8. Erp Managing The Implementation Process

European Journal of Operational Research 146 (2003) 302–314
www.elsevier.com/locate/dsw

Enterprise resource planning: Managing
the implementation process
Vincent A. Mabert 1, Ashok Soni *, M.A. Venkataramanan 2

Department of Operations and Decision Technologies, Kelley School of Business, Indiana University, 1309 E. Tenth St.,
Bloomington, IN 47405-1701, USA

Abstract

Over the past few years, thousands of companies around the world have implemented enterprise resource planning
(ERP) systems. Implementing an ERP system is generally a formidable challenge, with a typical ERP implementation
taking anywhere from one to five years. The story of the success of ERP systems in achieving the stated objectives is
mixed. Some companies have had very successful implementations while others have struggled. This paper empirically
investigates and identifies key differences in the approaches used by companies that managed their implementations on-
time and/or on/under-budget versus the ones that did not using data collected through a survey of US manufacturing
companies that have implemented ERP systems. Logistic regressions are used to classify on-time and on/under-budget
firm groups based on the survey responses and to identify the significant variables that contribute to on-time and on/
under-budget implementation performance. The results indicate that many different factors ranging from pre-imple-
mentation planning to system configuration influence performance, which managers should be sensitive about when
implementing major systems like ERP.
Ó 2002 Elsevier Science B.V. All rights reserved.

Keywords: Enterprise resource planning; Survey methodology; Logistic regression models

1. Introduction vide, at least in theory, seamless integration of
processes across functional areas with improved
Since the mid-1990s, enterprise resource plan- workflow, standardization of various business
ning (ERP) systems have been installed in thou- practices, and access to real-time up-to-date data.
sands of companies worldwide. ERP systems are ERP systems are complex and implementing one
enterprise-wide on-line interactive systems that can be a challenging, time consuming and expen-
support cross-functional processes using a com- sive project for any company (Davenport, 1998).
mon database. ERP systems are designed to pro- An ERP implementation can take many years to
complete, and cost tens of millions of dollars for a
*
moderate size firm and upwards of $100 million
Corresponding author. Tel.: +1-812-855-3423; fax: +1-812- for large international organizations (Mabert et al.,
856-5222.
E-mail address: soni@indiana.edu (A. Soni).
2000). Even with significant investments in time
1
Tel.: +1-812-855-2661. and resources, there is no guarantee of a successful
2
Tel.: +1-812-855-3491. outcome.

0377-2217/02/$ - see front matter. Ó 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 7 - 2 2 1 7 ( 0 2 ) 0 0 5 5 1 - 9

V.A. Mabert et al. / European Journal of Operational Research 146 (2003) 302–314 303

Despite the large installed base of ERP systems, Everdigen et al. in a survey of 2647 European
academic research in this area is relatively new. companies across all industry types determined
Like many other new Information Technology adoption and penetration of ERP by functionality.
(IT) areas, much of the initial literature in ERP Mabert et al. surveyed manufacturing companies
consists of articles or case studies either in the in the US to study penetration of ERP, moti-
business press or in practitioner focused journals. vation, implementation strategies, modules and
Many of these articles provide anecdotal infor- functionalities implemented, and operational ben-
mation based on a few successes or failures. These efits in the manufacturing sector.
publications have chronicled both some high pro- While the above practitioner and academic re-
file failures and extensive difficulties at such search provides valuable insights into both ERP
companies as FoxMeyer and Hershey Food Cor- effective use and implementation process issues, we
poration (Deutsch, 1998; Diederich, 1998; Nelson feel a more systematic empirical analysis of ERP
and Ramstad, 1999), and some model implemen- implementations is essential for understanding key
tations (Kirkpatrick, 1998). Also, several authors factors that lead to a successful implementation, as
(Piturro, 1999; Trunk, 1999; Zuckerman, 1999) measured by on-time and on/under-budget per-
emphasize that ERP is a key ingredient for gaining formance. This paper addresses this issue by re-
competitive advantage, streamlining operations, porting on and analyzing the results of a survey of
and having ‘‘lean’’ manufacturing. As a testimo- companies who have implemented ERP systems.
nial for this viewpoint, they point to tens of More specifically, this paper empirically investi-
thousands of companies around the world who gates whether there are key differences in the
have implemented or are planning to implement approaches by companies that managed their im-
ERP systems. plementations ‘‘on-time’’ and/or ‘‘on/under-budget’’
More recently, several academically oriented versus the firms that did not. These are two success
papers have dealt with various aspects of ERP measures often cited by companies for ERP im-
(Davenport, 1998; McAfee, 1999; Stratman and plementations (Mabert et al., 2000, 2001). Logistic
Roth, 1999; Van Everdingen et al., 2000; Mabert regression models are used to classify companies
et al., 2000, 2001). Davenport looks at the reasons that are able to accomplish their implementations
for implementing ERP systems and the challenges on-time and then on/under-budget based upon a
of the implementation project itself. McAfee, and set of input variables. All results are based on the
Stratman and Roth both look at operational per- responses from this survey.
formance. McAfee reports on a longitudinal ex- The rest of the paper is organized as follows: In
periment at a computer manufacturing facility to the next section we discuss the research issues and
determine the impact of an ERP system on op- framework employed to conduct the investigation,
erational performance. His research shows that and the relevant research germane to this evolving
operational performance measures improve sig- area. Section 3 outlines the systematic data col-
nificantly on pre-ERP levels four months after lection methodology in this study. In Section 4 we
implementation. McAfee proposes a longer time develop logistic regression models to classify
frame for such a study. Stratman and Roth pro- companies based on on-time and on/under-budget
pose an integrated conceptual model of ‘‘ERP measures and present our findings. Section 5
Competence’’ which they define as comprised of summarizes our observations and conclusions.
several organizational aptitudes including strategic
planning, executive commitment, project manag-
ement, IT skills and change management. They 2. Research issues and research framework
argue that a firmÕs ERP Competence must be used
effectively in order to truly harness the capabilities The research reported in this paper is part of a
of an ERP system for competitive advantage. Van long-term on-going project aimed at studying the
Everdigen et al. and Mabert et al. both use surveys state-of-the-art of ERP practice and implementa-
to systematically study a variety of issues. Van tions. This project has been carried out using a

304 V.A. Mabert et al. / European Journal of Operational Research 146 (2003) 302–314

two-phased approach. In Phase I, a case study during the planning and implementing stages.
methodology was employed to study ERP im- In the larger firms the team members were fully
plementations at 12 manufacturing firms using dedicated to the ERP project and were often co-
structured interviews of key managers, IT profes- located in a ‘‘War Room’’.
sionals, and users associated with each companyÕs • The implementation team spent extra time up
implementation. The size of these case study firms front to define in great detail exactly how the
ranged from $30 million in revenues to over $35 implementation would be carried out. This in-
billion. Five firms had revenues over $1 billion, cluded what modules and process options
three had revenues between $200 million to $1 would be implemented and how the senior man-
billion and four were in the $30 million to $200 agement priorities would be incorporated.
million range. These implementations involved • These companies laid out clear guidelines on
four different ERP vendor packages. In addition to performance measurements. These metrics were
the case studies, senior consultants at six consult- not just technical ones but also included busi-
ing firms specializing in ERP implementations ness operations.
were interviewed at length to get their perspectives • Modifications to the ERP system code were
on ERP (Mabert et al., 2000, 2001). Their expertise kept to a minimum.
covered five different packages. The primary ob- • Organizational change and training strategies
jective for conducting the case studies and the were developed in advance and were continu-
consultant interviews was to obtain reliable ally updated during the implementation.
and detailed information on the current status of • Implementations where several key modules
ERP practice and implementations. This research were implemented at the same time took a
methodology has been used successfully in a shorter time (implementing several modules at
number of studies (Orlikowski, 1992; Yin, 1993, the same time is usually referred to the Mini
1994; Robey and Sahay, 1996), and was employed Big-Bang approach. Implementing the entire
here to provide a foundation for our future efforts. system at the same time is called the Big-Bang
Several key findings emerged from the case approach) than phasing in modules a few at a
studies. First, while most implementation projects time (usually referred to as the phased-in ap-
are unique in many ways, there are still many proach).
underlying issues, activities and strategies that are • Key technology issues, such as data integrity
common to all of them, irrespective of the package and technology infrastructure, were addressed
implemented. Second, the findings from the case early.
studies strongly suggest that the overriding objec- • Only minor reengineering efforts were carried
tive of most companies is to complete the project out up front.
on-time and within the budgeted resources. Third, • The implementation plan and subsequent pro-
to meet on time and budget targets, ERP projects gress was communicated regularly to employ-
have to be planned very carefully and managed ees, suppliers and customers.
very efficiently. And fourth, the companies that
stayed on-time and on/under-budget for their ERP After analyzing the set of characteristics relative
implementation had many common characteris- to their nature and timing during the implemen-
tics. These include the following: tation process, we chose to group these character-
istics into three categories consisting of planning
• Senior executives were very involved through- effort, implementation decisions, and implemen-
out the project, from the outset to completion, tation management. Planning effort refers to all
and also established clear priorities. factors that have to be addressed in the planning
• A cross-functional ERP Steering Committee stages before the start of the project. These include
with executive leadership was established to such variables as executive support and involve-
oversee the project. The Steering Committee ment in the planning of the project, the makeup of
was empowered to make key decisions, both the implementation team, and addressing key


Table 1
Key planning and implementation management variables
Planning variables Implementation management variables
Development of a business case Strong executive involvement
Defined very clear desired outcomes Strong executive support
Defined performance metrics Communicated progress regularly
Strong executive sponsorship Benchmarked implementation progress
Strong executive involvement ERP committee able to make key decisions
An empowered ERP steering committee Communicated with personnel impacted
An ERP implementation team Created ‘‘Super-Users’’ and ‘‘Trouble-Shooters’’
Clear organizational change strategies Trained all users
Clear education and training strategies Kept suppliers/customers informed
Communicated ERP plan to the enterprise
Addressed data conversion and integrity
Technology infrastructure in place

technology issues. Implementation decisions refer study the individual and/or collective impact of the
to strategic options on how to conduct the im- planning variables, strategic decision variables,
plementation. These include such decisions as and implementation management variables on
whether to implement using the Big-Bang ap- ERP implementation costs and timelines.
proach or the phased-in approach, and the amount The proposed methodology of studying key
of software customization and reengineering to factors behind ERP implementations is very simi-
complete. The third critical area is implementation lar to the approach used in a variety of studies in
management itself, referring to all variables/actions Information Technology (IT) implementation re-
during the implementation. The key variables search. Some of these proposed factors are the
identified by the case studies in each of these three ones that have been found to be significant in
categories are listed in Tables 1 and 2. other IT implementations. These include top
While the case studies proved useful in under- management support, IT design, and appropriate
standing the general nature of implementations, it user training interaction and understanding (Fu-
was based on a small sample. To confirm our ini- erst and Cheney, 1983; Schultz, 1984; Sanders and
tial finding of the differences and commonality of Courtney, 1985; Kwon and Zmud, 1987). Other
ERP implementations, a survey instrument was studies of IT implementation also provide some
developed in order to obtain a broader perspective guidelines for the current study (Ginzberg, 1981;
of ERP practice and experiences relating to plan- McFarlan, 1981) but much of this deals with leg-
ning, customization, managing, costs, time-lines, acy systems. The technology closest to ERP sys-
performance, and success factors. More specifi- tems for manufacturing firms is MRP II. ERP
cally, the primary objective of the project is to systems have evolved from MRP and MRP II and
can be considered as extensions of MRP II with
enhanced and added functionalities. In the MRP
literature, several authors have studied imple-
Table 2
mentations (Schroeder et al., 1981; White et al.,
Key strategic decision variables
1982) using different key variables and regression
Implementation decision variables
based modeling.
Single ERP package versus multiple packages While IT and MRP II research sheds some light
Big-Bang or mini Big-Bang versus a phased-in approach
Number of modules implemented
on ERP implementations, there are significant
Order of implementation differences between both legacy systems and ERP,
Modifications to system and MRP II and ERP. For example, many legacy
Major reengineering upfront versus limited reengineering systems, like early MRP, were custom-designed
An accelerated implementation strategy requiring unique designer–user interactions. ERP


systems, on the other hand, are package systems The survey, cover letter and response envelope
that usually require organizational or process were mailed to 270 firms that were randomly
changes for implementation. Further, legacy sys- drawn from a list of approximately 2700 US
tems usually operated in functional silos whereas companies that have or are implementing ERP
ERP systems are integrative and require significant systems. This list has been developed from a va-
changes to processes across the entire organization riety of sources including ERP user-group com-
for successful implementation. The key difference panies, customer lists provided by ERP vendors,
between MRP II and ERP systems is that ERP consultant clients and through research of com-
includes functionalities, such as human resources pany Web sites. The contact names at these 270
planning, decision support applications, regula- companies were developed from APICS members,
tory control, quality, elements of supply chain ERP user-group participation lists, and a sub-
management and maintenance support that are scriber database list from InfoWorld, a technical
beyond the traditional focus of MRP II (Yusuf trade journal. The mailing was sent out in the first
and Little, 1998). week of October 2000. By mid-December 2000, 78
responses had been received, for an overall re-
sponse rate of 28.8%. Given the length and com-
3. Methodology prehensive nature of the survey, this response rate
was concluded to be reasonable. Respondents were
The survey questionnaire asked for information not asked to provide company-identifying infor-
on implementation in six key areas: The respon- mation to enhance sharing of company perfor-
dentÕs and the companyÕs characteristics, the ERP mance based information. Of the 78 responses, 77
planning process, implementation decisions, man- had already implemented ERP systems. One was
agement of the implementation process, timelines in very preliminary stages of implementation and,
and, budgets and costs. It was four pages long and consequently, had only minimal information.
had a total 26 questions. Most questions in the Another two had contradictory information and
survey required multiple responses. The responses were thrown out, leaving 75 usable responses.
were encoded using a mix of check boxes, open- The size of the responding companies ranged
ended answers, and a binary scale with Yes or No from $30 million to $32 billion in annual revenues.
responses. The case studies provided the guidance The number of employees ranged from 140 to
for the encoding scheme in terms of what type of 60,000. Eighty percent of the respondents were at
questions required what responses. For example, the manager or above level, with over 35% listing
the total cost was segmented into buckets because their job titles to be director or vice-president.
the interviews showed that respondents were more Approximately 25% had some form of ERP des-
comfortable with providing approximate figures ignation in their title (Director––SAP Applica-
instead of exact values. The planning and man- tions, Director of ERP Systems, etc.). A more
aging variables were encoded using a binary scale detailed breakdown of the statistics from the re-
(Yes or No) because either the companies used sponding companies is presented in Table 3.
these approaches or they did not. After the initial
development of the survey questionnaire, it was
sent to two ERP project leaders from our case 4. Results
study companies. The primary objective was to
test whether the instrument provided consistent The data show there are distinct differences in
and accurate information. Their responses were what the sample companies have done in terms of
checked against the information collected during their implementations. Just over 89% of the com-
the case studies. In addition, the questionnaire was panies have implemented a single ERP system, of
checked by two ERP consultants. Based on the which just over half have other systems attached to
information provided by these experts, the instru- their ERP system. The rest had implemented
ment was fine-tuned and finalized. modules from multiple ERP systems. A total of


Table 3
Basic data of responding companies
RespondentÕs position Percent (%) Annual revenues Percent (%) Number of employees Percent (%)
Executive/VP 10.0 <$500 million 28.5 <1000 22.8
Director 25.7 $501 million to $1 billion 10.2 1000–2500 14.3
Manager 44.3 $1 billion to $5 billion 32.8 2501–5000 21.4
Other 20.0 >$5 billion 28.5 >5000 41.5

65% of the sample were SAP companies. Given its 4.1. Planning variables
market share is approximately 33%, SAP is rep-
resented disproportionately in the sample. How- For the 12 planning variables, the respondents
ever, our case study and consultant analysis were asked to answer either a ÔYesÕ or a ÔNoÕ de-
indicates that the type of system implemented is pending on whether their company undertook this
not a factor in determining implementation out- planning task. The results are summarized in
comes, time lines, and success rates. The strategy Table 4. The interpretation of the sample statistics
used for the implementation is one of the most in this table is as follows: Of the companies that
important factors in determining the outcome of were on-time, late, on/under-budget or over-budget,
an ERP project. Fifty-two percent (52%) of the the percentage statistic shows the proportion of
sample companies chose to implement several key companies that completed that task. Thus, of all
modules at the same time. Only two of these 52% the on-time companies, 83% had developed a
could be truly classified as Big-Bang implementa- business case. Similarly, of all the companies that
tions. The rest were Mini Big-Bang. Twenty-nine were late, 78% had developed a business case. The
percent (29%) of the companies chose to do sig- results for both the timeline categories and
nificant reengineering up front. The rest either did the budget/cost categories show similar trends.
minimal reengineering or left it till after the system While the data suggest that both sets of categories
was installed. Approximately 55% chose to use an are very similar for such dimensions as ‘‘Defined
accelerated implementation methodology such as Performance Metrics’’ and ‘‘Had an ERP Imple-
ASAP. mentation Team/War Room’’, there are a number
To evaluate the impact of the planning, decision of very striking contrasts for tasks such ‘‘Had
and implementation management variables on the Strong Executive Involvement’’, ‘‘Developed Clear
costs and the timelines, all the companies are first Education and Training Strategies’’ and ‘‘Had
separated into two timeline categories, ‘‘on-time’’ Technology/Infrastructure in Place’’. These des-
and ‘‘late’’ based on their response to the question criptive comparisons are further supplemented sta-
whether their ERP implementation was completed tistically using contingency tables. There are several
on-time. Next, the companies were separated into significant differences, as indicted by the reported
two budget/cost categories, ‘‘on/under-budget’’ p-values. (An ÔNSÕ indicates a non-significant
and ‘‘over-budget’’, again based on their response value.) Overall, these results indicate that the
on whether their implementation was on/under- companies who are on-time and on/under-budget
budget or over-budget. Separating the companies do a better job in planning for their ERP project.
into these categories provides an easy but useful
way of determining if there is a relationship be- 4.2. Management variables
tween the planning, decision and implementation
variables and whether the project was on-time or The questions in the survey for the nine man-
late, and whether it is on/under-budget or over- agement variables were very similar to the plan-
budget. This analysis is presented below for each ning variables and required ÔYesÕ and ÔNoÕ
set of planning, decision and management vari- responses. The results for these variables are tab-
ables. ulated in Table 5. The trends here are very similar


Table 4
Impact of planning variables
Planning variables On-time Late (%) p-value On/under- Over-budget p-value
(%) budget (%) (%)
Developed a business case 83 78 NS 87 72 NS
Defined very clear desired outcomes 83 61 0.075 77 56 0.073
Defined performance metrics 56 51 NS 50 53 NS
Had strong executive sponsorship 94 80 NS 93 75 0.047
Had strong executive involvement 89 47 0.002 77 42 0.004
Had an empowered ERP steering committee 94 80 NS 100 69 0.001
Had an ERP implementation team/war room 100 94 NS 100 92 NS
Developed clear organizational change strategies 67 55 NS 67 50 NS
Developed clear education and training strategies 94 76 0.075 97 67 0.002
Communicated ERP plan to the enterprise 100 88 NS 93 92 NS
Addressed data conversion and integrity issues early 100 75 0.019 87 78 NS
Had technology/infrastructure in place 100 63 0.002 97 50 0.000

Table 5
Impact of implementation management variables
Management variables On-time Late p-value On/under- Over-budget p-value
(%) (%) budget (%) (%)
Had strong executive involvement 89 51 0.003 73 50 0.026
Had strong executive support 100 84 0.065 100 78 0.006
Communicated progress regularly across the company 100 84 0.065 93 86 NS
Benchmarked implementation progress against clear 83 67 NS 87 56 0.006
milestones or performance metrics
Allowed ERP committee to make key decisions 100 88 NS 100 83 0.019
Communicated regularly with all who would be im- 94 76 0.075 93 72 0.027
pacted
Created ‘‘Super-Users’’ who served as trouble-shooters 89 94 NS 93 92 NS
Trained all who would be using the system 94 86 NS 100 78 0.006
Kept suppliers/customers informed 78 63 0.070 87 63 0.001

between the on-time and the late companies, and the planning and the management variables, the
the on/under-budget and the over-budget compa- impact of these variables can be potentially much
nies, with the on-time and on/under-budget com- more significant. For example, our case studies
panies managing their implementations better. indicate that the amount of modifications to the
These descriptive comparisons are also supple- system can significantly increase both implemen-
mented statistically using contingency tables. tation times and costs. This shows up very clearly
There are several significant differences, as shown in these results. Only 11% of the on-time compa-
by the reported p-values. Overall, these results nies undertook major modifications versus 53% of
suggest that the companies that managed their the late companies. The results are very similar for
implementations more systematically were more the budget categories. One obvious result is for the
likely to complete their implementations on-time accelerated implementation strategy. Companies
and on/under-budget. that use these strategies are more likely to com-
plete their implementations on time.
4.3. Decision variables
4.4. Logistic regressions
The impact of the decision variables is sum-
marized in Table 6. While the differences for both While the above planning, decision and imple-
sets of categories are not as striking as the ones for mentation variables provide valuable insights into


Table 6
Impact of implementation decision variables
Implementation decision variables On-time Late p-value On/under-budget Over-budget p-value
Implemented single ERP package 61% 41% NS 53% 39% NS
Used a mini Big-Bang implementation 65% 47% NS 59% 44% NS
Average number of modules implemented 8.24 8.51 NS 8.66 8.31 NS
Made major modifications to system 11% 53% 0.002 23% 56% 0.008
Undertook limited reengineering upfront 67% 51% NS 63% 50% NS
Used accelerated implementation strategy 89% 47% 0.005 53% 56% NS

completion of implementations on-time or on/ The coefficients of the logistic regression model
under-budget, a more systematic statistical analysis are determined as maximum likelihood estima-
is necessary to determine which of these variables tors, i.e., by maximizing the likelihood functions
are individually or collectively significant. Also, it (Fienberg, 1983). A negative sign for a coefficient
will be useful to explore if some composite vari- means that the variable is more likely to increase
ables created using combinations of variables are the probability of achieving an on-time or an on/
better predictors of success. For example, would under-budget implementation. In regression anal-
a composite variable representing all planning ysis, the modeler has the flexibility to specify the
factors be better at predicting on-time or on/un- functional form of the independent variables, with
der-budget implementations than the individual several open-ended options available. One strategy
planning variables. The composite variables would is to use all available independent variables. An-
reflect a collective set of things that need to be other is to collapse some of the variables into
done rather just a few individual ones. A number ‘‘composites’’ reflecting total effort. For this anal-
of statistical tools are available for such analysis ysis, we use both these options. For the latter, two
including discriminate analysis and regression. Dis- ‘‘composite’’ independent variables are created.
criminate analysis would help a researcher classify The first is a composite variable for the planning
variables whereas regression allows for more rig- variables. It is set up by adding all the ‘‘1’’ re-
orous analysis by providing an overall predictive sponses to the 12 planning variables. The objective
model and a significance level for the chosen here is to capture the overall amount of the plan-
variables. Since our response variables are di- ning effort. Thus, if a company used all the 12
chotomous (on-time versus late, and on/under- planning variables, its composite score would be
budget versus over-budget), we chose to use the 12. The second composite variable is for the im-
binary logit model of logistic regression analysis. plementation management variables. It is also set
Green et al. (1977) first proposed the applica- up by adding the ‘‘1’’ responses to the nine im-
tion of logit analysis. Since then numerous re- plementation management variables to capture the
searchers have used it to analyze categorical data overall management ‘‘effort’’. A similar composite
(Robinson and Satterfield, 1998; White et al., variable is not feasible for the decision variables
1999). To use a logistic regression, the response since the impact of each decision variable is indi-
variable needs to be coded as 0 or 1. For the study vidual in nature and not collective.
reported here, two sets of logistic regressions are
executed. The first uses completion of an imple- 4.5. Logistic regressions for on-time versus late
mentation on-time as the response variable. Com- implementations
pletion times are coded as 0 if the implementation
was completed on-time, and 1 if it was late. The A stepwise tool was used to develop the logistic
second logit regression uses completion of the regression models. For each response variable, two
implementation on/under-budget (coded 0) or logistic regressions are shown, one using at least
over-budget (coded 1) as the response variable. one composite variable and the other using all the


Table 7
Significant variables for the on-time vs late logistic regression (with at least one composite variable)a (Panel A) and classification table
for on-time vs late (with at least one composite variable) (Panel B)
Panel A
Variables Coefficient Standard error p-value
Composite of planning variables )1.820 0.682 0.008
Modifications to the system 7.590 2.751 0.006
Accelerated methodology )3.492 1.748 0.046
Strong executive involvement during implementation )2.292 1.270 0.071
Training all users 8.319 3.431 0.015
Keeping suppliers and customers informed )3.117 1.793 0.082
Constant 15.651 6.131 0.011
Panel B
Predicted on-time Predicted late Total Percentage correct (%)
Actual on-time 12 4 16 75.0
Actual late 1 46 47 97.9
Overall percentage 92.1
a
À2 log(likelihood) model: Marginal ¼ 74:706, Full ¼ 29:815, Diff ¼ 44:891, Sig ¼ 0:000.

individual variables. The results obtained from the The results for an on-time logit model with all
fit of these logistic regression models for on-time singular variables are shown in Tables 8 (Panel A)
implementations using the planning composite with the classification table in Table 8 (Panel B).
variable are summarized in Table 7 (Panel A) with The stepwise analysis only brings three variables
the resulting classification analysis in Table 7 into the equation, two planning and one decision.
(Panel B). The analysis was performed on 16 on- Both the planning variables, strong executive
time and 47 late implementations. Of the 63 cases involvement and defining very clear desire objec-
used to fit the model, 75% of the on-time compa- tives, contribute positively to an on-time imple-
nies were classified correctly and 97.9% of the late mentation whereas modifications result in delays.
companies were classified correctly, for an overall This model is not as powerful as the one with the
accuracy of 92.1%. The significant variables in the planning composite variable.
logit model are two decision variables, the com-
posite planning variable and three implementation 4.6. Logistic regressions for on/under-budget versus
management variables. The model predicts that over-budget implementations
extensive planning in advance of the implementa-
tion, an accelerated implementation strategy, For the budget logit model, the response vari-
strong executive involvement during implementa- able is set equal to 0 for an on/under-budget im-
tion, and keeping customers and suppliers in- plementation and equal to 1 for an over-budget
formed all contribute positively to an on-time implementation. The analysis for on/under-budget
implementation. On the other hand, the model versus over-budget was performed on 29 on/under-
indicates modifications to the systems and training budget and 33 over-budget companies for a total
users are likely to result in delays. While training of 62 cases. The results for this logistic regression
can add to the implementation time, it must be with composite variables are presented in Table 9
pointed out that this function is critical to the (Panels A and B). The significant variables in this
eventual success of the project. Like several of our logit model include the planning composite vari-
case study companies, it is quite possible that our able, one decision, and two implementation man-
sample companies did not estimate the extent of agement variables. The model indicates that the
the time required for training. planning composite variable contributes positively


Table 8
Significant variables for the on-time vs late logistic regression (without composite variables)a (Panel A) and classification table for on-
time vs late (without composite variables) (Panel B)
Panel A
Strong executive involvement during planning )3.520 1.177 0.003
Defining very clear desired objectives )1.766 0.931 0.058
Constant 3.953 1.294 0.002
Panel B
Predicted on-time Predicted late Total Percentage correct (%)
Actual on-time 10 6 16 62.5
Actual late 3 44 47 93.6
a

Table 9
Significant variables for on/under-budget vs over-budget logistic regression (with at least one composite variable)a (Panel A) and
classification table for on/under-budget vs over-budget implementations (with at least one composite variable) (Panel B)
Panel A
Composite of planning variables )0.963 0.269 0.000
Communicated progress regularly 4.516 1.790 0.012
Kept suppliers and customers informed )2.385 1.005 0.018
Constant 4.951 2.048 0.016

Panel B
Predicted on/under-budget Predicted over-budget Total Percentage correct (%)
Actual on/under-budget 24 5 29 82.8
Actual over-budget 6 27 33 81.8
a

to an on/under-budget implementation. As one that 96.6% of the on/under-budget companies are
would expect, modifications impact budgets ad- classified correctly and 84.8% of the over-budget
versely. Just like the model without the composite companies are classified correctly, for an overall
variables, this regression shows that keeping sup- percentage of 90.3%. The significant variables in
pliers and customers informed has a favorable the logit model include three planning, one deci-
impact on implementations that are completed on sion, and two implementation management vari-
or under-budget whereas communicating progress ables. The model indicates that the three planning
regularly to the entire company impacts budgets variables all contribute to an on/under-budget
adversely. It is also interesting to note that the implementation. Modifications impact budgets
implementation management composite variable adversely. Of the two management variables,
was not significant in any of our models. keeping suppliers and customers informed has a
The results for this logistic regression without favorable impact on budgets whereas communi-
composite variables are presented in Table 10 cating progress regularly to the entire company
(Panels A and B). The classification table shows impacts budgets adversely. Since our case studies


Table 10
Significant variables for on/under-budget vs over-budget logistic regression (without composite variables)a (Panel A) and classification
table for on/under-budget vs over-budget implementations (without composite variables) (Panel B)
Panel A
Strong executive involvement during planning )4.586 1.986 0.021
Clear education and training strategies )2.815 1.582 0.075
Technology and infrastructure in place )6.278 2.707 0.020
Communicated progress regularly 6.043 2.563 0.018
Kept suppliers and customers informed )4.457 1.753 0.011
Constant 4.668 2.555 0.068
Panel B
Predicted on/under- Predicted Total Percentage correct
budget over-budget (%)
Actual on/under-budget 28 1 29 96.6
Actual over-budget 5 28 33 84.8
a

indicated that communications is a critical func- results from this more comprehensive investigation
tion in the success of ERP implementations, this confirm analytically what we had learned through
counter-intuitive outcome is discussed later. the case studies.
First, all of the companies in our case studies
stressed that upfront planning is a key element for
5. Discussion and conclusions a successful implementation. This is confirmed by
the composite planning variable being highly sig-
Our case studies suggest that because of the nificant in our models. At a more detailed level,
investment required for an ERP project, both in planning of education and training programs, and
terms of the resources and the resulting organiza- having a technology infrastructure are two of the
tional changes, companies are very sensitive about significant individual planning variables common
implementation times and budgets, going to great to successful implementation. Several companies
lengths to set realistic timelines and budgets. One in our case studies indicated that planning educa-
indicator of this phenomenon is the case study tional and training programs had given them a
companies that started their implementations la- much better understanding of the magnitude of
ter. Such companies tended to have shorter com- costs involved in this process.
pletion times and smaller budgets, reflecting that Second, the case study companies also empha-
implementations have become more ‘‘efficient’’ sized keeping modifications to the source code to a
over time because of the learning curve effect. minimum. Because the integrative design of ERP
While the authors recognize that both the timelines systems increases the complexity involved in
and the budgets are set by the companies and self- source code modifications, most companies sig-
reported in the survey, there is no reason to believe nificantly underestimate the effort required for
they are not representative of actual experience. modifications. Modifications not only lead to in-
The logistic regression equations provide a valu- creased costs and implementation times, they also
able mechanism for identifying significant input make future upgrades of the system difficult to
variables in classifying whether a particular ERP implement. Every model in this study determined
implementation is likely to be on-time and within modifications to be a highly significant variable
budget based on the survey data. Many of the with adverse impact.


Third, the case study companies emphasized the to find answers to complex processes often creates
importance of the implementation management new questions. The authors note some counter-
effort itself. Surprisingly, those variables, individ- intuitive outcomes that require future exploration.
ually and collectively, are not as significant in
predicting on-time and on/under-budget imple- References
mentations as anticipated. This could be due to the
fact that since an ERP implementation is a key Davenport, T., 1998. Putting the enterprise into the enterprise
project, companies who put in significant up front system. Harvard Business Review 76 (4), 121–131.
planning continued to do a good job with the Deutsch, C., 1998. Software that can make a grown company
cry. The New York Times CXLVIII (51), 1, 13.
management of the implementation process. For
Diederich T., 1998. Bankrupt firm blames SAP for failure.
example, the correlation between the planning ComputerWorld August 28.
composite variable and management composite Fienberg, S.E., 1983. The Analysis of Cross-Classified Cate-
variable is 0.72 (p-value < 0:01). The data does gorical Data. MIT Press, Cambridge, MA.
suggest that keeping suppliers and customers in- Fuerst, W.L., Cheney, P.H., 1983. Factors affecting the
perceived utilization of computer-based decision support
formed was one significant implementation man-
systems in the oil industry. Decision Sciences 13 (4), 554–
agement variable, which was identical to our 569.
earlier study. For three of the case study compa- Ginzberg, M.J., 1981. Early diagnosis of MIS implementation
nies involving their suppliers and customers in the failure: Promising results and unanswered questions. Man-
implementation process led to two key outcomes. agement Science 27 (4), 459–478.
Green, P.E., Carmone, F.J., Wachpress, D.P., 1977. On the
One, they were able to design the inter-firm inter-
analysis of quantitative data in marketing research. Journal
action processes better, leading to fewer modifi- of Marketing Science 14, 52–59.
cations to the system. And two, they were able to Kirkpatrick, D., 1998. The e-ware war: Competition comes to
‘‘stress-test’’ the likely volume and types of trans- enterprise software. Fortune 138 (11), 103–112.
actions to be handled. Kwon, T.H., Zmud, R.W., 1987. Unifying the fragmented models
of information systems implementation, Critical Issues in
Finally, we noted a counter-intuitive outcome
Information Systems Research. John Wiley, New York.
as identified by the positive sign on the regression Mabert, V.M., Soni, A., Venkataramanan, M.A., 2000. Enter-
coefficient for the communication of progress, prise resource planning survey of US manufacturing firms.
which suggests that more communication tends to Production and Inventory Management Journal 41 (20), 52–
increase the likelihood of cost overruns. One pos- 58.
Mabert, V.M., Soni, A., Venkataramanan, M.A., 2001. Enter-
sible explanation is that the communications might
prise resource planning: Common myths versus evolving
cause reexamination of processes as well as ne- reality. Business Horizons 44 (3), 69–76.
cessitating more customization. Unfortunately, the McAfee, A., 1999. The impact of enterprise information
collected data do not allow for a more systematic technology adoption on operational performance: An em-
evaluation of this issue. This does suggest that pirical investigation, Harvard Business School, Cambridge
Working Paper.
further research is needed in this area. However,
McFarlan, F.W., 1981. Portfolio approach to information
all of our case study companies felt this is a criti- systems. Harvard Business Review 59, 142–159.
cally important step for user acceptability of the Nelson, E., Ramstad, E., 1999. HersheyÕs biggest Dud has
system as well as its productive use. turned out to be new computer system. The Wall Street
This study provides valuable insights towards Journal CIV (85), A1–A6.
Orlikowski, W.J., 1992. The duality of technology: Rethinking
understanding ERP implementations and impor-
the concept of technology in organizations. Organization
tant factors influencing success. In particular, it Science 3 (3), 398–427.
analytically verified the importance of planning, Piturro, M., 1999. How midsize companies are buying ERP.
execution and strategy on implementation time Journal of Accountancy 188 (3), 41–48.
and budgets. While the findings have some com- Robey, D., Sahay, S., 1996. Transforming work through
information technology: A comparative case study of
mon elements with other IT implementation
geographic information systems in county government.
studies, there are many that are unique to ERP Information Systems Research 7 (1), 93–110.
implementations because of the integrative char- Robinson, E.P., Satterfield, R.K., 1998. Designing distribu-
acteristics of ERP systems. However, the research tion systems to support vendor strategies in supply


chain management. Decision Sciences 29 (3), 685– White, E.M., Anderson, J.C., Schroeder, R.G., Tupy,
706. S.E., 1982. A study of the MRP implementation pro-
Sanders, G.L., Courtney, J.F., 1985. A field study of organi- cess. Journal of Operations Management 2 (3), 145–
zational factors influencing DSS success. MIS Quarterly 9 153.
(1), 77–93. White, R.E., Pearson, J.N., Wilson, J.R., 1999. JIT manu-
Schroeder, R.G., Anderson, J.C., Tupy, S.E., White, E.M., facturing: A survey of implementations in small and
1981. A study of MRP benefits and costs. Journal of large US manufacturers. Management Science 45 (1), 1–
Operations Management 2 (1), 1–9. 15.
Schultz, R.L., 1984. The implementation of forecasting models. Yin, R.K., 1993. Applications of Case Study Research. Sage
Journal of Forecasting 3, 43–55. Publications, Newbury Park, CA.
Stratman, J.K., Roth, A.V., 1999. Enterprise resource planning Yin, R.K., 1994. Case Study Research: Design and Methods.
(ERP) competence: A model and pre-test, design-stage scale Sage Publications, Thousand Oaks, CA.
development, University of North Carolina Chapel Hill Yusuf, Y., Little, D., 1998. An empirical investigation of enter-
Working Paper. prise-wide integration of MRPII. International Journal
Trunk, C., 1999. Building bridges between WMS & ERP. of Operations and Production Management 18 (1), 66–
Journal of Transportation and Distribution 40 (2), 6–8. 86.
Van Everdingen, Y., Van Hillegerberg, J., Waarts, E., 2000. Zuckerman, A., 1999. ERP: Pathway to the future or yester-
ERP adoption by European midsize companies. Communi- dayÕs buzz? Journal of Transportation and Distribution 40
cations of the ACM 43 (4), 27–31. (8), 37–43.

8. Erp Managing The Implementation Process

More Related Content

What's hot (20)

Viewers also liked (11)

Similar to 8. Erp Managing The Implementation Process (20)

More from Donovan Mulder (10)

Recently uploaded (20)

8. Erp Managing The Implementation Process