Sesi 03 - 1 - Article ERP and Its Impact

Gattiker & Goodhue/After ERP Implementation
RESEARCH ARTICLE
WHAT HAPPENS AFTER ERP IMPLEMENTATION:

UNDERSTANDING THE IMPACT OF INTERDEPENDENCE AND DIFFERENTIATION ON
PLANT-LEVEL OUTCOMES1
By: Thomas F. Gattiker
Department of Networking, Operations,
and Information Systems
College of Business
Boise State University
1910 University Drive
Boise, ID 83725
U.S.A.
TomGattiker@boisestate.edu
Dale L. Goodhue
Department of Management Information
Systems
Terry College of Business
University of Georgia
Athens, GA 30602
U.S.A.
dgoodhue@terry.uga.edu
Abstract
phase has occurred. Organizational information

processing theory states that performance is
influenced by the level of fit between information
processing mechanisms and organizational context. Two important elements of this context are
interdependence and differentiation among subunits of the organization. Because ERP systems
include data and process integration, the theory
suggests that ERP will be a relatively better fit
when interdependence is high and differentiation
is low. Our model focuses at the subunit level of
the organization (business function or location,
such as a manufacturing plant) and includes
intermediate benefits through which ERPs overall
subunit impact occurs (in our case at the plant
level). ERP customization and the amount of time
since ERP implementation are also included in the
model. The resulting causal model is tested using
a questionnaire survey of 111 manufacturing
plants. The data support the key assertions in the
model.
We present a model of the organizational impacts

of enterprise resource planning (ERP) systems
once the system has gone live and the shake-out
1
Peter Todd was the accepting senior editor for this

paper. Carol Brown was the associate editor. Judy
Scott served as a reviewer. The second reviewer chose
to remain anonymous.
Keywords: Organizational information processing

theory, enterprise systems, ERP, data integration,
interdependence, differentiation, manufacturing
planning and control
MIS Quarterly Vol. 29 No. 3, pp. 559-585/September 2005
559
Introduction
Enterprise resource planning (ERP) systems are
commercial software systems that automate and
integrate many or most of a firms business
processes. Sometimes called enterprise systems,
ERP systems promise integration of business
processes and access to integrated data across
the entire enterprise (Davenport 1998). Furthermore, companies that implement the systems have
the opportunity to redesign their business practices
using templates imbedded in the software (often
called best practices) (Scheer and Habermann
2000). Among medium and large companies, ERP
adoption is approximately 75 percent for manufacturing and 60 percent in services (Scott and
Shepherd 2002) and is 80 percent among Fortune
500 firms (META Group 2004).
However, while some firms have achieved impressive benefits from their ERP systems, others have
experienced difficulty in gaining the benefits they
expected. Thus, as suggested by Markus and
Tanis (1999), for both researchers and executives,
one of the key questions is will investments in ERP
pay off? Several firm-level econometric studies
(Anderson et al. 2003; Hitt et al. 2002) suggest
that the most likely answer is yeson average.
However, the impacts vary from firm to firm.
Therefore, the executives next questions are likely
to be, How can I maximize the positive impact?
How do I avoid the problems? Is there some way
to predict what the ultimate impact will be for my
firm?
These questions are worth addressing. As an
attempt to do so, we draw on organizational information processing theory (OIPT) (Daft and Lengel
1986; Galbraith 1973; Goodhue et al. 1992; Miller
1992; Thompson 1967; Tushman and Nadler
1978). One of the defining characteristics of ERP
is the extensive integration it provides among the
subunits (such as manufacturing plants or different
functional departments) of a business. OIPT
suggests highly integrated systems will fit some
organizational subunits better than othersand
that interdependence and differentiation are two
characteristics that might influence the level of fit.
Specifically, when ERP is implemented, subunits
560
MIS Quarterly Vol. 29 No. 3/September 2005
that are highly interdependentthat is, very

dependent on other subunitsmay benefit
substantially. However, subunits that are very
different from the other subunits in the ERP implementation may incur costs (such as suboptimal
business processes or dependence on employee
work-arounds). In brief, our supposition is that
since ERP systems provide integration and standardization, their impact will be influenced by the
interdependence and differentiation between subunits of the organizationas depicted in Figure 1.
Before introducing a model designed to test these
ideas, we discuss two principles on which our work
on this topic is based. To effectively study this
issue, it is appropriate to (1) focus on intermediate
benefits at a local level, as opposed to the level of
the organization as a whole, and (2) take a postimplementation focus.
Intermediate Benefits at Individual

Manufacturing Plants
Barua et al. (1995) argue that truly understanding
how IT investments create value for the organization requires (1) a fine-grained unit of analysis:
the business function, such as operations or
marketing, rather than the firm as a whole, and
(2) a research model that includes the intermediate
benefits or intermediate variables (p. 8) through
which the functional impacts occur.
Understanding the intermediate benefits (the
second recommendation above) helps us explain
why certain overall impacts do or do not occur. At
the aggregate level, an ERP might help a firm
survive because it leads to higher profits. However, this result does not happen automatically
there must be pathways through which ERP
causes higher profits (and causes some firms to
achieve them to a greater degree than their peers).
Conceptually, since ERP systems provide integrated data and (arguably) so-called best practice
business processes, key intermediate benefits for
ERP might include higher quality data for decision
making, efficiency gains in business processes,
and better coordination among different units of
Interdependence
+
Intermediate ERP
Benefits
ERP
Local level
overall ERP
benefits
Differentiation
Figure 1. Conceptual Model of ERP Impact at the Local Level
the firm. By studying the factors leading to these

intermediate benefits, and the extent to which each
intermediate benefit contributes to overall impact,
we can better understand the pathways through
which ERP does (and does not) help organizations.
Following the first recommendation of Barua et al.,
we use individual manufacturing plants as our unit
of analysis. If an information system is to have a
long-lasting transformative effect on organizations,
then a substantial portion of that effect probably
will come from the technologys impact on the core
value-adding activities of the company, such as
operations, rather than on administrative functions
(Barua et al. 1995). Within manufacturing, we
focus on the manufacturing planning and control
area (as opposed to quality, engineering, and so
on), which includes developing a general production plan, detailed material and capacity planning
(e.g., MRP, CRP), and execution (i.e., purchasing
and shop floor) (Vollmann et al. 1992).
A local (plant) level of analysis is consistent with
our theory base. One can approach interdependence from the firm levelas a property of organizations (e.g., as the overall or average level of
interdependence among all plants of an organization). Or one can approach the construct as a
property of subunitsby considering the level of
interdependence between a particular plant and

the other parts of the organization (e.g., other
plants) (Tushman and Nadler 1978, p. 616). In
keeping with our goal of understanding intermediate variables at the functional level, we adopt
the latter conceptualization. Similarly we conceptualize differentiation as a property of an individual
plantas the level of difference between that plant
and the organizational norm on a number of key
characteristics, such as the volume and variety of
products produced (as we describe in more depth
when we introduce our differentiation hypothesis).
Similarly, it is appropriate to approach the outcome
variables from a local level of analysis. For most
information systems, some positive impacts will
accrue globally and some locally. For example, a
global ERP benefit could be the ability of corporate
accounting to close the books more quickly
because all inventory and order status information
is on one database, rather than on numerous sitespecific legacy systems. By contrast, a local
benefit would occur if ERP provides plant personnel with better visibility of orders from other
plantsperhaps with decreased expediting and
overtime as a result. These local impacts add up
to form a substantial portion of the total global
effects. Understanding local effects is necessary
to understand how the overall benefits occur.
561
Post-implementation Focus
Recent literature reviews (Dong et al. 2002;
Esteves 2001; Jacobs and Bendoly 2003), suggest
that most existing ERP research focuses on
selection and implementation, not on ERPs postimplementation impacts. Selection and implementation are critical areas, and numerous valuable
insights have emerged, including the importance of
relationships with consultants, core team characteristics, business process reengineering approach, change management approach, user
training approach, top management support, project champion, user involvement, package choice,
module choice, package customization, project
management, implementation approach, and a
clearly stated business case (Akkermans and van
Helden 2002; Brown and Vessey 1999; Gefen
2002; Hirt and Swanson 1999; Holland and Light
1999; Hong and Kim 2002; Ng et al. 1999; Robey
et al. 2002; Scheer and Habermann 2000; Sumner
1999). In addition to improving implementation
success itself, these factors influence later results.
A number of articles suggest also that ERP
systems go through a post-implementation breakin phase, in which performance may not be typical
of the longer-term effects that the organization
might experience (Cosgrove Ware 2003; Markus
and Tanis 1999; Ross and Vitale 2000). According
to Staehr et al. (2002), however, the ultimate
impacts of ERP on the organizationonce the
system has been implemented and has been
shaken downare not as thoroughly researched.
Therefore, in this paper, we focus on explaining
differences in impacts among plants that have
been running ERP for long enough to be through
the implementation and break-in phase.
Research Model
Dependent Variables
Figure 2 shows our general model of the local
impact of interdependence and differentiation. In
addition, it shows several other control variables,
which are suggested by the existing ERP literature. As we suggested in the introduction to this
562
paper, we believe there are three important

intermediate benefits through which ERP could
deliver overall plant level benefits to firms: better
information (data quality), more efficient internal
business processes (task efficiency), and better
coordination between different units of the firm
(coordination improvements). Thus we show
arrows from each of them to local level overall
benefits. Below we explain our rationale for the
key links in the model.
Interdependence Influences
ERP Benefits
OIPT states that, in order to prosper, organizations
must resolve uncertainty. The amount and types
of uncertainty vary from organization to organization and include the stability of the external
environment, the predictability of core processes,
how tasks are subdivided, and the level of interdependence among those subdivisions (Galbraith
1973; Thompson 1967; Tushman and Nadler
1978). Therefore an organization must select and
deploy the subset of information processing
mechanisms (hierarchies, different schemes of
departmentalization, lateral relations, computer
systems, etc.) that fits the particular uncertainties
that it faces (Daft and Lengel 1986). ERP systems can be viewed as a particular class of
information processing mechanism. Thus OIPT
suggests ERPs impact depends, at least in part,
on the amount and types of uncertainties at hand.
Early OIPT theorists (e.g., Galbraith 1973, 1977)
focused on uncertainty at the company level.
Tushman and Nadler (1978) moved the focus to
the subunit level. They suggest that the impact of
an integrative coordination mechanism on a subunit, such as a plant, may depend on the level of
interdependence between that plant and other
plants in the organization: The greater the interdependence that one subunit shares with another,
the greater the need for them to share information,
because changes in conditions in one subunit may
require some adjustment in the other. In such
cases, each subunits information processing
mechanisms must facilitate the exchange of infor-
In te rm e d ia te b e n e fits
H1
In te rd e p e n d e n c e
T a s k e ffic ie n c y
H2
D iffe re n tia tio n
H 6a
H3
C u s to m iz a tio n
T im e e la p s e d
s in c e
im p le m e n ta tio n
C o o rd in a tio n
im p ro v e m e n ts
H 6b
L o c a l (p la n t)
le v e l o v e ra ll
b e n e fits
H4
H 6c
H5
D a ta q u a lity
Figure 2. Research Model
mation across subunit boundaries within the

organization, and therefore highly integrated
mechanisms would be beneficial (Tushman and
Nadler 1978). For example, when two manufacturing plants within an organization serve a
common customer or when one plant provides
inputs to another plant, the level of interdependence between these two plants is high. Changes
in the master production schedule or inventories of
one plant may necessitate adjustments by the
other plant, and an information system that
integrates the data of both plants may thus
improve coordination.
Applying this reasoning to ERP, we would expect
ERP to improve coordination between one plant
and others in the business because ERP
increases information links among plants (and
among other subunits in the business). However,
the degree to which these stronger information
links produce a valuable benefit will vary among
organizations and even among plants within an
organization because interdependence varies
among them. For example, when plant A provides
plant B with materials (such as a stamping plant
supplying materials to an assembly plant), ERP
enables A to look at Bs projected material needs
(perhaps by accessing Bs material requirements
planning module), and ERP enables B to anticipate shipment quantities and late deliveries (perhaps by accessing As master production schedule
or capacity requirements planning system).

However, when interdependencies such as these
are not present or are less strong (i.e., when A and
B do not have common materials, customers, and
so on), opportunities for ERP-enabled coordination
improvements are not as prevalent.
This perspective is proposed by Goodhue et al.
(1992). Specifically, their first proposition states,
All other things being equal, as the interdependence between subunits increases,
the benefits of data integration will
increase and the amount of data integration in rational organizations should also
increase (p. 301).
Note that this is a two-part proposition. Wybo and
Goodhue (1995) investigated the second half of
the proposition (i.e., that greater interdependence
should lead to greater data integration) on the
assumption that the first half was correct.
However, they found no significant relationship.
This makes a compelling case for explicitly testing
the first half of Goodhue et al.s first proposition,
which we do here. In other words, is it really true
that the benefits of data integration increase with
interdependence?
To apply this issue to ERP, our first hypothesis
states that interdependence influences the degree
563
to which intermediate benefits from implementing

ERP are realized. Specifically,
H1a: In a plant within an ERP implementation, the greater the interdependence of one plant with other plants in
the organization, the greater the ERPrelated coordination improvements
accrued by that plant.
If we were comparing plants that had implemented
ERP with plants that had not, we would investigate
the presence of an interaction between interdependence and ERP. However, because we
intend to test this assertion only in plants that have
all already implemented an ERP (ERP implementation is held constant) what is conceptually a
moderator effect becomes a main effect.
In addition to improving coordination, ERP is more
likely to enhance task efficiency when interdependence is high. Without integrated systems,
interdependent subunits need to resort to relatively
time consuming methods of sharing information
with one another (fax, telephone). By contrast,
ERP can provide instant access to information,
making employees more efficient. The more interdependent plants are, the more ERP will improve
efficiency. Again, what is conceptually a moderating relationship becomes a main effect when
ERP implementation is held constant. Thus H1b
(which is parallel to H1a) states,
H1b: In a plant within an ERP implementation, the greater the interdependence of one plant with other plants in
the organization, the greater the ERPrelated task efficiency improvements
Differentiation Influences
ERP Benefits
Tushman and Nadler (1978) also stated that when
an individual subunits local task characteristics or
its local external environment differ from other
organizational subunits, then that subunit may well
require unique, nonstandard systems in order to
564
cope with its particular circumstances. By contrast, ERP systems tend to impose standard
processes and data on organizationsand on the
plants in those organizations (Davenport 1998).
Existing research has documented that the fit
between an ERPs standard processes and the
organizations business conditions is an important
issue (Somers and Nelson 2003). However, we
must also consider the possibility that there can be
a poor fit between an ERP and an individual
plants business conditions.
Once organizations have chosen a particular ERP
vendor and system, they must configure2 the
system by considering the overall corporate needs.
(We are assuming, for now, that the organization
avoids customization, bolt-ons, and the like.) In
other words, standard processes and data definitions are defined to meet the needs of the overall
company and its plantsa type of intra-company
consistency which many organizations consider
beneficial (Cooke and Peterson 1998; Kumar et al.
2002; Mabert et al. 2000). However, because all
subunits are subject to the same set of configuration decisions that are made at the organization
level, if one plant has very different business processes than the majority, that plant may experience problems because the ERP gives it little local
level flexibility (Gattiker and Goodhue 2002;
Jacobs and Bendoly 2003; Jacobs and Whybark
2000).
For example, a high-volume repetitive manufacturing plant was part of an ERP implementation
in a division that mostly consists of job shops. The
ERP system apparently was configured mainly to
fit the needs of the job shops. Using the system
created substantial operational problems for the
oddball plant (Gattiker 2002).
Another example comes from subunits that consume or ship discrete inventory items that need to
be denominated in continuous units (e.g., feet or
Configuration is the process of setting software switches

or modifying tables that define workflows and so on.
Configuration allows changes only within certain boundaries limited by what capabilities the vendor includes in
the software.
pounds on hand rather than just the number of

units on hand). A plant that produces engineered
beams and cuts them to unique lengths for each
customer order (i.e., not to stock) has this inventory problem (Gattiker and Goodhue 2004). In
advance of customer orders, the plant produces
blanks in a few lengths, such as 48 feet. When a
customer order arrives, beams are then cut to the
customer-specified length. They are cut from
either a new blank or from material left over from
when earlier orders were cut. Because they are
remainders, the lengths of leftovers are random,
and thus there are no part numbers for them.
Thus it is difficult for inventory control personnel to
track the lengths of individual leftovers in inventory. This is a problem because, for example, it is
not enough to know that there are 40 linear feet of
a particular type of beam in remainder inventory.
The plant must have a system that knows the
individual lengths of each remainder in stock,
because, for example, having two 20-foot-long
sections will not allow the plant to fill an order for a
customer who needs a 24-foot continuous length.
The plant has a relatively uncommon inventory
tracking problem. Therefore, if the plant is part of
an ERP implementation that includes mostly plants
that produce standard-length discrete products
(and thus do not have the same inventory
challenge), it may well experience difficulties.
When an ERP system is not a good fit for a plants
unique business processes, making do might
compromise performance (for example, the
engineered lumber producer might rely on time
consuming physical stock-checks before accepting
orders). Or plant personnel might revert to
informal, nonintegrated systems (e.g., spreadsheets, legacy systems) that meet local needs but
do not facilitate coordination beyond plant boundaries (Gattiker and Goodhue 2004; Soh et al.
2000). Either way there is a performance drop.
Such misalignments are a serious problem (Berry
and Hill 1992). Sia and Soh (2002) categorize
ERP misfits as surface (having to do with user
interface and the like) or deep structure (fundamental misfit between the model/package and
reality) and as pervasive (exogenous, stemming
from external sources) or non-pervasive (such as
different part numbers in different plants). Misfits

that are both deep-structure and pervasive are the
most problematic. Clearly many misfits between
an ERP configuration and a manufacturing facility
are deep structure misfits because they deal with
fundamental processes (i.e., physical transformation processes). Furthermore, many are pervasive
because many plant-level business processes are
tied to the products and markets for which the
plant is responsible and these are determined by
company-level strategy (Berry et al. 1991; Miller
1981). We do realize that not all misfits are deep
structure and pervasive. Certainly some intracompany process differences result merely from
managerial preferences or divergent evolutionary
drifts (for example, there are rarely fundamental
grounds for two plants to use inconsistent part
numbering systems). In order to make our focus
on deep structure and pervasive misfits clear, we
define differentiation as between-plant differences
that are related to products produced and markets
served.
In general, when differentiation (product and
market-related differences) between plants is
greater, it is less likely that a system that standardizes data and processes among plants will
meet all plants needs equally well. Thus, differentiation moderates the degree to which benefits
from implementing an ERP are realized. As with
H1, when ERP implementation is held constant,
this moderating effect translates into a main effect
of differentiation. Specifically,
H2a: In a plant within an ERP implementation, the greater the differentiation of a plant from the other plants
in an organization, the lower the ERPrelated coordination improvements
H2b: In a plant within an ERP implementation, the greater the differentiation of a plant from the other plants
in an organization, the lower the ERPrelated task efficiency improvements
565
Translating differentiation (which OIPT theorists

have conceptualized in very general terms) to the
manufacturing domain requires turning to the
product-process literature (Hayes and Wheelwright
1979; Hill 2000; Miller 1991; Safizadeh et al. 1996;
Van Dierdonck and Miller 1980), which states that
for good performance, a plants processes should
be aligned with the characteristics of the products
that the plant produces and of the markets that it
serves. Table A2 of Appendix A enumerates key
manufacturing strategy and process characteristics. These are fundamental dimensions along
which plants within an organization can and often
do differ from one another (Hayes and Wheelwright 1979; Hill 2000; Taylor 1980; Turner 1998).
Control Variable: ERP

Customization
Organizations have a variety of options when they
recognize that the process options within their
ERP system are not well aligned with the business
process that the organization desires. These are
detailed in Table 1.
Options 1 through 3 can be carried out at the
discretion of the plant, but may entail costs associated with making do with a suboptimal process
and/or sacrificing some of the potential functionality of the ERP (Soh et al. 2000). Option 4
typically requires global-level authorization and
resources (e.g., endorsement from the companys
IS management, programming expertise from the
companys IS department) because this approach
entails customization to the ERP itself.
Customization (option 4 in the table) may be a
response to a lack of fit between the organizations
business processes and those envisioned by the
ERP package designers. However, customization
could potentially also be used to bring the ERP
into line with the requirements of a nonstandard
plant. Customization may, therefore, be an effective strategy for dealing with the unique needs of
the extremely different plants discussed in the
previous section. Since we have a theoretical
interest in differentiation, our model needs to
include customization as well.
566
We conceptualize customization as changes to the

ERP system to meet the needs of an individual
plant. At this point we are considering ERP
customization in general terms rather than looking
at the various means by which these systems
could be customized (e.g., bolt-ons, code changes,
etc.). If this type of customization is indeed an
effective way to deal with local-level ERP-business
misalignment (perhaps due to high differentiation),
then we would expect to see a positive relationship
between customization and intermediate benefits
from ERP, especially when local differentiation is
high. Thus,
H3a (control hypothesis): In a plant
within an ERP implementation, customization moderates the negative
impact of an individual plants differentiation on coordination improvements from ERP accrued to that plant
H3b (control hypothesis): In a plant
within an ERP implementation, customization moderates the negative
impact of an individual plants differentiation on task efficiency improvements of ERP accrued to that plant.
Control Variable: Time Elapsed Since

ERP implementation
Several process models (Markus and Tanis 1999;
McAfee 2002; Ross and Vitale 2000) suggest that
ERP impacts on the organization may improve
with time. A survey by CIO Magazine (Cosgrove
Ware 2003) suggests that most companies do not
achieve the anticipated benefits after one year, but
the majority do reap them beginning in the second
year. In general, it appears that companies (and
the subunits that make up those companies) may
experience a performance dip initially after implementation (Ross and Vitale 2000). However, often
performance improves thereafter. We expect this
improvement to apply to both of our focal intermediate benefits: coordination improvements and
task efficiency. Although time elapsed does not
relate directly to our OIPT theoretical framework,
it is an important control variable. Therefore,
Table 1. ERP Misfit Resolution Strategies (adapted from Figure 1 in C. Soh, S. S.

Kien, and J. Tay-Yap, Cultural Fits and Misfits: Is ERP A Universal Solution,
Communications of the ACM (43:4), 2000, p. 50. Copyright 2000, ACM, Inc.
Reprinted by permission.)
1
Adapt to the new functionality in ERP (adopting the new operating processes embedded in
ERP)
Accept shortfall in ERP functionality (compromising on the requirements of the organization)
Workarounds to provide the needed functionality without touching the ERP scripts
Manual (by hand rather than using a computer system)
ERP alternative (finding an alternative way to perform the function with the package
Customization to achieve the required functionality
Non-core customization (interfacing with add-on module or through query/report writer

facility)
Core customization to amend the base code

within an ERP implementation, greater
time elapsed since ERP implementation is associated with greater coordination improvements of ERP accrued
to that plant.
time elapsed since ERP implementation is associated with greater task
efficiency improvements of ERP
accrued to that plant.
benefits. This is the basis for the following

hypotheses:
data quality is associated with greater
coordination improvements for the
plant.
task efficiency for the plant.
Control Variable: Data Quality
Overall Benefits at the Local Level
In attempting to discern the impacts of interdependence and differentiation on coordination and task
efficiency benefits, it is important to control for the
effects of data quality on those benefits. Without
data quality, many other benefits from an ERP will
likely not occur (Vosburg and Kumar 2001). More
specifically, without data quality (accurate and
relevant data), an organization is severely constrained in the coordination and task efficiency
benefits it can achieve from its ERP system. ERP
provides easy access to corporate-wide data, but
if that data is inaccurate or irrelevant to the
business process in the subunit, there will be few
Our ultimate dependent variable is local (plant)

level overall benefits, which we define as the
overall business impact of ERP on that plant. Our
model proposes that the overall benefits at the
plant level will come substantially through the
intermediate benefits of task efficiency, coordination improvements, and data quality. Thus we
present the following hypotheses:
H6a: In a plant within an ERP implementation, greater task efficiency is
associated with greater local (plant)
level overall benefits from ERP.
567
H6b: In a plant within an ERP implementation, greater improvements in

coordination with other subunits are
associated with greater local (plant)
level overall benefits from ERP.
H6c (control hypothesis): In a plant
local (plant) level overall benefits from
ERP.
Instrument Development
and Data Analysis
Instrument Development
A survey instrument was developed to measure
the constructs needed to test the above hypotheses. We used Bagozzis (1980) framework to
develop and validate the instrument. In order to
develop clear definitions of the constructs and their
interrelationships in a well-specified theoretical
context, we reviewed relevant bodies of literature,
and we conducted four studies of ERP systems
that were running in individual manufacturing
plants in four different companies (Gattiker 2002;
Gattiker and Goodhue 2000, 2004).
This
sharpened our understanding of how the relevant
constructs occurred in practice, allowing us to
define and operationalize them accurately.
We borrowed questions from existing scales where
possible. As an additional means of ensuring that
the questionnaire items corresponded to the
theoretical constructs as intended, we conducted
interviews with nine managers in local manufacturing facilities. (The positions of these individuals mirrored well the job titles of the respondents
that we later surveyed.) During the process, these
practitioners filled out a prototype questionnaire
and were asked to explain their interpretation of
the items. We also elicited informal verbal descriptions of interviewees business environments and
ERP systems, and we compared these to their
responses to the questionnaire items. The interview process resulted in refinements of many
questionnaire items. These final changes resulted
568
in 43 items used to measure the 8 constructs in

the model as shown in Appendix A.
Data Collection
Survey participation was solicited from two
different groups of APICS members and from user
associations of two of the major ERP packages.
APICS is the American Production and Inventory
Control Society, whose members are primarily
materials and scheduling personnel at plant and
central levels, as well as consultants and employees of IT vendors. The user groups mostly
consisted of IT staff but did include some operations people. Potential participants were either
sent or given a pencil and paper survey with a
traditional cover letter, or were given an e-mail
solicitation inviting them to visit a Web site with a
parallel version of the survey. IT consultants and
non-operations people were removed from the
pool of responses, as were individuals who
indicated that their plant had not implemented
ERP. This left 129 usable responses. (In six
cases, two surveys from the same plant were
averaged to make three responsesone per
plant.) Surveys from APICS mailing lists and
APICS list-serves accounted for 81 percent of the
usable responses.
Since e-mail solicitations were sent to list-serves
and the composition of the list serve subscribers
(practitioners, academics, consultants, manufacturing versus service, and so on) is unknown,
computing a response rate is problematic. Our
response rate on pencil and paper surveys sent to
mailing lists was approximately 9 percent.
However, even among the traditional mailing lists,
we could not cull out companies that had not
implemented ERP, and presumably these
individuals had little motivation to return the
survey. Since computing the overall response rate
(especially the response among companies with
ERP) is problematic, we need to consider the
possibility that we have a response bias. Indeed,
we may well have a bias. Even for individuals in
organizations that have implemented ERP, the
motivation to fill out the survey is probably higher
when the system is relatively successful. Thus we
might assume that our results are reflective of
organizations that have been relatively successful

with ERP. We follow up on this notion in the
discussion section.
Sample Characteristics
In the final sample, each of the following industries
was represented by at least 5 percent of respondents: automotive, chemicals, consumer, electronics, and other processing. We did not collect
data from more than one plant per company. The
average number of plants in the ERP implementations represented was 7.3. All of the companies
in this sample had implemented manufacturing
modules as part of their ERP system. Over 70
percent had the following applications: MRP/
scheduling, purchasing, shop floor control, and
accounting. Over 25 percent included engineering
applications and human resources in their systems. Tables 2 through 5 characterize responses
on other characteristics. Of the responses, 21
percent were for packages that do not have the
instant name recognition of SAP or JD Edwards
and dominant vendors. Although these packages
are consolidated as other in Table 4, respondents indicated their package name on the survey.
We looked these up on the Web site that
describes software offerings by vendor in order to
ensure that they were indeed ERP systems.
Measurement Validity
of the Constructs
Because of the sample size (n = 129), we opted to
use exploratory factor analysis and Cronbachs
alpha to establish discriminant validity and internal
consistency (reliability). Our model includes constructs at three levels as shown in Figure 1 above:
interdependence and customization for the first
level3; coordination benefits, data quality, and task
Differentiation and time elapsed since implementation

were not included in the exploratory factor analysis.
Differentiation is a formative construct as opposed to a
reflective construct, so there is no presumption that the
indicators are correlated (Bollen 1991; Chin 1998). We
have a single indicator for months since implementation.
efficiency for the second level, and overall impact

for the third level. We conducted an exploratory
factor analysis for each level, using the eigenvalue
cutoff of 1.00 to determine the number of factors,
and SAS Promax (oblique) rotation. In an
exploratory factor analysis, each question must
load more strongly on its primary factor than on its
secondary factor. Our criteria for screening items
using EFA were (1) that the item load on the
expected factor (i.e., load with the other items
intended to measure the intended construct) and
(2) that the loading on the primary factor must be
substantially greater (a difference of .50 or more)
than the loading on any other factor. Furthermore, we omitted any question that decreased the
reliability (Cronbachs alpha) of the scale of which
it is intended to be a part. As discussed below, the
survey items generally performed as intended.
Appendix A contains the wording of all items and
indicates the items that were dropped.
For the first level (Table 6), the interdependence
items loaded as intended. The questions for
customization split into two factors. After reexamining the items, it was clear that the focus of
items 1, 4, and 5 was on customization per se,
while for the other factor (tapped by questions 2
and 3) the focus was on plant level user participation in the implementation. We therefore
dropped the user participation factor because
factor 2, rather than the user participation factor, is
consistent with our a priori theoretical description
of customization (presented in the research model
section of the paper).
For the second level (coordination benefits, data
quality, and task efficiency) three factors were
extracted (Table 7). Task efficiency and coordination benefits separated cleanly. The questions
on data quality included both questions on
accuracy and questions on relevant data. These
all loaded on the same factor. While other
research has shown these to be two distinct
constructs, respondents in our study apparently do
not distinguish between the two. Two of the data
relevancy questions had high loadings on a secondary factor (a difference of less than .50 in
loadings on the primary and secondary factors).
Thus, these two questions were dropped as
problematic from a discriminant validity point of
569
Table 2. Frequency Breakdown

by Company Size in Employees
Table 3. Frequency Breakdown by

Respondents Job Function (Plant Level)
Company Size
Percent of
Total
1-1,500
29
Scheduler/Planner/Buyer
16
1,500-10,000
37
34
Materials Manager/ Purchasing

Manager
43
10,000+
Operations Manager
18
Plant Manager
11
Other Manufacturing Position
12
Table 4. Frequency Breakdown

by Software Vendor
Software Vendor
Percent of
Total
Job Function
Percent of
total
Table 5. Time Elapsed ERP Since

Implementation at Plant (ince go-live)
Time Elapsed (months)
Percent of
Plants
SAP
39
Less than 12
14.0
JD Edwards
19
12 to 17
21.7
QAD
18 to 23
13.2
Oracle
24 to 35
20.2
Baan
36 to 47
12.4
PeopleSoft
48 to 59
10.9
Other
21
60 to 72
7.8
570
Table 6. Rotated Factor Pattern: Model Level 1

3 Eigenvalues > 1
Factor1
Factor2
Factor3
CUSTOM1
0.07738
0.90494
0.15156
CUSTOM2
0.03274
0.07652
0.85956
CUSTOM3
0.00112
0.08509
0.88020
CUSTOM4R
0.07728
0.86658
0.00775
CUSTOM5
0.00813
0.86773
0.14714
INTER1
0.91012
0.02091
0.14463
INTER2
0.81811
0.01940
0.10844
INTER3
0.90680
0.02740
0.03864
INTER4
0.90052
0.01864
0.07271
INTER5
0.91360
0.00149
0.07501
INTER6R
0.83909
0.03367
0.04090
INTER7
0.88784
0.04985
0.14168
Table 7. Rotated Factor Pattern: Model Level 2

3 Eigenvalues > 1
Factor1
Factor2
ACC1R
0.65325
0.09558
0.12735
ACC2
0.84943
0.08159
0.02030
ACC3
0.83765
0.03781
0.01071
ACC4
0.81136
0.06097
0.01805
REL1
0.44071
0.11760
0.32895
REL2R
0.82871
0.02913
0.09850
REL3R
0.70097
0.06017
0.03251
REL4R
0.63481
0.02294
0.22796
COOR1
0.05358
0.87135
0.00005
COOR2
0.00108
0.87745
0.09741
COOR3
0.04262
0.93462
0.01362
COOR4
0.00890
0.92983
0.10807
TSK_EF1
0.07944
0.10779
0.85610
TSK_EF2
0.02698
0.07292
0.85658
TSK_EF3R
0.04889
0.06006
0.84416
0.17651
0.10620
0.67323
TSK_EF4
Factor3
571
Table 8. Rotated Factor Pattern:

Model Level 3
1 Eigenvalue > 1
IMPACT1
0.92118
IMPACT2
0.90726
IMPACT3R
0.67203
IMPACT4
0.93242
Table 9. Construct Reliabilities and Descriptive Statistics

No. of
items
Cronbachs
Alpha
Mean (1-7
scale)
Std. Dev.
Local level (plant) overall benefit
.93
4.7
1.3
Interdependence with other plants
.95
4.2
1.8
Improvements in coordination with other

plants
.93
4.4
1.7
Customization
.86
4.1
1.7
Task efficiency
.86
4.6
1.4
Data quality
.88
4.7
1.2
Construct
view. For the third level (overall impact), one

factor was extracted, as expected (Table 8).
Three of the questions loaded very highly on that
factor (above .90), and one question (impact3r)
had a loading of .67.
After EFA, we examined all of the remaining
questions to determine if the value of Cronbachs
alpha for any construct would be increased by
dropping the question. Only one item (impact3r,
the problematic item mentioned in the previous
paragraph), failed on this criterion so this item
was dropped. Cronbachs alpha and descriptive
statistics for the final scales appear in Table 9.
The constructs have strong discriminant validity
and reliability.
572
Testing the Hypotheses

Several researchers (Markus and Tanis 1999;
Ross and Vitale 2000) have presented compelling evidence that performance dips immediately after ERP goes live in an organization.
They suggest that organizations require substantial time to work through this initial shakedown phase and that performance during this
phase is not indicative of an organizations
longer-term ERP situation. Thus, we might
expect plants that have been running ERP less
than 1 year to have any number of unpredictable
problems that might mask the systematic, longerlived impacts that our model considers. Therefore, in order to focus on ERP impacts after the
break-in period, we excluded 18 plants whose

ERP systems had been in place for less than 1
year. The 1-year cutoff is somewhat arbitrary.
However, limited empirical evidence (e.g., Cosgrove Ware 2003; McAfee 2002) seems to
suggest that one year is sufficient time for
organizations to overcome break-in problems.
Any other cutoff would have to be more carefully
justified, and the ERP literature lacks the empirical data to provide such justification. Table 10
displays the correlations among the constructs in
the post 1-year data set.
We tested the hypotheses using regression (SAS
version 8.2, PROC REG procedure).4 A composite was created for each reflexive construct
(those for which factor analysis was performed)
by summing the factor-weighted indicator scores.
A straight sum was used for differentiation. The
months elapsed since ERP implementation is a
single indicator. The regression results are shown
in Table 11.5 Statistically significant coefficients
are highlighted in Figure 3.
Table 12 summarizes the results by hypothesis.
The model contained four hypotheses suggested
by organizational information processing theory:
two for interdependence and two for differentiation. H1a and H1b stated that interdependence
influences the degree to which coordination and
task efficiency benefits are realized when ERP is
The resulting sample size (n = 129 minus 18 equals n

= 111) raises questions about using structural equation
modeling (SEM) tools such as LISREL to test the
hypotheses. There is certainly disagreement about this.
Different authors recommend different minimum sample
sizes for LISREL: at least 100 (Hair et al. 1998), at
least 150 (Bollen 1989), at least 200 for degrees of
freedom of 55 or lower (MacCallum et al. 1996), or 5
times the number of parameters to be estimated
(Bagozzi and Edwards 1998). Therefore we used
regression. However, we compared LISREL with
regression results as a double-check. Due to the
difficulties in modeling interactions in SEM, we ran main
effects only regressions and compared them to a main
effects only SEM. The SEM results indicate a good fit
to the data (RMSEA .065, SRMR .071, RNI .964, NNFI
.959), and the parameter estimates are substantively
equivalent to the results found in the regression model.
5
Removing the interactions (which are nonsignificant)

does not substantively change the results.
implemented. This was supported for coordination improvement (H1a), but not supported for
task efficiency (H1b). H2a and H2b stated that
differentiation negatively influences the degree to
which intermediate benefits are realized when
ERP is implemented. This was supported for
both of the intermediate benefits we examined:
coordination improvement (H1a) and task
efficiency (H1b). H3a and H3b suggested that
customization would moderate the effect of
differentiation. The data do not support this;
however, there is a significant main effect of
customization on task efficiency. The model also
controlled for time elapsed since ERP
implementation (H4) and data quality (H5). Of
these, only data quality was a significant predictor
of coordination improvements, but both were
significant predictors of task efficiency. Finally,
coordination benefits, task efficiency and data
quality were all significant predictors of overall
local impact, supporting H6.
Post Hoc Analysis

Time Elapsed Since ERP implementation
The ERP literature contains considerable discussion about the effect of the time elapsed since
implementation on various ERP outcomes. Our
data allow us to do some exploratory analysis on
this issue in order to provide some potentially
useful information to future researchers. Our
original analysis showed that time elapsed since
ERP implementation is a significant antecedent of
task efficiency. To further explore this effect, we
segmented the observations into five categories
(using (1,0) coding) based the number of years
elapsed since implementation (0 to 11 months, 12
to 23 months, etc.). We ran a regression using
the full data set with only these categorical
variables as antecedents to task efficiency. The
results are in Table 13 and Figure 4, which plots
each coefficient. Each coefficient can be interpreted as the incremental impact of being in that
age group, compared with being in the 0 to 11
month age group. In other words, plants in the 12
to 23 month group have task efficiency scores
573
Table 10. Correlation Matrix (n = 111)

Overall
Coord
Task Eff
Data Q
Interdep
Differ
Custom
Time
Overall
Coord
.34
Task Eff
.77
.22
Data Q
.71
.22
.60
Interdep
.11
.81
.03
.02
-1.65
-.03
-.25
-.16
.10
Custom
.11
.17
.12
-.04
.14
.16
Time
.10
-.07
.29
-.02
-.10
.01
Differ
.07
Table 11. Regression Results (n = 111)

Dependent
Variable
p Value of
Regression
Adj
R2
Independent
Variables
Interdep
Differ
Coord
Improvement
0.000
0.000
0.71
0.48
Task
Efficiency
Local (plant)
level overall
benefits
574
0.000
0.71
Standardized
Regression
Coefficients
Unstandardized
Regression
Coefficients
.81
.44
15.31
< .001
< .05
t Value
.09
.06
1.78
Customization
.07
.10
1.41
DiffXCustom.
.06
.01
1.27
Time elapsed
.01
.01
0.28
DataQual
.19
.22
3.72
Interdep
.04
.02
0.55
Differ
p Value if
Significant
< .001
.19
.09
2.67
< .01
Customization
.14
.15
2.03
< .01
DiffXCustom.
.09
.01
1.32
Time elapsed
.30
.09
4.34
< .001
DataQual
.57
.48
8.16
< .001
CoordImprov
.15
.09
2.73
< .01
TaskEff
.52
.44
8.09
< .001
DataQual
.36
.25
5.61
< .001
Intermediate mechanisms
Interdependence
.81***
-.09*
Coordination
improvements
.15**
.19***
Differentiation
-.19**
.36***
Data quality
Customization
Time elapsed
since
implementation
.14**
.30***
.57***
Local (plant)
level overall
benefits
.52***
* p < .05
** p < .01
***p < .001
Task efficiency
Figure 3. Overview of Significant Regression Coefficients (Standardized)
Table 12. Results by Hypothesis

Supported
Hypothesis
H1
H2
H3
H4
H5
H6
Results
Yes
Interdependence greater intermediate

benefits
a. Coordination Improvements
Differentiation lower intermediate

benefits
b. Task efficiency
DifferXCustomization greater
intermediate benefits
b. Task efficiency
Time elapsed greater intermediate

benefits
b. Task efficiency
Data quality greater intermediate

benefits
b. Task efficiency
Intermediate benefits greater local

(plant) level overall benefits
a. Coordination benefits
b. Task efficiency
c. Data quality
b. Task efficiency
No
575
Table 13. Time Elapsed Since ERP Implementation as an Antecedent to Task

Efficiency (n = 129)
Category (years elapsed since ERP
implementation
Coefficient (improvement over 1-12 month

group)
12 to 23 months
2.62
24 to 35 months
3.96
36 to 47 months
5.16
48 or more months
5.44
2
Dependent Variable: Task Efficiency; Adjusted R = 0.10 ; p < .01; n = 129
Increm ental benefits above first

year benefits
6
5
4
3
2
1
0
1-11
12-23
24-35
36-47
48-60
Months since ERP "go l ive "
Figure 4. Effect of Time Elapsed Since ERP Implementation on Task Efficiency
that are 2.67 points higher than those in the 1 to

11 month group. Plants in the 24 to 35 month
group have scores 3.96 points higher; and so on.
These results clearly indicate that performance
improves over the first yearincreasing year by
year but at a decreasing rate.
2
Note that the adjusted R in Table 13 is 0.10. As

a point of comparison, we added the other
576
constructs from our a priori research model

predicting task efficiency (interdependence, differentiation, customization, and data quality). With
these constructs added to the regression model in
2
Table 13, the adjusted R jumps to .42. These
results are instructive. A number of published
articles have indicated that ERP impacts become
increasingly positive with time. Our results support
this. However, this analysis shows that elapsed
time is far from the whole story. By adding interdependence, differentiation, customization, and
data quality, we explain four times as much of the
variance in task efficiency as we can explain with
time elapsed alone.
Additional Antecedents Tested

We tested the impact of several other possible
explanatory factors: firm size, division size, plant
size (all measured as number of employees), ERP
vendor, specific modules implemented, number of
modules implemented, and number of plants in the
implementation. We added these possible control
variables one by one to the a priori regression
models in Table 11. (We used dummy coding for
categorical variables). Each variable was tested
separately for effect on coordination improvements, data quality, task efficiency, and plant level
overall benefits. We had no a priori notions about
the directionality of effects so we used twosided
tests at p = .05. None of the potential predictors
were significant, even without correcting for
multiple comparisons.
In addition, we investigated the effect of several
possible interactions. First we investigated interactions between time elapsed since ERP implementation and our focal constructs of interdependence and differentiation. The question of interest
here is whether the coordination or task efficiency
impacts of interdependence or differentiation
change with timefour interactions in all. Beyond
these, we explored the possibility of two other
interactions involving interdependence: (1) an
interaction between interdependence and the
number of plants in the ERP implementation and
(2) an interaction between interdependence and
the number of ERP modules. Each of these could
be expected to increase coordination benefits or
task efficiency.
In each case we added the interactions one at a
time to the baseline regression model in Table 10.
To avoid getting significant results by chance, we
used a threshold of p < .05 adjusted for multiple
comparisons (.05 divided by six interaction terms).
Although the effect on coordination benefits of
differentiation time and interdependence

number-of-plants were significant (p < .05) when
uncorrected for multiple comparisons, none of the
six interactions were significant when corrected.
Discussion
Our model shows that, among plants that have
cleared the implementation hurdle, task efficiency,
coordination improvements, and data quality
2
explain a tremendous amount (R = .71, Table 11)
of the variance in overall plant level benefits.
Further we explain a substantial amount of the
variance in two of these three predictors of overall
plant level benefits.6
Differentiation explains a significant amount of
plant-to-plant variation in both of the intermediate
benefits we sought to predict (coordination
improvements and task efficiency). Interdependence only affects coordination improvements.
However, the cumulative impacts of interdependence and differentiation on plant level overall
benefits are roughly equivalent to one another
(Multiplying the standardized regression coefficients on the path from interdependence to local
level overall benefits in Figure 3 above yields .81
.15 = .12. Doing the same on the paths from
differentiation to local level overall benefits yields
.09 .15 + .19 .52 = .11).
We suggested that one way to respond to distinct
local level needs is through ERP customization.
However, our results do not show that customization moderates that effect of differentiation. On
the other hand, our findings do suggest that ERP
customization (as a main effect) can improve local
efficiency. We should point out that our power for
the interaction analysis is less that .30 (effect size
of approximately .015). In other words, there is a
less than one in three chance that our sample size
could detect a significant interaction if one exists.
In the data collection section, we discussed the possibility of a bias in our sample toward relatively successful
ERP implementations. The relationships we find may be
most valid among such plants.
577
Continuing to examine the intermediate benefits,

task efficiency increases with the amount of time
elapsed since ERP implementation. Data quality
affects both coordination improvements and task
efficiency.
Contributions to Academia
According to a number of literature reviews (e.g.,
Dong et al. 2002; Esteves 2001; Jacobs and
Bendoly 2003), the majority of published ERP
research is descriptive or prescriptive. These
studies have amassed valuable findings, but there
is also a need to place the ERP phenomenon in
the context of existing theoretical frameworks and
to generate and test hypotheses.
We have provided a large-sample, cross-sectional
study that investigates relationships posited by
previous OIPT researchers. Tushman and Nadler
(1978) proposed that interdependence and
differentiation should affect the fit between particular information processing mechanisms and
particular companies or subunits. Goodhue et al.
(1992) applied these ideas to a particular information processing mechanism: data integration. Our
study confirms the relationships posited by these
earlier scholars. The non-finding of Wybo and
Goodhue (discussed earlier) regarding the
absence of a relationship between interdependence and managerial decisions to implement data
integration makes our findings on the impact of
integration and differentiation on data integration
outcomes particularly important for scholars
interested in OIPT.
Our focus on differentiation within the organization
adds to the body of ERP-organization alignment
research (Hong and Kim 2002; Koh et al. 2000;
Sia and Soh 2002; Soh et al. 2000; Somers and
Nelson 2003). This stream has taught the
research community much about misfits, their
sources, their severity, and their resolution. However, most of this research has focused on the
alignment of the assumptions in the ERP package
with the firm or divisiontreated as a single unit of
analysis. Our results suggest that misfits between
the organizations ERP configuration and the indi-
578
vidual subunit within the company are also important. The development of a measure of differentiation among manufacturing plants is an important
contribution. Of course, this measure could benefit from additional testing and refinement.
To further relate our study to other ERP scholarship, we compare our perspective and results to
the suggestions of Markus and Tanis (1999).
They consider three theoretical perspectives
through which one could view ERP: rational actor,
external control, and emergent process. They
advocate the third perspective because unpredictable events (such as project champion or project
leader departures) and external factors (such as
vendor failure) can seriously affect ERP at any
stage in its lifecycle. On its surface, our study
seems to contradict Markus and Tanniss
assertion. We average across many unpredictable
events and, using only a few predictor variables,
we are able to explain a large portion of the
variance in ultimate impactat least in one
functional area. But if we remember that our
survey data may have a response bias that
focuses us on those firms that have successfully
negotiated the difficulties of implementation, it
becomes clear that our findings do not necessarily
contradict Markus and Tanniss assertions. Their
focus was primarily on implementation success,
where an emergent process perspective is
arguably essential, because it is so hard to predict
implementation results in advance. Our focus is
on the impacts of those ERP systems after they
are successfully implemented. Our results suggest that a rational actor perspective is justified
when studying successful ERP systems.
Contributions to Practice
At the beginning of this paper, we invoked Markus
and Tanniss executive, who asks, Will ERP pay
off? While there is already some evidence (e.g.,
the studies by Hitt et al. and Anderson et al. discussed earlier) that on average the answer is
yes, we suggested that the executive will also
want to understand why ERP results vary from
company to company. Clearly the first part of the
answer is implementation practicesan area
others address but we do not. However, our

model explains much of the plant-to-plant variation
in ERP impacts among plants that have implemented ERP successfully.
The influence of interdependence suggests that it
is a mistake to expect large coordination-related
benefits to accrue automatically from successfully
implementing ERP, even though these benefits are
highly touted by vendors and in the trade press.
The level of interplant interdependence varies
among and within organizations; therefore, the
potential for reaping coordination-related benefits
from ERP varies as well. This finding may be
particularly instructive for firms that implement
ERP mainly for technical reasons (e.g., replacing
aging infrastructure). In such cases, collateral
impacts, such as coordination benefits, may not
automatically materialize. Put differently, Wybo
and Goodhue (1995) suggest that managers have
not based decisions about integration on interdependence; our study suggests that they should.
Our findings for differentiation show that ERP can
create operational difficulties for a subunit that
differs from its peers in its products and manufacturing processes. Certainly some local level
protests about new information systems are rooted
in reluctance to change, in power dynamics, and
so on; however, our results suggest that IS implementers should not dismiss out of hand managers
who claim, It will not work here because we are
different.
While we do not find that customization moderates
the effect of differentiation, we do find that customization by itself has an impact (as a main
effect). This is important since exhortations
against customization are plentiful in industry and
academia (e.g., Pereira 1999). Note, however,
that we do not investigate whether this benefit of
customization outweighs the initial and ongoing IT
costs related to programming, potential future
upgrade difficulties, and other risks. Since differentiation does not influence the relationship
between customization and benefits, we should
acknowledge the possibility that our questionnaire
did not measure customization for the benefit of
the individual plant (as we intended), but rather
measured customization directed at making the

ERP fit better with the organization as a whole.
Our finding that ERP benefits improve with time is
significant because it runs counter to hints in the
results from Hitt et al. (2002) about ERP benefits
after implementation. Those researchers found
that after implementation some performance
indicators seemed to drop back toward previous
levels. However, Hitt et al. are quite cautious
about this particular finding due to their few after
implementation data points (p. 84).
ERP-enabled interplant coordination improvements lead to local level overall ERP benefits.
However, coordination benefits do not predict
overall ERP benefits as strongly as do task
efficiency and data quality (as the standardized
regression coefficients in Figure 2 indicate). This
limited role is surprising, since improving coordination among subunits is an often-cited motivation
for implementing ERP. One possible explanation
is that, at the plant level, the operational consequences of the interdependence between plants
have been well understood and addressed by
existing practices, such as those typical of just-intime. If that is the case, ERP systems may not
add as much incremental coordination benefit
within the production function (Flynn and Flynn
1999; Piszczalski 2000). However, we suspect
that coordination-related benefits are more important for other types of relationshipsfor example,
between plants and sales or between engineering
and purchasing.
Limitations and Future Research

Our study examines varying levels of interdependence and differentiation among plants with ERP
systems. We did not attempt to measure different
levels of data integration. Testing a model that
captures different levels of data integration is an
important item for future research.
We focus at the local (plant) level. As we mentioned earlier, our analysis does not include global
costs and benefits, such as the ability to quickly
answer corporate-wide questions involving multiple
579
plants. Additionally, the plant level is not an appropriate level of analysis for capturing IT costs such
programming and maintenance costs. However,
given the arguments of Barua et al. and others discussed in our introduction, we believe that the
tradeoffs entailed in a local focus enable us to
make worthwhile contributions.
Our measures of customization consider the role
of this variable generally. Including measures that
distinguished between different customization
strategies would have increased our understanding of customization and may have increased
the likelihood of detecting a stronger statistical
effect of customization. We also recognize that
ERP packages may have more built-in flexibility
and capabilities than the companies represented
in our data are using. However, our goal was to
understand ERP as enacted by businesses.
Finally, we treat ERP systems as a class. We
assume that the similarities among this class of
systems are more important than the differences.
However, research that focuses on differences
among systems is also important.
Conclusion
Past research has suggested that, on average,
ERP outcomes appear to be positive at the organization level. However, the effects of ERP vary
considerably from company to company. We
sought explore some of the reasons why. We
argued that gaining this type of understanding is
facilitated by understanding why some local level
intermediate benefits do (and do not) occur.
Organizational information processing theory
guided this inquiry. The theory suggests that interdependence and differentiation both affect the
level of benefit that occurs from data and process
integration. Our analysis of data from 111 manufacturing plants supports the notion that interdependence is associated with increased plantlevel benefits from ERP while differentiation is
associated with the opposite. Our data also show
a positive role for customization (considered
generally) and the amount of time elapsed since
ERP implementationtwo variables that have
been the subject of much discussion in the ERP
literature.
580
References
Akkermans, H., and van Helden, K. Vicious and
Virtuous Cycles in ERP Implementation: A
Case Study of Interrelations Between Critical
Success Factors, European Journal of
Information Systems (11), 2002, pp. 35-46.
Anderson, M. C., Banker, R. D., and Ravindran, S.
The New Productivity Paradox, Communications of the ACM (46:3), 2003, pp. 91-94.
Bagozzi, R. P. Causal Models in Marketing, John
Wiley and Sons, New York, 1980.
Bagozzi, R. P., and Edwards, J. A General Approach for Representing Constructs in Organizational Research, Organizational Research
Methods (1:1), 1998, pp. 45-87.
Barua, A., Kriebel, C. H., and Mukhopadhyay, T.
Information Technologies and Business Value:
An Analytic and Empirical Investigation, Information Systems Research (6:1), 1995, pp. 3-23.
Berry, W., Bozarth, C., Hill, T., and Klompmaker,
J. Factory Focus: Segmenting Markets from
an Operations Perspective, Journal of Operations Management (10), 1991, pp. 363-387.
Berry, W. L., and Hill, T. Linking Systems to
Strategy, International Journal of Operations
and Production Management (12:10), 1992, pp.
3-15.
Bollen, J. A. Structural Equations with Latent Variables, John Wiley and Sons, New York, 1989.
Bollen, K. Conventional Wisdom on Measurement: A Structural Equation Perspective, Psychological Bulletin (110:2), 1991, pp. 305-314.
Brown, C., and Vessey, I. ERP Implementation
Approaches: Toward a Contingency Framework, in Proceedings of the 20th International
Conference on Information Systems, P. De and
J. I. DeGross (Eds.), Charlotte, NC, 1999, pp.
411-416.
Chin, W. W. Issues and Opinion on Structural
Equation Modeling, MIS Quarterly (22:1), 1998,
pp. vii-xvi.
Cooke, D. P., and Peterson, W. J. SAP Implementation: Strategies and Results, Report R1217-98-RR, The Conference Board, New York,
July 1998.
Cosgrove Ware, L. By the Numbers: Enterprise
Systems Show Results, CIO Magazine,
November 1, 2003 (available online at http://
www.cio.com/archive/110103/tl_numbers.html).
Daft, R. L., and Lengel, R. H. Organizational

Information Requirements, Media Richness and
Structural Design, Management Science
(32:5), 1986, pp. 554-571.
Davenport, T. Putting the Enterprise in the Enterprise System, Harvard Business Review (76:4),
1998, pp. 121-131.
Dong, L., Neufield, D., and Higgins, C. The
Iceberg on the Sea: What Do You See, in
Proceedings of the 8th Americas Conference on
Information Systems, R. Ramsower and J.
Windsor, Dallas, TX, 2002, pp. 857-864.
Esteves, J. Enterprise Resource Planning Systems Research: An Annotated Bibliography,
Communications of the AIS (7:8), 2001.
Flynn, B. B., and Flynn, E. J. Information Processing Alternatives for Coping with Manufacturing
Environmental Complexity, Decision Sciences
(30:4), 1999, pp. 1021-1052.
Galbraith, J. R. Designing Complex Organizations, Addison-Wesley, Reading, MA, 1973.
Galbraith, J. R. Organization Design, AddisonWesley, Reading, MA, 1977.
Gattiker, T. F. Anatomy of an ERP Implementation Gone Awry, Production and Inventory
Management Journal (43:3/4), 2002, pp. 96105.
Gattiker, T. F., and Goodhue, D. L Software
Driven Changes to Business Processes: An
Empirical Study of Impacts of Enterprise
Resource Planning Systems at the Local Level,
International Journal of Production Research
(40:18), 2002, pp. 4799-4814.
Gattiker, T. F., and Goodhue, D. L. Understanding the Local Level Costs and Benefits of
Erp Through Organizational Information Processing Theory, Information and Management
(41:4), 2004, pp. 431-443.
Gattiker, T. F., and Goodhue, D. L. Understanding the Plant Level Costs and Benefits of
ERP Systems: Will the Ugly Duckling Always
Turn into a Swan?, in Proceedings of the 33rd
Hawaii International Conference on System
Sciences, R. H. Sprague (Ed.), IEEE Computer
Society Press, Los Alamitos, CA, 2000, pp.
7017-7027.
Gefen, D. Nurturing Clients Trust to Encourage
Engagement Success During the Customization
of ERP Systems, OMEGAInternational Jour-
nal of Management Science (30), 2002, pp.

287-299.
Goodhue, D. L., Wybo, M. D., and Kirsch, L. J.
The Impact of Data Integration on the Costs
and Benefits of Information Systems, MIS
Quarterly (16:3), 1992, pp. 293-311.
Hair, J. F., Anderson, R. E., Tatham, D. L., and
Black, W. C. Multivariate Data Analysis, Prentice Hall, Upper Saddle River, NJ, 1998.
Hayes, R., and Wheelwright, S. Link Manufacturing Processes and Product Life Cycles,
Harvard Business Review (57:1), 1979, pp.
133-140.
Hill, T. Manufacturing Strategy: Text and Cases
(3rd ed.), Irwin, Boston, 2000.
Hirt, S. G., and Swanson, E. B. Adopting SAP at
Siemens Power Corporation, Journal of Information Technology (14:3), 1999, pp. 243-252.
Hitt, L. M., Wu, D. J., and Zhou, X. Investment in
Enterprise Resource Planning: Business Impact
and Productivity Measures, Journal of Management Information Systems (19:1), 2002, pp.
71-98.
Holland, C. P., and Light, B. A Critical Success
Factors Model for ERP Implementation, IEEE
Software (16:3), May-June 1999, pp. 30-36.
Hong, K.-K., and Kim, Y.-G. The Critical Success
Factors for ERP Implementation: An Organizational Fit Perspective, information and Management (40:1), 2002, pp. 25-40.
Jacobs, F. R., and Bendoly, E. Enterprise
Resource Planning: Developments and Directions for Operations Management Research,
European Journal of Operational Research
(146:2), 2003, pp. 233-240.
Jacobs, F. R., and Whybark, C. Why ERP?, Irwin/
McGraw-Hill, New York, 2000.
Koh, C., Soh, C., and Markus, M. A Process
Theory Approach to Analyzing ERP Implementation and Impacts: The Case of Revel
Asia, Journal of Information Technology Cases
and Applications (2:1), 2000, pp. 4-23.
Kumar, V., Maheshwar, B., and Kumar, U.
Enterprise Resource Planning Systems Adoption Process: A Survey of Canadian Organizations, International Journal of Production
Research (40:3), 2002, pp. 509-522.
Mabert, V. A., Soni, A., and Venkataramanan, M.
A. Enterprise Resource Planning Survey of
U.S. Manufacturing Firms, Production and
581
Inventory Management Journal (41:2), 2000,

pp. 52-58.
MacCallum, R., Browne, M., and Sugawara, H.
Power Analysis and Determination of Sample
Size for Covariance Structure Modeling, Psychological Methods (1:2), 1996, pp. 130-149.
Markus, M. L., and Tanis, C. The Enterprise System Experience: From Adoption to Success, in
Framing the Domains of IT Management, R. W.
Zmud (Ed.), Pinnaflex Educational Resources,
Cincinnati, OH, 1999, pp. 173-207.
McAfee, A. The Impact of Enterprise Technology
Adoption on Operational Performance: An
Empirical Investigation, Production and Operations Management (11:1), 2002, pp. 33-53.
META Group. Market Research: The State of
ERP Services (Executive Summary), META
Group, Inc., Stamford, CT, 2004.
Miller, D. Environmental Fit Versus Internal Fit,
Organization Science (3:2), 1992, pp. 159-178.
Miller, J. Fit Production Systems to the Task,
Harvard Business Review, (59:1) JanuaryFebruary 1981, pp. 145-154.
Ng, J., Ip, W., and Lee, T. A Paradigm for ERP
and BPR Integration, International Journal of
Production Research (9), 1999, pp. 2093-2108.
Pereira, R. E. Resource View Theory Analysis of
SAP as a Source of Competitive Advantage,
DATA BASE (30:1), 1999, pp. 38-46.
Piszczalski, M. Lean vs. Information Systems,
Automotive Manufacturing & Production
(112:8), August 2000, pp. 26-28.
Robey, D., Ross, J. W., and Boudreau, M. C.
Learning to Implement Enterprise Systems: An
Exploratory Study of the Dialectics of Change,
Journal of Management Information Systems
(19:1), 2002, pp. 17-46.
Ross, J. W., and Vitale, M. The ERP Revolution:
Surviving Versus Thriving, Information Systems Frontiers (2:2), November 2000, pp. 233241.
Safizadeh, M. H., Ritzman, L. P., Sharma, D., and
Wood, C. An Empirical Analysis of the
Product-Process Matrix, Management Science
(42:11), 1996, pp. 1576-1587.
Scheer, A., and Habermann, F. Making ERP a
Success: Using Business Process Models to
Achieve Positive Results, Communications of
the ACM (43:4), 2000, pp. 57-61.
582
Scott, F. and Shepherd, J. The Steady Stream of

ERP Investments, AMR Research Outlook,
August 26, 2002 (available online at http://
www.amrresearch.com/).
Sia, S. K., and Soh, C. Severity Assessment of
ERP-Organization Misalignment: Honing in on
Ontological Structure and Context Specificity,in
Proceedings of the 23rd International Conference on Information Systems, L. Applegate,
R. Galliers, and J. I. DeGross (Eds.), Barcelona,
2002, pp. 723-729.
Soh, C., Kien, S. S., and Tay-Yap, J. Cultural Fits
and Misfits: Is ERP a Universal Solution,
Communications of the ACM (43:4), 2000, pp.
47-51.
Somers, T. M., and Nelson, K. G. The Impact of
Strategy and Integration Mechanisms on
Enterprise System Value: Empirical Evidence
from Manufacturing Firms, European Journal of
Operational Research (146:2), 2003, pp. 315338.
Staehr, L., Shanks, G., and Seddon, P. Understanding the Business Benefits of Enterprise
Resource Planning Systems, in Proceedings of
the 8th Americas Conference on Information
Systems, R. Ramsower and J. Windsor (Eds.),
Dallas, TX, 2002, pp. 899-905.
Sumner, M. Critical Success Factors in Enterprise Wide Information Management Projects,
in Proceedings of the 5th Americas Conference
on Information Systems, W. D. Haseman and
D. L. Nazareth (Eds.), Milwaukee, WI, 1999, pp.
232-234.
Taylor, S. G. Are the Process Industries Different?, in Proceedings of the 23rd Conference
of the American Production and Inventory
Control Society (APICS), Los Angeles, 1980,
pp. 94-96.
Thompson, J. D.
Organizations in Action,
McGraw-Hill, New York, 1967.
Turner, J. The Unique Requirements of Process
Industry ERP, in APICS: The Performance
Advantage (8:6), 1998, pp. 58-60.
Tushman, M. L., and Nadler, D. A. Information
Processing as an Integrating Concept in
Organizational Design, Academy of Management Review (3), 1978, pp. 613-624.
Van Dierdonck, R., and Miller, J. G. Designing
Production Planning and Control Systems,
Journal of Operations Management (1:1), 1980,

pp. 37-46.
Vollmann, T. E., Berry, W. L., and Whybark, D. C.
Manufacturing Planning and Control Systems
(4th ed.), Irwin, New York, 1992.
Vosburg, J., and Kumar, A. Managing Dirty Data
in Organizations Using ERP: Lessons from a
Case Study, Industrial management and Data
Systems (101:1), 2001, pp. 21-31.
Wybo, M. D., and Goodhue, D. L. Using Interdependence as a Predictor of Data Standards:
Theoretical and Measurement Issues, Information and Management (29), 1995, pp. 317-328.
About the Authors

Thomas F. Gattiker is an assistant professor in
the Networking, Operations and Information Systems Department at Boise State University, where
he teaches operations management and procurement/supply chain management (he was formerly
employed at Miami Universtiy). His research
appears in The International Journal of Production
Research, Information and Management, Production and Inventory Management Journal, Quality
Management Journal, and The Decision Sciences
Journal of Innovative Education. His current
research is in the application of information technology to the operations and supply chain areas.
He was the 1999 APICS E&R Foundation George
and Marion Plossl Fellow.
Dale L. Goodhue is a professor and Department
Head of the MIS Department, and the C. Herman
and Mary Virginia Terry Chair of Business Administration at the University of Georgias Terry College of Business. He has published in Management Science, MIS Quarterly, Decision Sciences,
Sloan Management Review and other journals,
and is an associate editor for Management
Science and a former associate editor for the MIS
Quarterly. His research interests include measuring the impact of information systems, the
impact of task-technology fit on individual performance, the management of data and other IS
infrastructures/resources, and the impact of ERP
systems on organizations.
Appendix A
Questionnaire Items
The items used to test the model are shown below, with their sources, if any. Except as noted, each
question used a Likert scale: 1 = strongly disagree to 7 = strongly agree. R indicates reverse scoring for
the analysis. Instead of being sorted by construct, as they are below, the items were intermixed on the
actual questionnaire. Time elapsed since implementation was measured with the single open-ended item,
how long (in months) has ERP been running live at this plant.
Based on manufacturing strategy and manufacturing systems literature, we identified and refined a list of
critical product and process characteristics (Table A2). Our practitioner interview process, described in the
methodology section of the paper, resulted in further refinements. For each characteristic on the list, survey
respondents were asked to describe how their plant compares to other plants that are a part of their ERP
implementation on a nine-point response scale (1 = plant has much less of the characteristic; 9 = plant has
much more of the characteristic). In addition to our ex ante practitioner interviews, we examined correlations among the items in our data in order to ensure that they related to one another in a fashion that is
consistent with theory. The items performed as expected except for volume, which was dropped. In the
pretest, many practitioners interpreted the item as overall manufacturing volume rather than volume per unit
583
(volume per end item or end configuration). We attempted to reword the item to solve this problem, but
apparently we did not succeed.
Table A1. Reflexive Survey Items

Data Quality
ACC1R
The information from the ERP system has numerous accuracy problems that make it
difficult for employees to do their jobs (based on Goodhue 1995)
ACC2
The information that the ERP system provides to employees in this plant is accurate
ACC3
The data plant employees receive from the ERP system is true
ACC4
The ERP data that plant employees (planners, supervisors, etc.) use or would like to
use are accurate enough for their purposes (based on Goodhue 1995)
REL1
The data that the ERP system provides is exactly what plant employees need to carry
out their tasks (item deleted)
REL2R
It is difficult for plant employees to do their jobs effectively because some of the data
they need is missing from the ERP system
REL3R
The data accessible from the ERP system lacks critical information that would be useful
to plant employees
REL4
The ERP system provides the right data to meet plant employees needs (item deleted)
Task Efficiency
TIME1
Since we implemented ERP, plant employees such as buyers, planners and production
supervisors need less time to do their jobs
TIME2
ERP saves time in jobs like production, material planning and production management
TIME3R
Now that we have ERP it is more time-consuming to do work like purchasing, planning
and production management
TIME4
ERP helps plant employees like buyers, planners, and production supervisors to be
more productive
Overall Business Impact of ERP on the Plant
IMPACT1
In terms of its business impacts on the plant, the ERP system has been a success
IMPACT2
ERP has seriously improved this plants overall business performance
IMPACT3R
From the perspective of this plant, the costs of ERP outweigh the benefits
(Item Deleted)
IMPACT4
ERP has had a significant positive effect on this plant

ERP Customization to Meet Local Plant Needs
CUSTOM1
The ERP system was altered to improve its fit with this plant
CUSTOM2
The ERP implementation team was responsive to the needs of this plant (item deleted)
CUSTOM3
Individuals from this plant had a great deal of influence on how the ERP system was
set-up. (item deleted)
CUSTOM4R
A standard version of the ERP software was implemented without changes being made
to fit the particular requirements of this plant
CUSTOM5
When the ERP system was being implemented in this plant, the package was changed
to better meet the needs of this plant
584
Table A1. Reflexive Survey Items (Continued)

Interdependence with Other Plants
INTER1
To be successful, this plant must be in constant contact with these other plants
INTER2
If this plants communication links to these other plants were disrupted things would
quickly get very difficult.
INTER3
Frequent information exchanges with these other plants are essential for this plant to
do its job
INTER4
Close coordination with these other plants is essential for this plant to successfully do
its job
INTER5
Information provided by these other plants is critical to the performance of this plant
(based on Wybo and Goodhue 1995)
INTER6R
This plant works independently of these other plants
INTER7
The actions or decisions of these other plants have important implications for the
operations of this plant (based on Wybo and Goodhue 1995)
Improvements in Coordination with Other Plants
CB1
ERP helps this plant adjust to changing conditions within these other plants
CB2
ERP has improved this plants coordination with these other plants
CB3
ERP makes this plant aware of important information from these other plants
CB4
ERP helps this plant synchronize with these other plants
Table A2. Differentiation Items

Volume
Number of units produced monthly per model or configuration or formulation (item

dropped)
Variety
The number of different model numbers, configurations or formulations produced
Part number
complexity
The number of different active part numbers or material code numbers, excluding
finished goods part numbers or finished goods code numbers
BOM
complexity
Number of levels in the typical bill of materials
Postponeme
nt strategy
The degree to which products are made to customer specifications, instead of to

stock
Design
stability
The average number of design changes per month
NPIs
The number of new design introductions per month
Mfg. Cycle
time
The average amount of time that passes between the time an order is put into
production and the time it is completed
Lot control
The need to identify or segregate material by individual piece or lot rather than
merely by part number
Dominant
technology
Amount of production activity dedicated to processing (blending, purifying,

converting, etc.) as opposed to assembly or fabrication.
585

Sesi 03 - 1 - Article ERP and Its Impact

Transféré par

Informations du document

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Sesi 03 - 1 - Article ERP and Its Impact

Transféré par

Droits d'auteur :

Formats disponibles

Gattiker & Goodhue/After ERP Implementation

WHAT HAPPENS AFTER ERP IMPLEMENTATION:

phase has occurred. Organizational information

We present a model of the organizational impacts

Peter Todd was the accepting senior editor for this

Keywords: Organizational information processing

MIS Quarterly Vol. 29 No. 3, pp. 559-585/September 2005

Gattiker & Goodhue/After ERP Implementation

MIS Quarterly Vol. 29 No. 3/September 2005

that are highly interdependentthat is, very

Intermediate Benefits at Individual

Gattiker & Goodhue/After ERP Implementation

Figure 1. Conceptual Model of ERP Impact at the Local Level

the firm. By studying the factors leading to these

interdependence between a particular plant and

MIS Quarterly Vol. 29 No. 3/September 2005

Gattiker & Goodhue/After ERP Implementation

MIS Quarterly Vol. 29 No. 3/September 2005

paper, we believe there are three important

Gattiker & Goodhue/After ERP Implementation

Figure 2. Research Model

mation across subunit boundaries within the

or capacity requirements planning system).

MIS Quarterly Vol. 29 No. 3/September 2005

Gattiker & Goodhue/After ERP Implementation

to which intermediate benefits from implementing

MIS Quarterly Vol. 29 No. 3/September 2005

Configuration is the process of setting software switches

Gattiker & Goodhue/After ERP Implementation

pounds on hand rather than just the number of

different part numbers in different plants). Misfits

MIS Quarterly Vol. 29 No. 3/September 2005

Gattiker & Goodhue/After ERP Implementation

Translating differentiation (which OIPT theorists

Control Variable: ERP

MIS Quarterly Vol. 29 No. 3/September 2005

We conceptualize customization as changes to the

Control Variable: Time Elapsed Since

Gattiker & Goodhue/After ERP Implementation

Table 1. ERP Misfit Resolution Strategies (adapted from Figure 1 in C. Soh, S. S.

Accept shortfall in ERP functionality (compromising on the requirements of the organization)

Manual (by hand rather than using a computer system)

Customization to achieve the required functionality

Non-core customization (interfacing with add-on module or through query/report writer

Core customization to amend the base code

H4a (control hypothesis): In a plant

benefits. This is the basis for the following

Control Variable: Data Quality

Overall Benefits at the Local Level

Our ultimate dependent variable is local (plant)

MIS Quarterly Vol. 29 No. 3/September 2005

Gattiker & Goodhue/After ERP Implementation

H6b: In a plant within an ERP implementation, greater improvements in

MIS Quarterly Vol. 29 No. 3/September 2005

in 43 items used to measure the 8 constructs in

Gattiker & Goodhue/After ERP Implementation

organizations that have been relatively successful

Differentiation and time elapsed since implementation

efficiency for the second level, and overall impact

MIS Quarterly Vol. 29 No. 3/September 2005

Gattiker & Goodhue/After ERP Implementation

Table 2. Frequency Breakdown

Table 3. Frequency Breakdown by

Materials Manager/ Purchasing