systems in product development adopted later because they require having various
products considered in the development plan; some-
Table 2 (Panel A) provides descriptive statistics thing that is likely to happen later on in the life of a
on the percentage of companies adopting each of company, once the initial product has been devel-
the product development systems identified for this oped and released to the market.19 Fig. 2 shows
research and the time-to-adoption since the found- the adoption time evolution; it plots the percentage
ing of the company (Appendix B reproduces the rel- of companies that had adopted each of the systems
evant questionnaire items). Project milestones are as a function of company age. Fig. 3 plots the same
the system most companies have adopted and the information but against company size (proxied by
fastest to adopt. Time-to-adoption is significantly number of employees).
different from the other systems at least at the Panel B in Table 2 summarizes the results to the
10% (two-tailed t-test) in pair-wise comparisons,18 open-ended question ‘‘what are the three most
except for product concept testing process where important measures that top management uses to
the difference in time of adoption is not significant. evaluate the progress of a development effort?”
Companies that adopted product concept testing included in the questionnaire. The five measures’
process, while relatively fewer (45%), did so earlier descriptions were coded by two researchers indepen-
in their lives. Project selection process is adopted dently. Panel B in Table 2 gives the frequency distri-
significantly slower than all other systems except
product portfolio roadmaps. Roadmaps are also
19
slower than the other systems except product con- We also tested for differences in percentage of adoption.
‘‘project milestones” is significantly different from all other
systems except ‘‘comparing actual progress to plan” at least at the
18
We also performed an ANOVA test that was significant at the 5% level (two-tailed). ‘‘Comparing actual progress to plan,”
1% level. We also tested for differences using Wilcoxon matched- ‘‘budget for development projects,” ‘‘project selection process,”
pairs signed-ranks test and testing whether the median of the and ‘‘product portfolio roadmap” are significantly different from
differences were different from zero. The conclusions remained ‘‘project team composition guidelines” and ‘‘product concept
are the same as those using the parametric t-test. testing process” at least at the 10% level (two-tailed).

12
334 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347

0.8

Percentage of Companies Adopted

0.6

0.4

0.2

0
0 1 2 3 4 5

Company Age (years)

Project Milestones Actual Progress to Plan Project Budget
Project Selection Product Portfolio Roadmap Product Concept Testing
Project Team Composition

Fig. 2. Early-stage entrepreneurial company sample: evolution of product development management control systems over time. Each of
the seven systems (project milestones, reports comparing actual progress to plan, budget for development projects, project selection
process, product portfolio roadmap, product concept testing process, project team composition guidelines) is coded yearly as 1 (if the
company reports having adopted the system) and 0 otherwise. For each year since funding (x-axis) we add the number of companies that
have adopted the system over the total population to obtain the percentage in the y-axis.

0.8
Percentage of Companies Adopted

0.6

0.4

0.2

0
<10 <20 <30 <40 <50 <60 <70 <80 <90 <100 <110 <120 <130 <140 <150

Company Size (Number of Employees)

Project Milestones Actual Progress to Plan Project Budget
Project Selection Product Portfolio Roadmap Product Concept Testing
Project Team Composition

Fig. 3. Early-stage entrepreneurial company sample: evolution of product development management control systems over size (number of
employees). Each of the seven systems (project milestones, reports comparing actual progress to plan, budget for development projects,
project selection process, product portfolio roadmap, product concept testing process, project team composition guidelines) is coded
yearly as 1 (if the company reports having adopted the system) and 0 otherwise. For different number of employees (x-axis) we add the
number of companies that have adopted the system over the total population to obtain the percentage in the y-axis.

bution for the 69 companies of the types of measures sures are reported by 24 respondents (34.8%) and
used to track product development and how often quality-related measures 15 respondents (21.7%).
they are updated. A time-related measure is the most The iterative analysis of interview data identified
frequently used with 62 respondents (89.9%) refer- six different reasons-for-adoption of MCS. While
ring to it with updates averaging 2.81 times per we began our data analysis aware of the seven
month on average. Budgets and product functional- MCS roles identified in the literature (‘Reasons/
ity related measures are mentioned by 30 of the events and roles of MCS in product development’),
respondents (43.5%) while customer-related mea- the iterative analysis quickly converged to a

13
 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347 335

different set of categories. These categories better observations where it was a secondary reason-
described the experiences described in the interviews for-adoption.
as to why systems are adopted. As described in In certain companies, systems are adopted to
‘Data analysis’, our coding process was exploratory legitimize the company vis-à-vis external partners
without pre-imposing a set of categories. (Carruthers, 1995). They are not required from an
Panel A in Table 3 provides descriptive statistics internal management perspective, but they are
and illustrative quotes for each of the reasons-for- adopted as a symbol (Macintosh, 1994, chapter 9)
adoption. The number of firms using each type of to enhance the credibility of the company towards
reason-for-adoption and the importance of the rea- external parties – usually customers, partners, or
son-for-adoption as described in Table 3 are20, 21: investors.22 Establishing a framework for the inter-
action with external parties is another external rea-
Main Secondary son-for-adoption of MCS. External parties require
reason-for- reason-for- visibility into the organization’s processes to moni-
adoption adoption tor, coordinate, and control them (Pfeffer & Sala-
1. External reasons- ncik, 1978). Inter-organizational agreements often
for-adoption lack the constant interactions required to ground
Legitimize 2 9 informal management and they may need to formal-
Contract 7 8 ize this interface. We labeled this situation as con-
2. Internal reasons-for- tracting to indicate the relevance of formal systems
adoption to implement the contract between players (Dekker,
Proactive-manager 24 11 2004; Otley & Berry, 1994).23
background We also found much evidence consistent with
Proactive-need to 9 10 internal reasons for MCS adoption.24 Managers
focus may be proactive in that the systems are imple-
Reactive-chaos 11 3 mented ahead of the company requiring them. We
Reactive-learning 7 0 distinguish two different proactive reasons: manager
3. Kept approach 9 background and need to focus. The first one captures
informal those systems put in place when a particular man-
ager comes to the firm; the MCS is adopted because
of this person’s prior experience. The idea of import-
These reasons-for-adoption are not exclusive of in knowledge with the hiring of a manager domi-
each other, and more than one can be present nates this category. Rather than have a incumbent
at different stages or for different systems within managers design the system, a new manager is hired
a company. For each reason-for-adoption, we re- that brings this knowledge with her; a process
port the number of observations where it was known in the literature as grafting (Huber, 1991).
the main reason-for-adoption and the number of Background is closely associated with congenital
20 22
During the coding process, it became clear that more than one This role is consistent with the theoretical role number seven
reason-for-adoption was present in several of the adoption described in ‘Reasons/events and roles of MCS in product
events: interviewees described several forces around the event. development’ and Appendix A.
23
Rather than force each adoption event into a single category, we This role is consistent with the theoretical role number six
decided to identify a main reason-for-adoption as the most described in ‘Reasons/events and roles of MCS in product
important one and a secondary reason-for-adoption as also development’ and Appendix A.
24
present; but with less strength than the first one. We believe that In contrast to external reasons-for-adoption of adoption,
this coding approach better reflects the data. internal reasons-for-adoption do not directly map into the MCS
21
Legitimize, contract, and learning are reasons-for-adoption roles identified in Section ‘Reasons/events and roles of MCS in
associated with particular roles and labeled using the role that product development’ and Appendix A. During the coding
they fulfill. Diverse events may trigger the adoption; for instance, process, the categories that appeared to better describe the data
legitimize may be demanded by customers, suppliers, or regula- were not as much the particular roles that the systems adopted
tors. Manager background, need to focus, and chaos are were intended to fulfill as the reasons why they were adopted (the
associated with particular events (such as hiring a new manager) situation that led to the adoption). While this coding better
and labeled using these events. For instance, manager back- reflects the underlying data, the interviews provide illustrations of
ground may lead to the adoption of MCS that makes goals MCS’ roles consistent with those roles in the literature. Appendix
explicit, helps coordination, or plans the sequence of steps (or a A is included to illustrate each of the theory-based roles in our
combination of these goals). sample.

14
336 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347

Table 3
Early stage entrepreneurial company sample: adoption of management control systems in product development
Panel A: typology of reasons-for-adoption
External reasons
Legitimize ‘‘We’re also finding that the customers get a feeling of control when you give them more data (...). By
Main reasona: 2 seeing data that they’ve never seen before, we look good compared to the internal IT, which is one of our
Secondary reason: 9 strong competitors”
‘‘That was almost the number one question (how do you go about developing this stuff) from both
audiences – people we were bringing on, and our customers (. . .) and investors actually. Pretty almost
everyone. If you don’t have it, nobody’s comfortable with it. So again, I think it’s just like when you want
to buy a house. The first thing you want to do is inspect the foundation of the house”
Contract ‘‘Pfizer, as part of the agreement, put in place milestones that we had to achieve”
Main reason: 7 ‘‘When we won the Motorola contract, Motorola forced us to get our act together and so that was a
Secondary reason: 8 forcing function for all these (product development) processes. Motorola insisted on proper program
management, proper change control, proper project reporting, monthly business reviews, monthly
project reviews, etc. and they sent audit teams in to audit where we were and make specific
recommendations.” ‘‘We tie most of our payments to milestone payments”
Internal reasons: proactive
Manager background ‘‘Both of us had come from big company backgrounds where we had used project tracking systems, it
Main reason: 24 seemed natural (. . .), even though it was a very small amount that we were tracking, we were just used to
Secondary reason: 11 it. That was the normal thing to do. I’d have to say that we’ve had project tracking systems, red–yellow–
green status reports, and so forth, from the very beginning”
‘‘The founder and the initial engineering people knew engineering and knew that you can’t just have a
free-form, ‘‘let’s go be random”, there has to be some real milestone way of tracking it. (. . .) So when I
got there, the engineering team was a very competent team, very good architects, very hard-working.
They had a real desire for process: code reviews and configuration management and that was good”
Need to focus ‘‘Now we’re getting to the place where the return is lower compared to the investment, right? Because
Main reason: 9 we’ve already taken up the low hanging fruit. So now we’re having to use stepladders to go get fruit. And
Secondary reason: 10 now it makes more sense to be measuring exactly how many grapes am I getting for how much effort
moving the stepladder”
‘‘I did it (formalization) more to share my thinking with a bigger team, six, seven people”
‘‘I think looking back at it, and comparing it to other companies I’ve been working for, I think the
toughest step for a company to go through is the step from going from a small company where pretty
much everyone knows the main goals, the main focus and the main initiatives within the company, to the
steps where you have to stop passing the whole business and some people will know some, some people
will know others, but not everyone sees the big picture. And I think it’s a very crucial, it’s very easy to
start having a lot of people running in many different directions. And the difficulty at that stage is to keep
the whole energy focused. It’s like moving from a two-person boat where it’s easy to make sure you
paddle in the same rhythm, to a 10-person boat where suddenly you have to spend more time on making
sure it’s the same rhythm than anything else”
Internal reasons: reactive
To chaos ‘‘The original engineering team would give me dates when they were going to have certain things done
Main reason: 11 and never make the dates. They didn’t communicate amongst themselves, so even when they released
Secondary reason: 3 something, it didn’t do what they said it was going to do. There was no QA process, so when they did
release it, it continued to be buggy, they would introduce more bugs than they would fix”
‘‘We were picking projects in such a strange way. Whatever came in, displacing something else. Sort of
chaotic”
‘‘We wanted to formalize it, but the skill set wasn’t there, and also the realization that we needed that
wasn’t there. So I couldn’t push it too much. I said, okay, for now let’s focus on the other areas. (. . .) The
reason it’s happening now, actually this month, is because the last release we did it dragged out to a
month’s cycle, and it was very painful”
‘‘I think generally you find that a big change occurs when something goes wrong, and something did, and
so big change occurred”
To learning ‘‘It was around here that things became less ad hoc because of experience and documentation of process,
Main reason: 7 retaining history of what worked and didn’t work, finally came to place. (. . .) By that time, we’d had
Secondary reason: 0 enough experience with enough projects to say, ‘‘Okay, now we get what works and doesn’t work. We
can estimate accurately.” And it just took doing it a number of times in this new industry and both
succeeding and failing to understand okay now we can generalize, we can systematize, we can come up
with methodologies that really work”
‘‘It was just sort of the solidification of what was a more informal sort of milestone assessment process”
(continued on next page)

15
 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347 337

Table 3 (continued)
Kept approach informal: 9 ‘‘We can still afford to be fairly fluid and flexible and informal in our approach, because our development
team has been with Kevin, every one of them, eight to 15 years”
‘‘I think we’re not at that stage yet. I think we are a small company, where on the resources side, it’s
pretty clear from the onset, when we go into a product, we pretty much know how long it’s going to take
and how much it’s going to cost us”
‘‘We thought we were going to have to fire him because he was such a creative guy and would always
work on something new. (. . .) Some people are prim donnas. And you have to put up with that, but they
have to be able to sing. And so the worst thing that can happen in a company, is, prima donnas that don’t
deliver. But if you have people who deliver, it happens every year, what they do in between doesn’t
matter. And I know we’ve had managers that will fire that kind of person, because they come in late, and
they do weird things, and they don’t make the budget. But if you’ve been around long enough you know
that, 80% of your innovation will come from crazy people”
‘‘We tried to formalize it in ’97, in ’98, in ’99, but we said it was not possible. So we left it at an informal
level. Part of my job through most of my life has been doing skunk work things in bigger organizations.
In a sense, that was kind of the flavor of what was going to work in the company”

Panel B: limitations of management control systems

‘‘I think to a certain extent processes could slow things down, and so being slow to market is worse than being fast to market. The
whole goal of a biotech firm is to bring a product to market. That’s life or death. And anything that gets you closer to market is good
and anything that keeps you from the market is bad”
‘‘Fortunately on the engineering side of it – given the amount of money we had, we did put in place a very formal development cycle
with all the tools in place. So all the revision control systems for hardware and software, all those tool chains are in place, and have
been in from day one in the company. And if anything, that’s cost me six months of product development time, so now I have second
thoughts about the so-called sophisticated tool chains. But, they’ve always been in place here actually from day one”

Source of MCS Illustrative quotes

design
Panel C: sources of management control systems’ design
Internal people ‘‘Some of the systems were driven by key individuals with—given their experience and background, they drove
those processes”
Number of ‘‘We have a guy who again brought the experience with him. Most of us that have been in big companies and
companies: 61 that’s most of my senior staff; either had the PLC or a gate process. But in particular our product line manager
director brought the template. And the trick of bringing something that’s been used in a different company is
making it relevant and the steps involved relevant to the size of the company”
‘‘So I have one of my guys write the whole document, it’s about an 80 page document, about the development
process”
‘‘I give (Our VP of engineering) credit for designing (the plan of record) here.”
‘‘Working with people who were experienced, like the sales guys were experienced in product development too,
because it’s their project being developed. (The founder) brought quite a bit of the infrastructure too. My support
guy was an older guy in his 50’s and he had 30 years of product development experience, so those guys all brought
in a lot to the table”
‘‘I brought it from my previous experience. I’ve done product development for about twelve years and you kind of
know what works”
External parties ‘‘For example, there’s a company that makes backup software. And I needed backup software because if the worst
possible thing happens, I want to be able to put it back to the way it was, which means I need a backup. So I don’t
want to create my own backup system. This company had a backup system”
Number of ‘‘Pfizer, as part of the agreement, put in place milestones that we had to achieve”
companies: 3
a
Main reason indicates number of companies where the particular reason-for-adoption was identified as the main one; secondary reason
indicates if it was also identified. For each company we identified at least one reason (except if the process was still informal) and at most
two reasons (a main one and a secondary one). Quotes are particular examples of each of the reasons.

16
338 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347

learning, where ‘‘individuals (. . .) have knowledge into existence in ’99 during the boom; we needed
about (. . .) the processes the organization can use better managed projects because we had so many
to carry out its creator’s intentions” (Huber, 1991, things going on. We were forced to execute these
p. 91). These managers, usually with significant prior projects efficiently so we could get to the next one.”
experience, perceive MCS as management infra- But in most cases, the formalization grew out of the
structure required to facilitate growth. Their behav- periodic enactment of an informal routine.
ior can be interpreted as mimetic (Powell & Panel A in Table 3 also identifies cases where an
DiMaggio, 1991), where they emulate practices from informal management approach was used and Gre-
other organizations to reduce the cognitive uncer- iner’s (1972) crisis of control had not yet happened.
tainty; although the narratives suggest a strong func- We limit the count to companies that explicitly men-
tionalist aspect. The second proactive reason-for- tioned this approach. The stated reasons for main-
adoption, need to focus, reflects managers who taining an informal approach include: (1) team has
implemented systems because they perceived an worked together for a long time and their informal
emerging need. In contrast to the manager back- interactions are well-understood but not coded, (2)
ground that leads to implementation with the hiring management team believed that formal systems
of a new manager regardless of the systems’ immedi- would kill creativity, (3) the organization was not
ateness, need to focus responds to a particular emerg- considered to be large enough to grant MCS, and
ing need – such as coordination of geographically (4) management team did not have the knowledge
dispersed workforce, increase organizational effi- to implement these systems.
ciency, or improve communication. Once adopted, MCS with few exceptions remain
MCS adoption may also be reactive to unex- as part of the management infrastructure and
pected events, mistakes, or recurring problems evolve. Frequently, interviewees described the sys-
(Simons, 1995). We label it as reactive to chaos. tems as ‘‘becoming more and more sophisticated.”
Lack of skills or lack of resources may delay the In a few settings, however MCS can be a time-
adoption of systems until constant failures in the bounded solution to achieve a certain objective.
informal processes force its adoption (Flamholtz, For example, a new CEO in the sample formalized
2000). In most circumstances, chaos was unin- the project selection process to focus the organiza-
tended and managers were ill-prepared to deal with tion and then dropped it as the organization under-
it; however in one of our observations the manager stood the new strategy.
purposefully used chaos to drive the need for Several managers interviewed gave instances of
change.25 MCS stifling innovation (Panel B, Table 3 for illus-
A final category that we labeled as reactive to trative quotes) and the need to adapt MCS ‘‘within
learning was the outcome of the enactment process the context of a company of our size (. . .) the mini-
(Weick, 1995). While the outcome was a more effi- mum that we need to accomplish without putting
cient organization, this category is different from artificial requirements, barriers, or roadblocks that
the proactive focus category in that formal systems slow us down.”
emerged as the outcome of a learning process. It was Panel C in Table 3 presents the descriptive statis-
not managers who decided to implement systems tics related to the source of knowledge used in
because of a particular need or prior experience, designing management systems. Internal managers
but systems emerged to code existing practices. In designed MCS in the majority of cases. Typically,
some cases, the coding was triggered by certain the design is a process of knowledge creation where
events. For instance, in one of the companies the tacit knowledge (Nonaka, 1990) spread over various
growth associated with the economic boom of the people in the organization is codified. External
late nineties triggered the coding of processes: ‘‘we designers are rarer and usually reflect the contract-
developed (the systems) over the years doing them ing process with partners or buying external tech-
ourselves, we know what we need to have; we have nology (software) to manage processes.
a project binder that has sections in it. (. . .) It came The categories of reasons-for-adoption that
emerge from the structured analysis of the inter-
25 views have the following pattern. Certain reasons-
This approach to implementing MCS was unique and in all
other cases, chaos was unintended. While an ‘‘outlier,” this
for-adoption are associated with the need to fulfill
observation is interesting as a different way to drive MCS one of the roles identified in the literature. In other
adoption. words, managers’ description associated the adop-

17
 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347 339

tion with the need to fulfill a particular role. While informal routine and capture (4) the learning associ-
various events may have triggered the adoption, ated with the routine. While several events may lead
the adoption itself led to this role. We label these to the adoption, managers identify the particular
reasons using the role that they are adopted for. role rather than the event as the reason why the sys-
External reasons-for-adoption are often associated tem was formalized. Finally, MCS may be adopted
with particular roles. Yet, other MCS are adopted to (5) legitimize the company vis-à-vis customers,
because of a particular event happened regardless partners, or suppliers and (6) contract with external
of the role of the system. Managers describe the parties.
adoption as linked to an event, while the role(s) of
the MCS subsequently adopted varied across com- Results for time-to-adoption and reasons for MCS
panies. We label this reasons using the event that adoption
triggered the adoption. Internal reasons are often
associated with particular events. Fig. 4 illustrates This section examines whether the various rea-
this argument. Previous literature has not greatly sons-for-adoption identified in ‘Results for the
focused on these reasons-for-adoption. Only adoption of management control systems in product
Simons (1995) identifies response to a (1) crisis or development’ are associated with different timing of
problem (our chaos category) as a ‘‘trigger” of adoption of ‘‘project milestones.” Timing is mea-
adoption. Our analysis of interview data identifies sured relative to the year the company was legally
this particular reason-for-adoption. But we identify established. The six reasons-for-adoption described
also five additional reasons-for-adoption. MCS can in Table 3 are typically triggered by an event –
emerge because of events such as (2) hiring a man- whether an external party demands to see the pro-
ager with a particular background (import in); this cesses in place, a partnership is entered, a new man-
person comes with knowledge of certain processes ager is hired, needs emerge, problems arise, or
that she deploys in the company. Another event informal practices are formalized. Therefore, the
triggering the adoption of MCS is the (3) decision adoption depends on whether a particular event
of a manager in the company with experience in happens and there is no clear directional expecta-
product development who decides that the process tion. While existing theory has not addressed the
has reached a stage where there is a need to adopt question of why systems are adopted (Cardinal
a particular system. In all these cases, systems may et al., 2004), we can make some preliminary predic-
be adopted to fulfill various roles (there is not a tions. First, companies that still keep an informal
one-to-one relation between the reason-for-adop- approach will report adopting systems later than
tion and the role of the system) such as make goals the rest of the companies in the sample. Second, it
explicit, help coordination, plan a sequence of steps, is also plausible that certain events that trigger the
or facilitate control. MCS also emerge because an adoption happen early in the life of the companies.
event requires these systems to fulfill a specific role. For instance, learning is likely to be due to the fact
For instance, MCS may be adopted to formalize an that there was no event external or proactive inter-

Reasons for MCS Adoption

External Internal
Proactive Reactive
Legitimize Contract Manager Need to Chaos Learning
Background Focus
Make Goals
Explicit and Stable
Code Learning
From the Past
Help
Coordination
Roles of Plan Sequence of
MCS Related
Steps
Promote
Accountability and
Control
Contract with
External Parties Related
Symbols to
Legitimize Related

Fig. 4. Reasons for MCS adoption and roles of MCS.

18
340 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347

nal that triggered the adoption of MCS; therefore, Table 4

we expect later MCS adoption to be associated with Early stage entrepreneurial company sample: product develop-
ment mcs formalization
the learning adoption reason.
A Cox specification is used to examine the poten- Categories of reasons Project milestones
tial effect of the various reasons-for-adoption on the Hazard ratio Robust std. error
time-to-adoption of MCS in product development; Panel A: time taken to adopt product development and reasons-for-
this is consistent with prior literature (Baron, Bur- adoption
ton, & Hannan, 1999; Hellmann & Puri, 2002). External
Legitimize 2.425*** 0.66
The Cox specification models the time to an event Contract 2.032** 0.65
(for instance, adopting a particular system). We Internal – proactive
examine the time taken to adopt the project mile- Manager background 3.573***,+ 1.23
stones’ MCS; this is the system adopted most often Need to focus 2.283** 0.75
in our sample (Table 2, Panel A).26 We separately Internal – reactive
Learning 1.589 0.98
asked the CEO and business development manager Chaos 2.496 ** 0.90
about the adoption date; kappa statistic of inter- Number of observations 388
rater agreement was significant at the 1% level.27 Number of companies 69
Table 4, Panel A reports the Cox results. Our depen- Chi-square 16.89***
dent variable is the time-to-adoption of project mile-
Categories of reasons Project on-time performance
stones as reported by the business development (0 = late, 1 = early or on-time)
manager. The conclusions are unchanged if we use
Coefficient Robust std. error
the date of adoption reported by the CEO. The inde-
pendent variables are dummy variables that take Panel B: on-time development performance and reasons-for-
adoption
value of one for those companies where a particular External
reason-for-adoption of adoption was identified (dur- Legitimize 0.203à 0.61
ing the interview coding) regardless of whether it is Contract 2.486** 2.22
the main or secondary reason-for-adoption. The Internal – proactive
coefficients reported are hazard ratios. A coefficient Manager background 0.767 ++,+++ 0.65
Need to focus 1.454** 0.73
on the independent variables greater (or less) than Internal – reactive
one means that higher values for the independent Learning 0.065à 0.99
variables are associated with shorter (longer) time- Chaos 2.730**, , 1.12
to-adoption.28 Manager background, chaos, legiti- Number of observations 64
mize, need to focus, and contract are associated with Pseudo-R2 0.22
Chi-square 13.57**
significantly faster adoption than the reference
group which still uses an informal approach. Except Panel A: *, **, *** significantly different at 10%, 5%, and 1%,
respectively, from reference category. +Significantly different at
for background of managers leading to faster adop-
10% from legitimize. Table reports Cox proportional hazard
tion compared to legitimizing (the coefficient for model of the time-to-adoption of project milestones. Dependent
management background – 3.573 – is significantly variables are the reasons-for-adoption of MCS adoption identi-
larger and therefore is associated with faster adop- fied in the qualitative analysis. The category not included is
tion than the legitimize coefficient – 2.425), we find ‘‘informal.”
Panel B: **Significantly different at 5% from reference category.
that the different reasons-for adoption do not lead +++
Significantly different at 1% from chaos. ++5% from contract
to faster adoption (except when compared to an and need to focus. Significantly different at 10% from learning.
informal management approach). A potential expla- Significantly different at 5% from legitimize. àSignificantly dif-
ferent at 10% from contract. The table reports a logit model with
product development performance as the dependent variable (1 if
26
We emphasize the time to adopt the ‘‘project milestones” the project is early or on-time, 0 if it is late). Dependent variables
MCS in our results as it is the most frequently adopted individual are the reasons-for-adoption of MCS adoption identified in the
system in our sample. Similar results are found for the ‘‘reports qualitative analysis. The category not included is ‘‘informal.”
comparing actual progress to plan” and ‘‘budget for development
projects” systems. nation is that the events that lead to adoption hap-
27
We also ran the same specification including a dummy for
pen randomly. A crisis, the need to focus, an
industry and controlling for the presence of venture capital. The
conclusions remained unchanged. external contract, or hiring a manager with product
28
The hazard ratio is eb. The reported standard error is the one development MCS knowledge lead to the adoption
associated with the hazard ratio. of MCS; but there is no reason to believe that one

19
 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347 341

event is more likely to happen earlier than any other focus ( 1.454). Learning by doing companies per-
event. An alternative explanation is that 69 compa- form better than contract and chaos. Need to focus
nies do not give enough statistical power to unveil companies perform worse than informal manage-
these differences.29 Learning by doing is not signifi- ment firms (which perform better than contract,
cantly associated with faster adoption of the project need to focus, and chaos) and do not perform signif-
milestones’ MCS; this is consistent with the routine icantly better than any of the other categories.
having to be enacted several times before it is coded. The association between performance and rea-
The analysis in Panel A of Table 4 was repeated sons-for-adoption suggests various discussion
for the next two most widely used product develop- points. First, except for ‘‘background” and ‘‘learn-
ment systems – ‘‘reports comparing actual progress ing,” the other reasons-for-adoption have a negative
to plan” and ‘‘Budget for development projects.” coefficient in some cases at a significant level. This
The same inferences from Panel A of Table 4 result suggests that informal management is not
regarding the significance of individual reasons- necessarily bad and may outperform formalized sys-
for-adoption apply for these two systems. tems. The question that emerges is whether the
under-performance of contracting, need to focus,
and chaos indicates that these systems are worse
Results for product development performance and
than an informal management approach; or rather
reasons for MCS adoption
(a more likely explanation), that the under-perfor-
mance reflects the limitations of the informal
A pivotal measure of product development per-
approach that led to the adoption of these MCS
formance is on-time development – see Table 2
(and that the company is still in the transition per-
(Panel B) where time-based measures are mentioned
iod to learn how to use them). Alternatively, these
by 62 out of 69 companies. We examine whether
same product development processes without these
companies adopting MCS because of manager
systems would perform even worse with an informal
background to facilitate product development man-
approach. Our data do not allow separating these
agement perform better (in terms of on-time devel-
different explanations but disentangling which one
opment) than firms adopting because of external
better explains the phenomenon is an important line
reasons-for-adoption, reactive to chaos, and infor-
for future research. A second conclusion suggests
mal. Manager background is the single most cited
that MCS adopted because of the background of a
reason for adopting a product development MCS.
manager are associated with better performance
The dependent variable takes a value of 0 for
than adopting because of chaos, contract, or need
companies where product development projects
to focus. This result speaks to the importance of
are late (43 observations) and 1 if they are on-time
timely adoption and by a manager with the knowl-
or early (21 observations). We lost five observations
edge to deploy the system.
for companies that chose not to disclose on-time
When external reasons drive the adoption of sys-
product development performance.30 The depen-
tems (more so when the reason is an external con-
dent variables are the same as in Table 4, Panel A.
tract), MCS appear to fail to bring the required
Table 4, Panel B reports the results. Manager
structure to improve on-time performance. Finally,
background driven companies perform significantly
when systems are adopted as a reaction to crisis
better as regards to on-time development perfor-
and problems, performance is worse than other cat-
mance than companies adopting as reaction to
egories. This result suggests that the company may
chaos or need to focus. The coefficient for manager
design systems without accessing the appropriate
background is 0.767, which is larger than the nega-
knowledge. An important caveat is that we only
tive coefficients for chaos ( 2.730) and need to
look at one performance dimension: on-time sche-
29
The statistical difference between the background coefficient
dule attainment.
and the learning coefficient was only significant at 10% one-tailed.
30
We choose a 1/0 variable because companies have very Discussion
different types of projects running at the same time – large
projects to develop a new platform, medium projects to develop
Discussion of the main findings
particular functions, and shorter projects to adapt the product to
a particular customer. During the interview managers described
significant variation across projects and often felt unsure about The events that trigger MCS adoption have
what an appropriate ordinal number would be. received limited attention in the existing literature.

20
342 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347

Using a database that includes over 200 interviews adoption. We find support for this event being
with managers from 69 companies, we identify six associated with product development MCS adop-
different reasons-for-adoption. Two reasons arise tion in 20% (14 companies) of our sample of 69
from external factors – legitimize and contract. companies.
Financial accounting research has long stressed We find much support in our research for the
the role that external regulatory bodies such as roles current theory and literature identify for
the IASB or the FASB play in financial reporting MCS. Appendix A provides illustrative quotes from
method choice. Our analysis of product develop- our interviews for seven MCS roles. In some cases,
ment systems highlights the role of two external there is a one-to-one correspondence between the
reasons/events in MCS choice. The increasing role reason for adoption and a MCS role – see especially
of multi-company agreements (such as joint-ven- the legitimize and contract external reasons for MCS
tures and large-company reimbursement of early- adoption. In most cases, a given reason (such as
stage company research) is likely to see the impor- manager background or chaos) may be associated
tance of external factors in MCS adoption with the adoption of MCS performing different
increase.31 Four of the six factors we identify as roles in individual companies.
prompting MCS adoption are classified as ‘‘inter-
nal.” Two factors are predominantly of a ‘‘proac- Limitations and extensions of the research
tive” kind – manager background and need to
focus. One aspect of the manager background rea- An important caveat in interpreting our results
son is the import-in notion. Key individuals (such is causality. The interview protocol was designed
as the CEO or a board member) may perceive a to have managers describe why systems were
major capability gap in their company and then adopted. Thus, the analysis implies causality but
hire a manager who brings in (imports-in) a new only in the sense that managers perceived such a
MCS that is part of building up the internal capa- causality to exist. We cannot preclude that the
bility. This reason is more likely with early-stage perceived cause of adoption and the adoption
companies than established companies as the for- itself is due to an omitted correlated variable that
mer rarely have the full set of functional capabili- the manager and the coding of the interviews
ties from day zero. For example, companies with a failed to unveil. Another important issue in inter-
long time gap between startup date and first reve- preting the results is that the categories that we
nues (such as biotechnology companies) may delay identify as reasons-for-adoption of MCS are based
for considerable time the adoption of a customer on the analysis of 207 interviews. Because the lit-
relationship management system (CRM). The hir- erature provides limited guidance as to why
ing of a senior marketing/sales executive after sev- MCS emerge, the categories identified emerge
eral years of R&D/product development may be a from the analysis of data. We cannot rule out
prompt to that new executive implementing a the possibility that other reasons-for-adoption
CRM system in the early-stage company. Two of may also exist.
the four internal reasons/events for MCS adoption The sample is dominated by companies that
we describe are predominantly of a ‘‘reactive” nat- received venture capital and are concentrated in
ure kind – learning and chaos. Simons (1995) ear- two main industries: information technology and
lier identified the chaos event as a trigger for MCS biotechnology. These characteristics make the find-
ings most relevant to companies with similar profiles.
31
A fundamental premise of some observers on accounting is Our research design used three contemporaneous
that, apart from the influence of GAAP-based rules for interviews per company as the main source of data.
inventory valuation, management accounting choices are inter- We did not conduct longitudinal observation of
nally determined (in contrast to financial accounting where the
influence of external parties such as the IASB and FASB is
these companies (although our interviews probed
sizable). Indeed, some observers use phrases such as ‘‘internal emergence over time in much detail). An extension
accounting”/‘‘external accounting” as substitutable with ‘‘man- of the research would be to conduct interviews at
agement accounting”/ ‘‘financial accounting.” Our findings on different points in company history to further probe
the role of the two external reasons for management accounting the MCS emergence question. Such a study would
system adoption (what we call legitimizing the company and
contracting with external parties) highlights that the labeling of
enable us better understand the use of MCS within
management accounting as internal accounting is both simplistic the web of social interactions both within the com-
and wrong. pany and with outside parties.

21
 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347 343

We treat the adoption of MCS as an event itself. Acknowledgements

Prior literature (Davila, 2005; Sandino, 2007) have
used a similar research design. Simplifying the This paper was funded by the Center for Entre-
adoption to an event facilitates the analysis of preneurial Studies (CES), Graduate School of
the results, but leaves open the rich issue of what Business. We greatly appreciate the support of
drives (and when) future refinement of the adopted CES co-directors Professors Irv Grousbeck,
MCS. While there is some evidence on how this Charles Holloway, and Garth Saloner and CES
growth takes place (Cardinal et al., 2004; Moores program director Linda Wells. We also thank the
& Yuen, 2001), the phenomenon is still little under- comments by participants at the University of
stood. A consistent comment from managers we Maastricht and two anonymous reviewers. We
interviewed was the increase in sophistication of are also very grateful for the support of the people
their MCS over time. that assisted us on this project – Jakub Wilsz, Jen-
This study has focused on product development nifer van Steele, Aimee Noelle Swanson, Christo-
but it is silent on other processes and functions such pher Armstrong, Jan Chong, Merle Ederhof, and
as marketing, sales, human resources, or finance. Ravi Sarin.
Exploring these different settings would extend the
empirical basis to formulate a theory of emergence
and growth of MCS. Our research design did not Appendix A
collect data on the specific design of the MCS
adopted. Having this information may shed some A.1. Theoretical roles of MCS in innovative settings:
light on the association between reasons-for-adop- illustrations from early-stage entrepreneurial
tion and performance. For instance, a poor MCS company sample
design may explain why adoption as a response to
chaos is associated with worse on-time development MCS roles in Quotes from interviews that
performance. innovative illustrate the roles
The study can be extended to examine whether settings
reasons similar to the ones identified for early-stage
1. Make goals ‘‘Myself and a couple of the
firms are also relevant for explaining the evolution
explicit, stable other folks here thought we’d
of these systems in established companies. Adoption
and visible to better put a structure in place
is just the first step in the lifecycle of MCS; a com-
provide where we write a specification
parative study over this lifecycle to better under-
convergence in (. . .) and at some point we
stand why systems are redesigned would extend
meaning across freeze it and then we develop
and enrich the findings reported in this study.
actors against it. (. . .) And it’s a
A challenging aspect of research on early-stage
challenge because you have a
companies is their higher mortality. The sample of
moving target most
companies we examined had survived their startup
of the time, but the reality is
phase and developed a more established and possi-
if that specification changes
bly broader set of functional capabilities. Those
on too dynamic a basis,
companies that do not survive represent a fascinat-
there’s no way really a
ing but difficult to research sample. Anecdotal evi-
technology team to execute
dence from early-stage company failures cite
against it.”
factors that potentially have strong ‘‘absence” of
2. Code learning ‘‘It is very important that the
MCS aspect, e.g. failure to anticipate cash shortages
from past process is followed because we
and failure to meet promised product or customer
experience to have done terrible when we
deadlines. Detailed case studies rather than larger
avoid past don’t follow the process, so we
sample research designs of the kind adopted in this
mistakes and are now very meticulous about
study is likely a more fruitful avenue to examine this
increase following it.”
set of failed companies. Further research on the
efficiency
‘‘consequences” of not adopting MCS is a useful
(continued on next page)
complement to the research we report on surviving
early-stage entrepreneurial companies.

22
344 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347

Appendix A (continued) Appendix B

MCS roles in Quotes from interviews that
B.1. Early-stage entrepreneurial company sample:
innovative illustrate the roles
questionnaire items
settings
3. Help in ‘‘The company was somewhat 1. When was the company legally established?
coordinating more contentious and more (month/year) _________
innovation fractured, and (putting in 2. We are interested in mapping the evolution of
efforts across place product development your company in terms of number of equivalent
different parts processes) was the only way to full time employees and number of R&D employ-
of the get everyone unified on a ees. What were these measures for each of the
organization common plan of attack.” dates specified below?
4. Plan the ‘‘The focus is on understanding
sequence of reporting progress towards
Date Total full Total full time
steps that lead achieving an important
time R&D
to stated goals,milestone; what the underlying
employees employees
help causes are for what we’re seeing
coordination and what the plans are, next steps Dec 2003
over time that are being taken. (. . .) (...)
Schedule is important, but it is Dec. 1993
only the start of the conversation.”
5. Promote ‘‘The benefits of (product
accountability development processes) 3. How long is a typical product development pro-
and facilitate formalization is that you get ject, from initial concept development to product
control by more accountability. I think it launch? ______________
4. Compared to the original target launch date for a
exception is very easy from an
typical product, is the actual date the product is
when engineering perspective to say I
launched generally (circle one) . . . early/late by
innovation can do that. (. . .) What you get
an average of how much time __________
results deviate is (a) you can hit that 5. Please indicate when your company formalized
from expectation in a timely manner, each system below. ‘‘Formalized” is defined
expectations and (b) develop features that as having documented a process and / or
meet the market on time. (. . .) periodically and purposefully executing the
As the company matures, process.
people get more accountable”
6. Contract with ‘‘The first product road map
external was one specifically that we Year
partners developed for sharing with formalized
Microsoft” Project milestones
7. Symbols to ‘‘What (our customer) said was Budget for development projects
legitimize the ‘‘I want to see what processes Reports comparing actual progress
organization in have you instituted in your to plan
its system? We are going to buy Project selection process
environment version 5.0 of your product, I Product portfolio roadmap
want to see what you did from Product concept testing process
version 3.0, 4.0, 5.0. What were Project team composition
the milestones? What bugs did guidelines
you fix? What was testing that
went through?” We had done
all those things but we didn’t 6. What are the three most important measures that
have the documentation. So we top management uses to evaluate the progress of
had to show them and to be a development effort (for example, schedule
candid (recreate it).” attainment or on-budget)?

23
 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347 345

C.1.4. Strategy
Measure How often does top management 1. What does the competitive landscape look like?
check it (weekly, monthly, . . .)? In terms of technology? Market participants?
1 Any significant changes? If so, why?
2 2. What do your target customers value about your
3 company compared to your competitors?

C.1.5. Product development

Appendix C 1. How do new product development projects get
selected? How did this selection process evolve?
C.1. Early-stage entrepreneurial company sample: Why?
protocol questions for interviews
C.1.6. Systems
In this interview we are interested as much in cur- 1. Why did your company start to use the systems
rent practices as in changes in these practices over identified in the questionnaire?
time and the reasons for these changes. We want to 2. What factors drove the need to adopt the
understand the company’s history, its chronology, systems?
and the evolution of its management practices. During 3. Who designed the systems?
the interview emphasize the idea of evolution and iden- 4. Has any been modified? Why?
tify reasons why each of the variables in the research 5. What is the biggest challenge in using these
changed. (Italics are instructions for interviewers.) systems?
6. Has any been dropped? Why?
In this interview we want to better understand
the challenges that your company faced (and is still
How are the key performance dimensions mea-
facing) in moving from the startup phase to the
sured? Why are they key?
phase where professional management systems are
required to manage the company. We are interested
References
in your company’s history and chronology as well.
We already received your answers to the question- Abernethy, M. A., & Brownell, P. (1997). Management control
naire, which were very useful to focus the questions systems in research and development organizations: The role
in this interview. of accounting, behavior and personnel controls. Accounting,
Organizations and Society, 22, 233–249.
Abernathy, M., & Brownell, P. (1999). The role of budgets in
C.1.1. Products
organizations facing strategic change: An exploratory study.
1. Can you provide a brief description of your com- Accounting, Organizations and Society, 24(3), 189–204.
pany’s products? We are interested in the assess- Abernethy, M. A., & Lillis, A. M. (1995). The impact of
ment of the technology behind the products and manufacturing flexibility on management control system
their manufacturing complexity. design. Accounting, Organizations and Society, 20, 241–258.
Adler, P. S., & Borys, B. (1996). Two types of bureaucracy:
Enabling and coercive. Administrative Science Quarterly,
C.1.2. History 41(1), 61–89.
1. How has the business model of the company Ahrens, T., & Chapman, C. S. (2004). Accounting for flexibility
changed over time? and efficiency: A field study of management control systems in
2. What were the main turning points? a restaurant chain. Contemporary Accounting Research, 21,
3. What were the main phases in the history of the 271–301.
company? Amabile, T. M. (1998). How to kill creativity. Harvard Business
Review, 76, 76–81.
Amat, J., Carmona, S., & Roberts, H. (1994). Context and
C.1.3. Organization
change in management accounting systems: A Spanish case
1. How is the company organized? (functional, busi-
study. Management Accounting Research, 5(2), 107–122.
ness unit, matrix) Baron, J. N., Burton, M. D., & Hannan, M. T. (1999).
2. Was it always this way? (If it changed) When/why Engineering bureaucracy: The genesis of formal policies,
did it change? Who proposed or structured the positions, and structures in high-technology firms. The
change? Journal of Law, Economics, and Organization, 15, 1–41.

24
346 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347

Bisbe, J., & Otley, D. (2004). The effects of the interactive use of knowledge integration mechanisms. Accounting, Organiza-
management control systems on product innovation. tions and Society, 29, 401–422.
Accounting, Organizations and Society, 29, 709–737. Feldman, M. S., & Rafaeli, A. (2002). Organizational routines as
Brown, S. L., & Eisenhardt, K. M. (1995). Product development: sources of connections and understandings. Journal of Man-
Past research, present findings, and future directions. Acad- agement Studies, 39, 309–332.
emy of Management Review, 20, 343–378. Flamholtz, E. (2000). Growing pains: Transitioning from an
Brown, S. L., & Eisenhardt, K. M. (1997). The art of continuous entrepreneurship to a professionally managed firm San Fran-
change: Linking complexity theory and time-paced evolution cisco (2nd ed., 1st. edition, 1990). San Francisco CA: Jossey-
in relentlessly shifting organizations. Administrative Science Bass.
Quarterly, 42, 1–34. Gavetti, G., & Levinthal, D. (2000). Looking forward and
Burns, T., & Stalker, G. M. (1961). The management of looking backward: Cognitive and experiential search. Admin-
innovation. London: Tavistock. istrative Science Quarterly, 45, 113–137.
Cardinal, L. (2001). Technological innovation in the pharma- Granlund, M., & Taipaleenmaki, J. (2005). Management control
ceutical industry: The use of organizational control in and controllership in new economy firms – A life cycle
managing research and development. Organization Science, perspective. Management Accounting Research, 16, 21–57.
12(1), 19–36. Greiner, L. (1972). Evolution and revolution as organizations
Cardinal, L. B., Sitkin, S. B., & Long, C. P. (2004). Balancing and grow. Harvard Business Review, 50(4), 37–46.
rebalancing in the creation and evolution of organizational Hellmann, T., & Puri, M. (2002). Venture capital and the
control. Organization Science, 15(4), 411–431. professionalization of start-up firms: Empirical evidence.
Carruthers, B. G. (1995). Accounting, ambiguity, and the new Journal of Finance, 57(1), 169–198.
institutionalism. Accounting Organizations and Society, 20(4), Henri, J. F. (2006). Management control systems and strategy: A
313–328. resource-based perspective. Accounting, Organizations and
Chapman, C. S. (1998). Accountants in organisational networks. Society, 31, 529–558.
Accounting Organizations and Society, 23(8), 737–766. Howard-Grenville, J. A. (2003). Making it work: The resilience of
Chenhall, R. H., & Morris, D. (1986). The impact of structure, organizational routines. Boston University: Working paper.
environment, and interdependence on the perceived useful- Huber, G. P. (1991). Organizational learning: The contributing
ness of management accounting systems. Accounting Review, processes and the literatures. Organization Science, 2(1),
61(1), 16–35. 88–115.
Chenhall, R. H., & Morris, D. (1995). Organic decision and Innes, J., Mitchell, F., & Sinclair, D. (2000). Activity-based
communication processes and management accounting sys- costing in the UK’s largest companies: A comparison of 1994
tems in entrepreneurial and conservative business organiza- and 1999 survey results. Management Accounting Research,
tions. Omega, 23(5), 485–497. 11, 349–362.
Cohen, M. D., Burkhart, R., Dosi, G., Egidi, M., Marengo, L., Kalagnanam, S. S., & Lindsay, R. M. (1999). The use of organic
Warglien, M., et al. (1996). Routines and other recurring models of control in JIT firms: Generalising Woodward’s
action patterns of organizations: Contemporary research findings to modern manufacturing practices. Accounting,
issues. Industrial and Corporate Change, 5, 653–698. Organizations and Society, 1–30.
Cooper, R. G. (1995). Developing new products on time, in time. Khandwalla, P. N. (1972). The effect of different types of
Research Technology Management, 49–57. competition on the use of management controls. Journal of
Damanpour, F. (1991). Organizational innovation: A meta- Accounting Research, 46, 275–285.
analysis of effects of determinants and moderators. Academy King, G., Keohane, R. O., & Verba, S. (1994). Designing social
of Management Journal, 34(3), 555–590. inquiry: Scientific inference in qualitative research. Princeton
Davila, A. (2000). An empirical study on the drivers of NJ: Princeton University Press.
management control systems’ design in new product devel- Kvale, S. (1996). Interviews: An introduction to qualitative
opment. Accounting, Organizations and Society, 25(4, 5), research interviewing. Thousand Oaks, CA: Sage Publications.
383–409. Levitt, B., & March, J. G. (1988). Organizational learning.
Davila, A. (2005). An exploratory study on the emergence of Annual Review of Sociology, 14, 319–340.
management control systems: Formalizing human resources Lundberg, C. C. (1995). Learning in and by organizations: Three
in small growing firms. Accounting, Organizations and Society, conceptual issues. International Journal of Organizational
30(3), 223–248. Analysis, 3, 10–24.
Davila, A., & Foster, G. (2005). Management accounting Macintosh, N. B. (1994). Management accounting and control
systems’ adoption decisions: Evidence and performance systems. Toronto: John Wiley & Sons.
implications from startup companies. The Accounting Review, Marshall, C., & Rossman, G. B. (1995). Designing qualitative
80(4), 1039–1068. research. Thousand Oaks, CA: Sage Publications Inc.
Davila, A., & Foster, G. (2007). Management control systems in McGrath, M. D. (1995). Product strategy for high-technology
early-stage startup companies. The Accounting Review, 82(4), companies. New York, NY: Richard Irwin, Inc.
907–937. Miner, A. S., Bassoff, P., & Moorman, C. (2001). Organizational
Dekker, H. C. (2004). Control of inter-organizational relation- improvisation and learning: A field study. Administrative
ships: Evidence on appropriation concerns and coordination Science Quarterly, 46, 304–337.
requirements. Accounting, Organizations and Society, 29, Mintzberg, H. (1976). Planning on the left side and managing on
27–50. the right. Harvard Business Review, 54(4), 49–58.
Ditillo, A. (2004). Dealing with uncertainty in knowledge- Mintzberg, H. (1979). The structuring of organizations. Upper
intensive firms: The role of management control systems as Saddle River NJ: Prentice Hall.

25
 A. Davila et al. / Accounting, Organizations and Society 34 (2009) 322–347 347

Mintzberg, H. (1994). The rise and fall of strategic planning. New Rockness, H. O., & Shields, M. D. (1988). Organizational control
York, NY: Free Press. systems in research and development. Accounting, Organiza-
Moores, K., & Yuen, S. (2001). Management accounting systems tions and Society, 9, 165–177.
and organizational configuration: A life-cycle perspective. Romano, C., & Ratnatunga, J. (1994). Growth stages of small
Accounting, Organizations and Society, 26, 351–389. manufacturing firms: The relationship with planning and
Nelson, R. R., & Winter, S. G. (1982). An evolutionary theory of control. British Accounting Review, 26(2), 173–195.
economic change. Cambridge, MA: Harvard University Press. Sandino, T. (2007). Introducing the first management control
Nonaka, I. (1990). Redundant, overlapping organization: A systems: Evidence from the retail sector. Accounting Review,
Japanese approach to managing the innovation process. 82(1), 265–293.
California Management Review, 32, 27–38. Seidman, I. (1998). Interviewing as qualitative research: A guide
Oakes, L. S., Townley, B., & Cooper, D. J. (1998). Business for researchers in education and the social sciences. New York:
planning as pedagogy: Language and control in a changing Teachers College Press.
institutional field. Administrative Science Quarterly, 43(2), Simons, R. (1987). Accounting control systems and business
257–293. strategy: An empirical analysis. Accounting, Organizations and
Otley, D. T., & Berry, A. J. (1994). Case study research in Society, 12(4), 357–375.
management accounting and control. Management Account- Simons, R. (1995). Levers of control: how managers use innovative
ing Research, 5(1), 45–65. control systems to drive strategic renewal. Boston, MA:
Ouchi, W. (1979). A conceptual framework for the design of Harvard Business School Press.
organizational control mechanisms. Management Science, 25, Strauss, A., & Corbin, J. (1990). Basics of qualitative research.
833–848. Thousand Oaks, CA: Sage Publications.
Perrow, C. (1970). Organizational analysis: A sociological view. Ulrich, K. T., & Eppinger, S. D. (1995). Product design and
New York, NY: Tavistock Publications. development. New York, NY: McGraw Hill.
Pfeffer, J., & Salancik, G. R. (1978). The external control of Weick, K. E. (1995). Sensemaking in organizations. Thousand
organizations: A resource dependence perspective. New York, Oaks, CA: Sage.
NY: Harper & Row. Weick, K. E., Sutcliffe, K. M., & Obstfeld, D. (1999). Organizing
Powell, W. W. (1998). Learning from collaboration: Knowl- for high reliability: Processes of collective mindfulness. In R.
edge and networks in the biotechnology and pharmaceu- I. Sutton & B. M. Staw (Eds.), Research in organizational
tical industries. California Management Review, 40, behavior. Stamford: JAI Press.
228–240. Zirger, B. J., & Maidique, M. A. (1990). A model of new product
Powell, W. W., & DiMaggio, P. J. (1991). The new institutionalism development: An empirical test. Management Science, 36(7),
in organizational analysis. Chicago: University of Chicago 867–884.
Press. Zollo, M., & Winter, S. G. (2002). Deliberate learning and the
Quinn, J. B. (1978). Strategic change: ‘‘Logical incrementalism”. evolution of dynamic capabilities. Organization Science, 13,
Sloan Management Review, 20(1), 7–21. 339–352.

26