You are on page 1of 18

Abstract

Developing an This paper proposes the use of the Activity Based

Activity-based Costing (ABC) approach to software estimation. Like


other more traditional approaches to software
estimation, ABC provides man-day estimates. In
Costing Approach addition, it also provides detailed costing information
that is useful for management control and decision
for System making. The paper shows how the ABC approach can
be applied to software estimation by building an ABC

Development and model using data from twenty-two projects in a


financial services firm. The model is then used for
estimation, and comparisons between estimated and
Implementation actual man-day are computed (variance analysis). The
data generated by the model and from the variance
analysis are useful for management control and
decision making in areas such as resource allocation,
outsourcing of specific development activities, and
Ginny Ooi learning from adoption of new development tools and
Nanyang Technological University practices.

ACM Categories: D2.8, D2.9, K6.1, K.6.3


Christina Soh
Nanyang Technological University Keywords: IS Project Planning, Effort Estimation,
Time and Cost Estimation, Activity-based Costing,
Software Process Measurement, Organizational
Learning.

Introduction
Software development time and cost estimations are
important to management in deciding whether to
approve a software development project. They are
also used in departmental budgeting, project planning,
resource allocation, and outsourcing decisions. Both
understated and overstated costs have negative
impacts on IS management and organization
(Benjamin et al., 1984; Dugger, 1996; Lederer &
Prasad, 1995).
Unfortunately, studies have shown that software
development projects often run 100% to 200% over
budget (Maglitta, 1991; Rubin, 1991), and many
projects were abandoned because of severe cost
overruns and schedule slippages (Keil et al., 1995).
Budget overruns can arise from inaccurate estimation
of time and/or inaccurate allocation of resource cost.
Existing software estimation models focus on
estimating development time. Many traditional
estimation approaches estimate total project time, and
assume that accurate time estimation automatically
leads to accurate cost estimation. This assumption
Acknowledgement may not hold as the time spent in different
An earlier version of this paper appeared in the development and implementation activities differs
Research-in-Progress track of the Proceedings of the across projects. Different activities have different
Nineteenth Annual International Conference on costs, and some activities cost more than others. For
Information Systems, (Helsinki, Finland, December 13- example, one man-day of project management is more
16, 1998), pp. 341-345. costly than one man-day of programming because

54 The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3)
project management is performed by more Activity-Based Costing
experienced staff. In traditional approaches to costing
software development, two projects that require the Activity-based Costing (ABC) is a cost allocation
same number of man-days will have the same cost model pioneered by Harvard’s Cooper and Kaplan
estimate, even though one project may be more (1988), in the field of management accounting. ABC
complex and require a greater proportion of time to be has been successfully applied to manufacturing and
spent in more expensive activities such as project service industries (Helmi & Hindi, 1996; Kroll, 1996;
management. Reimann & Kaplan, 1990) for improving tactical and
Ideally, the estimation approach should also provide strategic decision-making and for enhancing corporate
useful information to management for three key cost control and customer profitability (Bradway &
management tasks: project approval, implementation Ross, 2000; Mabberley, 1998). ABC provides
monitoring, and post-implementation evaluation. To management with information to understand the use of
accomplish this, not only is information needed for scarce organizational resources in various business
total project time and costs, but also for time and cost activities. Management can then focus on areas of
by activity. This will enable IS and user managers to high cost, identify the factors that influence these
better manage and schedule resources for the project, costs, benchmark performance, and quantify
monitor the progress by activities, and evaluate and improvements in the time and cost of the activities
analyze the project by activities upon completion. performed. This contributes to the better management
Second, in order to provide relevant management of organizational resources in relation to product
information, the estimation model needs to be costing and customer profitability.
contextualized to the organization’s development
practices. The cost drivers used in the models should In the ABC approach, resources are first traced to
reflect the actual development practices of the project activities, and activity costs are then traced to
teams, and the cost driver rates should reflect their products/services, based on their consumption of the
skills and productivity levels, the different mix of activities. This principle is fundamentally different from
human resources in different activities, and the the traditional costing system that assumes that
compensation structure of the organization. Third, the products/services consume resources directly. The
cost drivers and cost driver rates must be easily generic ABC model is outlined in Figure 1.
understandable by both IS and user management so
as to facilitate negotiation between them regarding the The ABC model first groups the organization
scope of the project and the associated costs. resources into various resource pools (see Figure 1),
such as salaries, rental expense, license fees, and
Activity-based Costing (ABC) is an integrated depreciation. Various tasks performed in the
approach to estimating both project time and cost by
organization are then grouped into major and
providing a simple and contextualized basis for
functionally homogeneous activities, such as research
estimating costs by activity. In this paper, we compare
and development (R&D), receiving, and delivery. Each
and contrast the ABC approach with established
activity consumes different amounts of one or more
approaches to software estimation, develop an ABC resource pools. For example, R&D uses 5% of rental
model for software development and implementation, and 40% of salaries. The costs of performing the
and test it using data collected from twenty-two
activities are determined by allocating the costs from
projects in a large bank. Our findings indicate that the
the resource pools to each activity according to the
ABC approach provides reasonable estimation
percentage of the resource pool consumed.
compared to the actual effort and costs reported. The
approach also provides useful costing information that
Once the activity costs are determined, they can be
facilitates management decision making and
traced to cost objects (such as product) using cost
organizational learning.
drivers. Cost drivers measure the frequency and
The remainder of this paper is organized as follows: intensity of the demands placed on activities by cost
the ABC approach is first presented, followed by a objects. For example, the cost driver for the delivery
comparison with more established approaches to activity is the number of shipments. The organization
software estimation. The research methodology is then computes the total number of shipments in a given
described and the results of the study presented, time period, and traces the delivery activity costs to all
followed by a discussion of the management the products based on the number of shipments
information provided by the model and the potential for consumed by each product. The product with the
organizational learning through variance analysis. We largest number of shipments would bear the highest
conclude with implications for research and practice. cost for delivery activity.

The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3) 55
Total Dept
Resources
Total Costs

Resource A Resource B Resource C


Resource Pools Cost Pool Cost Pool Cost Pool

Resource Drivers
- % of time spent on
each activity

Activity 1 Activity 2 Activity 3 Activity 4


Activities

Cost Driver 1 Cost Driver 2 Cost Driver 3 Cost Driver 4


Cost Drivers

Project X Project Y
Cost Objects

Figure 1. Generic ABC Model

A comparison of the ABC approach with more models have improved in terms of providing early and
established approaches to software estimation accurate estimation of development effort.
provides an understanding of the relative advantages
FPA and COCOMO have traditionally been applied to
and limitations.
mainframe-based projects, using COBOL and 4GL
(Boehm, 1981; Dolado, 1997; Kemerer, 1987, 1993).
Major Approaches to Software Estimation
ANN and CBR can be used for different systems
There are many approaches to software project contexts as the models are capable of learning.
estimation. Widely cited estimation approaches include However, research studies of these techniques were in
three versions of Constructive Cost Models the context of projects that used COBOL (Finnie et al.
(COCOMO) (Boehm, 1981; 1984), Function Point 1997; Samson et al. 1997). ABC has largely been
Analysis (FPA) (Albrecht, 1979; Albrecht & Gaffney, used in manufacturing and service industries for
1983), Artificial Neural Networks (ANN) (Albus, 1981; estimating the cost of products and services.
Finnie et al., 1997), and Case-based Reasoning (CBR)
(Finnie et al., 1997; Mukhopadhyay et al., 1992; In the following section, the ABC approach is
Watson & Marir, 1994). More recent research has compared with the other established estimation
been conducted to advance these estimation models approaches along the following key dimensions: (1)
(for example, Ebrahimi, 1999; Kemerer & Porter, 1992; model inputs, (2) estimation process, (3) model
Kemerer, 1993; Samson et al., 1997), and these outputs, (4) relative accuracy, and (5) set up costs.

56 The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3)
Model Inputs. All the models require the estimation of and CBR are able to take into account individual
many project parameters as inputs to the estimation organizational practices in the estimation process, the
process. The inputs for FPA include five components use of standard function points as inputs limits their
of the proposed system, weighted by the levels of flexibility. ABC uses customized estimation rates to
complexity, and fourteen general application reflect the resource mix, cost, and development
characteristics. Simple COCOMO requires practices of the organization.
specification of the development modes and estimation
of delivered source instructions (DSI) as inputs. Model Outputs. All the established approaches focus
Intermediate COCOMO requires an additional fifteen on estimation of development time, usually in man-days.
effort adjustment factors (EAF). ANN and CBR both (Dolado, 1997; Finnie et al., 1997; Kemerer, 1993;
build on FPA as they use function points and general Mukhopadhyay et al., 1992; Samson et al., 1997). The
application characteristics as input parameters, as well main output for the ABC approach includes
as a database of prior projects. development time, resource mix required, and
estimated cost for each activity.
Prior research shows that inter-rater variance in the
estimation of inputs for FPA is about 30% (Low & FPA, basic and intermediate COCOMO, ANN and CBR
Jeffery, 1990; Rudolph, 1983). This is because input all provide a single estimate of total project time.
definition and estimation guidelines are not strictly Detailed COCOMO divides the project into six
standardized and are therefore subject to project predetermined phases, while ABC estimates time and
managers’ interpretation and judgement (Kemerer & cost for all the development activities of interest to the
Porter, 1992). These difficulties arise because FPA management of the organization.
attempts to provide a standardized model across many
organizations with different development practices and Relative Accuracy. Mean Absolute Relative Error
levels of resource capability. Since ANN and CBR use (MARE) has often been used in prior research to
function points as inputs, they are also subject to the determine the accuracy of the software estimation
problem of inter-rater variance. The main inputs for models. The formula for MARE is:
ABC are activity driver counts. These activity drivers
are customized to the organization’s context, and are ∑ {| Actual Time – Estimated Time|}
{Actual Time}/ Number of Projects
to be supported by customized examples to aid
organizational managers in their estimation of driver Finnie et al. (1997) reported that FPA has a MARE of
counts. 62.3%, while the MARE for ANN and CBR are 35.2%
Estimation Process. FPA has a standard formula that and 36.2%, respectively. Kemerer (1987) compared the
uses the inputs on five components of the proposed estimation models and reported MARE for FPA as
system and fourteen general system characteristics to 102.74%, Basic COCOMO as 610.09%, and Detailed
compute the number of function points. The function COCOMO of 607.85%, and concluded that that FPA is
points in turn determine the lines of code and the man- more accurate in estimating software development
days required for the project. Simple COCOMO uses a effort. While these estimation models have been studied
standard formula to arrive at delivered source and compared, ABC has not been applied to software
instructions, which are then translated into the man- development estimation. This study will however
month requirements. Intermediate COCOMO refines the compute the MARE for the sample of projects
estimated man-month estimate by considering an examined.
additional fifteen adjustment factors. Detailed COCOMO
Set-up Costs. Since FPA and COCOMO use standard
uses the adjustment factors, and computes the
models (inputs, rates and formulae), the set-up costs for
delivered source instructions for six specific phases in
the adopting organization is lower. The main cost to the
the development process, to arrive at total effort
organization is in training their managers to estimate the
required. ANN and CBR use function points as input,
inputs required by FPA. ANN and CBR need to make
and make use of previous cases to estimate effort
use of historical cases/ projects in the organization, in
required for the current project. ABC uses the
order to predict the software estimation effort. Set up
organizational resource mix and different activity drivers
time and costs are therefore required to set up the
to estimate the effort and cost required in each phase of
database of previous projects and to run the ANN and
the development process.
CBR models to generate organization specific
FPA and COCOMO use standard estimation rates in estimation rates. Managers will also need to be trained
estimating time. The use of standard estimation rates to provide inputs for the estimation of Function Points
may result in estimates that are unrealistic if an since these serve as inputs to both CBR and ANN. ABC
organization’s development practices and resource has significant set up costs as analysis is required to
capability (experience and ability of IT staff for example) identify organizational resource pools, development
are very different from the model’s norms. While ANN activities, and activity cost drivers.

The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3) 57
Comparing and contrasting these estimation models that could potentially be examined. The organization
provides insights into the strengths and weaknesses of also offered a number of other important advantages: it
each estimation model. From a management had a mix of in-house development and package
perspective, the common characteristics of existing implementation projects, and a mix of traditional and
estimation approaches result in some shortcomings, i.e. newer technologies such as host-based and client-
cost implications of resource mix are not considered, server systems. Finally, it had reasonably detailed
activity costs are not provided, input parameters are records of time spent by project team members. This
complex and difficult to understand, and estimation was necessary for the development of the ABC model.
rates are not contextualized to the organization. The
ABC approach potentially addresses these We selected twenty-two software development and
shortcomings. implementation projects for the study. The project
selection was limited by the requirement that the
First, ABC focuses on each activity in the software
projects in progress be completed within the duration
development process. This provides management with
of the study, so that variance analysis could be
more detailed data to monitor and evaluate performance
performed. Other selection criteria included project
by activity, and not merely at the overall project level.
size, technology platform, and adequate activity
Although COCOMO attempts to address the problem by
information in project work-plans. The twenty-two
having six activities in its complex model, the activities
projects consisted of five in-house enhancements, nine
may not be applicable to every organization. ABC model
in-house developments and eight package projects.
allows full customization to the development and
Five were host-based systems, and seventeen were
implementation practices of each organization. Second,
client-server. Projects ranged from 49 man-days to
ABC is more than an effort estimation tool in that it takes
2640 man-days, with an average man-day of 651.
into account the different types of resources and
associated resource costs consumed by each activity.
While a larger sample size is desirable, it is often
Instead of using a fixed charge-out rate per man-day,
difficult to obtain a large number of projects within the
ABC allows management to better understand the
time frame of a single study. Software development
actual cost of carrying out each development and
implementation activity. In order to provide this projects generally take a long time to complete and are
information, organizational specific data on organization not numerous. Also, it is difficult and often not a good
research strategy to accumulate projects across
structure and labor costs are incorporated into the
organizations, because one must then control for
model. Third, ABC is likely to be more easily understood
various organization specific effects and the
by top management and user departments because it is
consistency of measures. The sample size is
a widely accepted approach in cost accounting. Finally,
ABC supports organizational learning. It provides comparable to Albrecht and Gaffney’s (1983) and
analysis between estimated and actual time and cost Kemerer’s (1987) studies, which used twenty-four and
fifteen projects, respectively, for software estimation
used for each activity and provides a feedback loop for
model validation.
project managers.
The ABC approach does require more upfront set-up Data Collection
costs, in order to build a model that is customized to
the organization. It also does not have a history of use Various well-established ABC development and
in systems development, and therefore does not have implementation methodologies from American
a track record in terms of prior research in estimation Management Association, Institute of Management
accuracy. This study seeks to demonstrate the model Accountants, and ABC Technologies, Inc. (Miller,
building process, the benefits and limitations of the 1996) were synthesized and adapted to provide an
ABC approach, and to provide some initial indication of approach suitable for software development and
estimation accuracy, within the limits of the study. implementation. In developing and applying the ABC
approach, guidelines for developing successful
estimation models in software engineering were also
Methodology considered. These included transparency of the model,
Site and Sample Selection usefulness of the model, developer participation,
feedback, automated data collection, user training,
The research site in this study is a leading financial dedicated implementation team, and a goal-oriented
institution. The organization was selected for the study approach (DeMarco, 1982; Hall & Fenton, 1997;
because it had a large number of ongoing IT projects Grandy & Caswell, 1987; Pfleeger, 1993).

58 The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3)
Stages Main Activities
Study Current - Review general ledger and human resource reports for organization structure,
System staff mix, salary structures
Development - Review software development manuals and handbook for prescribed software
Process development and implementation activities
- Interviews with two process quality staff and two project managers to confirm
activities
Develop ABC - Gather data from two development projects to develop ABC prototype
Prototype - Hold focus group sessions with management, project managers, quality
assurance managers, and accountants to clarify any queries regarding ABC
prototypes, such as resource pools, activities, and cost drivers.
Refine ABC - Interviews with twenty project managers from twenty projects to refine
Prototype activities and cost drivers
- Gather driver counts for the twenty projects
- Analysis of project time sheets for the twenty projects
- Gather qualitative information about the individual projects to resolve any
discrepancies

Table 1. Process in Developing ABC Estimation Model

The activities involved in developing an ABC model for time to cost drivers. The cost drivers had previously
software development are summarized in Table 1, been identified by project managers during interviews,
under three broad stages: (1) Study current system and driver counts had been collected for all twenty-two
development process, (2) Develop ABC prototype, and projects. Activity time was obtained from project time
(3) Refine ABC prototype. sheets. The sample of twenty two projects was split
into two comparable sets based on the project types
One major advantage of the approach used in this and size, in order to perform a split-half regression and
study is the reliance on both quantitative and estimation analysis. The first set consisted of twelve
qualitative data throughout the data collection process. projects that were used to build the ABC model. The
Quantitative data from time sheets, financial accounts, model was then applied to the second data set that
and driver counts were collected to build and validate consisted of ten projects. Ten regressions were
the ABC model. Qualitative data from interviews and performed for the ten development activities identified,
focus group discussions on activities, drivers, and with the actual activity time as dependent variables
explanations of variances were also used to provide a and activity drivers as independent variables. The beta
better understanding of the software development and coefficient of the regression was the driver rate for
implementation process in the organization. It also each activity.
helped to clarify inconsistencies in quantitative findings
In order to mitigate the problem of small sample size,
and to explain variances. Triangulation of data from
only one independent variable was used, as far as
these different sources strengthens the findings and
possible, for each regression. Tests were run to check
increases the robustness of the results (Kaplan &
for violation of the assumptions required for regression
Duchon, 1988; Yin, 1984), and software engineering
analysis. These assumptions include linearity,
issues, in particular, have been said to be best
normality, independence of independent variables, and
investigated using a combination of qualitative and
non-multicollinearity. Other than linearity, the other
quantitative methods (Athey, 1998; Seaman, 1999).
assumptions appeared to be valid for the sample.
Analysis Prior studies suggest that there is a non-linear
relationship between function points and software
There are four parts to the analysis, namely cost
maintenance effort (Banker and Slaughter, 1997;
analysis, driver analysis, estimation using ABC model,
Banker et al., 1998). We therefore ran the box-cox
and variance analysis. Cost Analysis required the
transformation on each activity to determine the
identification of resource pool costs through the
linearity of the regressions (Kmenta, 1986). The
analysis of general ledger accounts. These costs were
general model used was:
then traced to activities by analyzing the time sheets
kept by project team members. (λ)
Yi = β0 + β1 X1i + …..+ βk Xk1 + εI
Driver Analysis employed regression to relate activity

The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3) 59
(λ)
Where Yi = Results: ABC Model
(λ)
for Software Development
Yi -1 for λ ≠ 0, and
λ The results of the analysis are presented in four
sections: (1) Cost Analysis, which includes the analysis
Loge for λ = 0 of resource pools, determination of total cost per
resource, identification of activities, total cost per
For the test of linear model the test statistic is: activity, and activity costs per day; (2) Cost Driver
Analysis, which reports the identification of cost
2[L (λ(hat)) - L (λ=1)] drivers, determination of driver rates, and the
estimation of man-days required based on the number
2 of drivers; (3) Estimation using the ABC model, where
This can be compared with the χ1 distribution.
we apply the estimation model to the second set of
Estimation was conducted by applying the regression projects and assess the degree of estimation accuracy
obtained above to the second data set consisting of is then presented; and (4) Variance Analysis, where
ten projects to estimate the time and cost required for budgeted and actual time and costs for each activity
every activity. The estimated and actual man-days are compared and variances analyzed. The overall
were then compared and the MARE computed for the ABC model resulting from the analysis is shown in
ABC model. Relatively low MARE would suggest that Figure 2.
over-fitting was not a serious problem.
Cost Analysis
Variance analysis was used to examine the differences
between the estimated and actual time required for
This analysis shows how cost from resource pools is
each project. Detailed variance analysis is a technique
traced to individual activities. Cost figures have been
adopted from the costing literature (Hongren, et al.,
transformed to protect the confidentiality of the
1997). The total variance between actual and
organization being studied.
estimated time and cost can be better understood if we
divide it into two components, namely volume variance
Resource Pool Costs
and flexible budget variance (Hongren et al., 1997).
Volume variance is the difference between the
standard cost and the estimated cost. It arises from Departmental costs were categorized into the resource
changes in production volume. In the context of pools, namely: (1) Project Managers, (2) System
systems development, changes in user requirements Analysts, (3) Business Analysts, (4) Programmers, and
after the commencement of the project result in (5) Development Support, as shown in Table 2.
changes to the cost driver counts, and hence the Development support included the cost of some
actual effort expended. Flexible budget variance technical staff, depreciation, and rental expense.
reflects the productivity of the resources used in the
project development. This can be due to changes in Activity Costs
resource mix, staff skill levels, use of productivity tools,
learning, development downtime, and others. Ten activities were identified, namely: (1) Project
Management, (2) Requirement Analysis, (3) Detailed
Projects with high variances for any activity were Design, (4) Front-end Programming, (5) Back-end
examined by reviewing interview notes and discussing Programming, (6) System Testing, (7) User
with project managers in order to understand the Acceptance Testing, (8) User Procedure and Training,
source of the variance. (9) Migration, Conversion and Rollout, and (10) Post

Project Roles Project System Programmer Business Development Total


Manager Analyst Analyst Support
Total Annual Costs $ 5,200,000 $ 10,500,000 $ 12,500,000 $ 1,000,000 $ 8,500,000 $ 37,700,000
1 3
Total Man-days 8,000 30,000 52,000 4,000 94,000 94,000
2
Cost per Man-day $ 50 $ 350 $ 240 $ 250 $ 90 $ 401

1 - Total Man-days = (Number of Staff) * (Number of Annual Working Days per Staff)
2 – Cost per Man-day = (Total Annual Costs) / (Total Man-days)
3 - 94,000 was computed by adding the total number of man-days for all staff in the development department.

Table 2. Computation of Rate per Man-day for Resource Pools

60 The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3)
Figure 2. ABC Model for Software Development and Implementation

The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3) 61
Project System Business Development Total project
Activities Manager Analyst Programmer Analyst Support Time
Project Management 37.43% 6.57% 0.85% 1.20% 7.36% 7.36%
Requirement Analysis 12.42% 13.44% 1.45% 35.91% 10.22% 10.22%
Detailed Design 11.70% 10.02% 0.27% 1.68% 5.69% 5.69%
Programming - Front-end 3.34% 9.13% 34.50% 2.41% 13.66% 13.66%
Programming - Back-end 2.82% 19.35% 41.12% 0.00% 27.25% 27.25%
System Testing 13.92% 14.54% 7.27% 8.90% 11.93% 11.93%
User Acceptance Testing 6.79% 10.23% 9.83% 14.15% 9.40% 9.40%
User Procedure and
Training 2.13% 1.93% 1.10% 32.45% 2.72% 2.72%
Migration, Conversion &
Rollout 9.11% 12.98% 3.27% 3.29% 10.73% 10.73%
Post Implementation Review 0.34% 1.82% 0.36% 0.00% 1.05% 1.05%
Total Project Time 100.00% 100.00% 100.00% 100.00% 100.00% 100.00%

Table 3. Time Allocation from Resource Pools to Activities (average based on 22 projects)

Implementation Review. The average percentage of this activity’s cost. Note that the total activity costs
time spent by each resource pool is presented in Table reflect the organizational-specific mix of different
3 tracing the percentage of time spent by each resource pools (see Table 4).
resource pool, development support cost was
Similar computation was carried out for each of the
allocated to each activity based on the total percentage
other nine activities. Table 5 shows the total annual
of time spent in each activity, as shown in the second
costs, total annual man-days and cost per man-day for
last column of Table 3.
each of the ten activities.
The resource pool costs were then allocated to the ten As expected, Project Management was the most
activities based on the time allocation percentages expensive activity ($619/ man-day) as much of the
shown in Table 3. For example, since project time spent was by project managers, the most
managers spent 12.42% of their total annual days in expensive resource pool. Programming activities were
Requirement Analysis activity, 12.42% of the total much cheaper, costing only $315 to $351 each day to
annual salary for project managers was assigned to perform.

% of Time Annual Man-days Cost Cost per


Resource Pool Spent Man-days Annual Costs Used Allocated Man-day
(a) (b) (c) (d) =(a) * (b) (e) = (a) * (c) (f) = (e) / (d)
Project Manager 12.42% 8,000 $5,200,000 994 $645,840 $650
System Analyst 13.44% 30,000 $10,500,000 4,032 $1,411,200 $350
Programmer 1.45% 52,000 $12,500,000 754 $181,250 $240
Business Analyst 35.91% 4,000 $1,000,000 1,436 $359,100 $250
Development
1
Support 10.22% 94,000 $8,500,000 7,216 $868,700 $120
Total Cost for RA 7,216 $3,466,090 $480
1 - Development support cost per man-day is derived by dividing total support costs of $868,700 by
total man-days spent in this activity, 7,216 man-days.

Table 4. An Example of Activity Cost Computation - Requirement Analysis Activity

62 The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3)
Activities Tot Annual Cost Tot Annual Man-days Cost / Day
(1) Project Management $ 3,379,757 5,456 $ 619
(2) Requirement Analysis $ 3,465,767 7,215 $ 480
(3) Detailed Design $ 2,194,477 4,148 $ 529
(4) Programming - Front-end $ 6,629,768 21,043 $ 315
(5) Programming - Back-end $ 9,634,239 27,412 $ 351
(6) System Testing $ 4,262,157 9,611 $ 443
(7) User Acceptance Testing $ 3,596,391 9,289 $ 387
(8) User Procedure and Training $ 1,005,581 2,615 $ 384
(9) Migration, Conversion & Rollout $ 3,189,450 6,453 $ 494
(10) Post Implementation Review $ 342,413 757 $ 452

Total Project Time $ 37,700,000 94,000 $ 401

Table 5. Costs Allocation from Resource Pools to Activities


1
Drivers Definition
(1) Project Duration Estimated man-days for all the activities except for Project Management, computed
automatically by the ABC model.
(2) Project Type Classify the project into (1) In-house development, (2) In-house enhancement, or (3)
Package projects.
(3) Number of Functions For systems with user screens, count the number of functions on each of the
screens, as perceived by the users. For systems with common interface programs
(no screen), count function as number of transaction types. If the system is an in-
house enhancement or a package, count only the number of functions that need to
be developed and tested by the project team, and the functions that need the team
to provide user training.
(4) Number of Back-end Count the number of back-end/ host programs in the system.
Programs
(5) Number of Front-end Count the number of front-end/ client programs in the system.
Programs
(6) Number of Files Sum the number of the following files: (1) Source files - files that the programs need
to extract data from, (2) Target files - files that the programs need to write to/ output
data to, and (3) Files to be converted – files that need to be converted before the
system can be rolled-out to the production environment.
(7) Number of Screens Count the number of screens in the system, as viewed by the users.
(8) Integration Factor Classify the project into 2 categories: (1) Project with up to 10 back-end/ host
programs, and (2) project with more than 10 back-end/ host programs.
1-Detailed examples accompanying each driver in the ABC manual are prepared for the organization’s use.

Table 6. Definition of Drivers

Cost Driver Analysis enhancement or package implementation), and


through number of back-end (mainframe host) and
Seven potential cost drivers were initially identified
front-end programs (client programs).
from interviews and discussions with the project team
members, including (1) Project Duration, (2) Project Stepwise regression was run for each activity, with
Type, (3) Number of Functions, (4) Number of Back- actual time spent as the dependent variable and
end Programs, (5) Number of Front-end Programs, (6) potential cost drivers identified as independent
Number of Files, and (7) Number of Screens. The variables. Out of the ten activities that we analyzed,
definitions for these drivers are shown in Table 6 that five of them were non-linear, three regressions were
follows. The nature of project is taken into account linear and two were not significant. Table 7 shows the
through the project type driver (development, summary results from the regressions.

The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3) 63
Non-Linear Regressions
(λ)
Yi -1
_ __ __ __ __ __ __ __ __ __ = β0 + β1 X 1i + …..+ βk X k1 + εi
λ

Activity Variables Estimate t-value p-value


Project Management Intercept 2.1554 5.272 0.001
Duration 0.0021 3.295 0.013
λ (0.0400)
R² 0.6080
Adjusted R² 0.5520
F-test (model) 10.859 0.013
Activity Variables Estimate t-value p-value
Requirement Analysis Intercept 2.9546 6.037 0.001
Functions 0.0457 3.509 0.010
Project type - Enhancement (1.5378) (2.344) 0.052
λ 0.1100
R² 0.7485
Adjusted R² 0.6766
F-test (model) 10.415 0.008
Activity Variables Estimate t-value p-value
Programming - Back-end Intercept 3.7714 9.160 -
Back-end Programs 0.0527 4.346 0.002
λ 0.0400
R² 0.6773
Adjusted R² 0.6414
F-test (model) 18.884 0.002
Activity Variables Estimate t-value p-value
User Acceptance Testing Intercept 5.5324 4.305 0.003
Total Files 0.0346 4.792 0.001
λ 0.4300
R² 0.7416
Adjusted R² 0.7093
F-test (model) 22.964 0.001
Activity Variables Estimate t-value p-value
Migration, Conversion, & Rollout Intercept 3.1003 6.583 -
Total Files 0.0081 2.686 0.028
λ 0.0400
R² 0.4742
Adjusted R² 0.4084
F-test (model) 7.214 0.028

Table 7. Regressions for ABC Model (Part 1 of 2)

Project Management, Requirements Analysis, Back- more time is required in management activities, such
end programming, User Acceptance Testing, and as progress meetings and supervision. The time spent
Migration, Conversion, & Rollout are four of the in Requirement Analysis depends on the number of
activities that are non-linear. All these activities functions and the project type. For enhancement
demonstrate diseconomies of scale when the number projects, the time required for this activity is
of drivers increases. These activities are characterized significantly less compared to development and
by significant user involvement. The larger the project package projects. This finding is intuitive, as
in terms of duration, functions, and files, the more time enhancement projects require less effort in analyzing
is required for IT personnel to communicate and requirements, compared to new development and
coordinate among themselves and with users. The package projects. Number of files is the driver for two
driver for Project Management is duration of the activities, namely User Acceptance Testing and
project, which implies that the bigger the project, the Migration, Conversion, and Rollout.

64 The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3)
Linear Regressions
Yi = β0 + β1 X1i + …..+ βk Xk1 + εi
Activity Variables Estimate t-value p-value
Detailed Design Intercept 9.1004 0.732 0.497
Functions 1.1476 3.122 0.026
R² 0.6610
Adjusted R² 0.5932
F-test (model) 9.750 0.026
Activity Variables Estimate t-value p-value
Programming - Front-end Intercept 54.7930 2.265 0.064
Functions 1.8569 2.749 0.033
R² 0.5574
Adjusted R² 0.4836
F-test (model) 7.556 0.033
Activity Variables Estimate t-value p-value
System Testing Intercept 3.0482 0.155 0.882
Functions 0.8355 1.608 0.159
Integration Factor 77.3510 3.268 0.017
R² 0.7752
Adjusted R² 0.7003
F-test (model) 10.346 0.011

Table 7. Regressions for ABC Model (Part 2 of 2)

Back-end Programming also shows diseconomies of Implementation Review, as most projects did not
scale when the number of back-end programs document the time expended for these activities. A
increases, even though it is not directly affected by the constant of 10 and 5 man-days were recommended by
number of users involved. This could be due to the many project managers to estimate the time required for
increased demand for integration among back-end User Procedures & Training, and Post Implementation
programs. As the number of programs increases, Review, respectively. With more project data being
programmers not only have to code the programs, but collected, the ABC model can be updated to include the
also to take into account the interrelationships among drivers and rates for these two activities.
programs.
These findings provide insights into the composition of
Linear regressions include Detailed Design, Front-end development costs. The non-linear component in this
Programming, and System Testing. These activities are study takes up about sixty percent of the total
largely performed by IT personnel and require less development time, and this may explain the results
interaction with users. An increase in each additional reported in prior studies, that the relationship between
driver does not result in a greater than proportional total development time and independent variables is
increase in effort for these activities. non-linear.

Efforts required in both Detailed Design and Front-end Estimation Using ABC Model
Programming are dependent on the number of functions
in the system. For System Testing, number of functions The regressions obtained from the analysis were
applied to the second set of ten projects to evaluate the
and integration factor were the main drivers. The
accuracy of the model. Total project costs were
integration factor is a new driver that was not identified
calculated by multiplying the estimated number of man-
during the earlier interviews. After the analysis of data,
and subsequent discussions with project managers, it days for each activity by the activity cost per man-day
was identified as one of the drivers for System Testing. for each activity and summing across activities. This
computation clearly shows that the number of man-days
This driver categorizes the project into two groups
is not the only factor determining the cost of the project;
(projects with up to 10 back-end programs, and projects
the mix of the resources is also important. For example,
with more than 10 back-end programs), effectively
one project is estimated to use 209 man-days and cost
capturing increased effort required in testing the
integration among programs in the back-end. $82,501; whereas another is estimated to require a
larger number of man-days (i.e., 252) but costs less, at
All the above regressions showed significant results $75,181. This is because the first project consumed
except for User Procedures & Training, and Post more time from an expensive activity.

The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3) 65
The accuracy of the model in predicting the time described in this paper (Finnie et al., 1997; Srinivasan &
required for the projects can be assessed in a number Fisher, 1995). 90% of the projects reported MRE of less
of different ways. When we apply our model to the than 100%, and the MARE was a relatively low 39.06%.
second set of ten projects, the percentage of error
(Boehm, 1981) is within 1.2 of the actual time (variance Variance Analysis
< 20%) 40% of the time (four out of the total of ten
projects); and the estimates are within a factor of 2 of An example of detailed variance analysis is presented
the actual time (variance < 100%) for all but one project. in Table 8. The ABC model only requires project
The magnitude of relative error (MRE) (Conte et al., managers to provide the estimated cost driver values
1986; Kemerer, 1987) is derived by taking the actual at the beginning of the project, and the actual cost
time spent for the activity less the time estimated, and driver values on completion of the project. Actual time
the difference in absolute terms was then divided by the spent is captured from project members’ time sheets.
actual time spent. The average of MRE is called the The estimated, standard and actual costs can then be
Mean Absolute Relative Error (MARE) previously computed using the ABC model.

Driver Count Estimated Actual Driver


Driver Counts
Counts
1. Project Type 0 0
2. # of Functions 1 4
3. # of B/End Programs 5 10
4. Integration Factor 0 0
5. Total # of Files 10 49

Time Variance Table


Flexible
Actual # of Standard # Estimated # of Total Volume Budget
Man-days of Man-days Man-days Variance Variance Variance
Activities (a) (b) (c) (d) = (a) - (c) (e) = (b) - (c) (f) = (a) - (b)
Project Management 15.5 15.2 (10.2) 0.3
5.0 (10.5)
Requirement Analysis 14.8 13.4 (0.8) 1.4
12.6 (2.2)
Detailed Design 13.7 10.2 3.4 (2.3)
11.4 1.2
Programming - F/E 56.6 56.6 19.3
75.9 19.3 -
Programming - B/E 52.8 42.1 (15.9) 10.7 (26.6)
26.2
System Testing 6.4 3.9 2.5 9.9
16.3 12.4
User Acceptance Testing 26.7 18.8 8.0 1.6
28.3 9.5
User Procedure and Training 10.0 10.0 (10.0) (10.0)
- -
Migration, Conversion & 26.4 20.0 6.5 (5.8)
Rollout 20.6 0.6
Post Implementation Review 5.0 5.0 7.9
12.9 7.9 -
Total Project Time 228.0 195.2 32.8 (18.8)
209.2 14.0
(a): Actual # of Man-days: Extracted from time sheets.
(b): Standard # of Man-days: Obtained from regression equations using actual drivers as independent variables (Actual
drivers * Man-days per driver)
(c): Estimated # of Man-days: Obtained from regression equations using estimated drivers as independent variables.
(Estimated drivers * Man-days per driver)

Table 8. Example of Variance Analysis (Part 1 of 2)

66 The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3)
Cost Variance Table
Activities Actual Cost Standard Cost Estimated Total Variance Volume Flexible Budget
(a) (b) Cost (d) = (a) - (c) Variance Variance
(c) (e) = (b) - (c) (f) = (a) - (b)
Project Management $ 3,097 $ 9,610 $ 9,426 $ (6,329) $ 184 $ (6,512)
Requirement Analysis $ 6,053 $ 7,115 $ 6,421 $ (368) $ 694 $ (1,062)
Detailed Design $ 6,031 $ 7,242 $ 5,421 $ 609 $ 1,821 $ (1,212)
Programming - F/E $ 23,913 $ 17,848 $ 17,848 $ 6,065 $ - $ 6,065
Programming - B/E $ 9,208 $ 18,561 $ 14,806 $ (5,598) $ 3,754 $ (9,352)
System Testing $ 7,229 $ 2,834 $ 1,722 $ 5,506 $ 1,112 $ 4,395
User Acceptance $ 10,957 $ 10,346 $ 7,263 $ 3,694 $ 3,083 $ 611
Testing
User Procedure and $ - $ 3,845 $ 3,845 $ (3,845) $ - $ (3,845)
Training
Migration, Conversion & $ 10,181 $ 13,066 $ 9,876 $ 305 $ 3,190 $ (2,885)
Rollout
Post Implementation $ 5,833 $ 2,261 $ 2,261 $ 3,572 $ - $ 3,572
Review
Total Project Costs $ 82,501 $ 92,726 $ 78,889 $ 3,612 $ 13,837 $ (10,226)
(a): Actual Costs: Actual man-days extracted from time sheets * Cost per man-day for each activity
(b): Standard Costs: Standard # of Man-days * Cost per man-day for each activity
(c): Estimated Costs: Estimated # of Man-days * Cost per man-day for each activity

Table 8. Example of Variance Analysis (Part 2 of 2)

Total Variance is the difference between the estimated Discussion


and actual time and cost recorded (column (d)). In the
example, the team exceeded the budget by $3,612. The data generated in the process of building and
The change in user requirements after the applying the ABC model, accumulated and interpreted
commencement of the project changes the cost driver over time, can provide a useful pool of knowledge for
counts, hence the volume of output. Volume variance the organization, particularly for resource allocation
is the difference between the standard cost and the and improving software development and
estimated cost (column (e)). In this example, the implementation processes. Senior managers view
positive variance of 32.8 man-days and $13,837 such knowledge as an important organizational
indicates that standard time and cost allocated is more resource (Adler, 1989; Baskerville & Pries-Heje, 1999;
than the estimated time and cost. This is because Larsen & Levine, 1999), and its acquisition,
changes in user requirements that led to an increase in articulation, and enhancement over time contributes to
actual driver counts, including functions, back-end the uniqueness of the organization (Dodgson, 1993).
programs, and total files. In this example, the total In the following sections, we discuss the potential of
negative flexible budget variance of 18.8 man-days different information components of the ABC approach
and $10,266 shows that the project team spent less for management control and decision-making, and for
time and cost in the project than expected based on organizational learning.
the actual driver counts. Analysis can be carried out at
the activity level to fully understand the reasons for the Resource and Activity Time and Cost
variances. The resource pool analysis required by the ABC model
provides cost and capacity information on the different
The four sections above demonstrate the complete types of resources used in the software development
cycle of the development and use of the ABC model, process. This information helps management in
from the building of the model, which includes resource planning and in making outsourcing
determining and validating the resource pools, decisions. During the course of this study, discussions
activities, activities’ drivers and rates, to its use in cost with senior IT management highlighted the usefulness
estimation and variance analysis. In the following of specific information provided by the model. For
sections, the usefulness of the data generated during example, senior managers were very interested in the
the ABC cycle is discussed. charge-out rate for each resource pool (Table 2)

The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3) 67
because this sensitized them to the resource mix Estimation and Variance Analysis
implications for each project. Senior managers were Using ABC Model
also very interested in the average percentage of time
At the end of each project, project managers and
spent by each resource pool in the major activities
management should investigate activities with
(Table 3). The data showed that relatively expensive
significant variance between estimated and actual time
resources such as project managers and system
and cost to surface the causes. This variance analysis
analysts were spending more time than was desirable
serves as a natural feedback loop for the managers to
in lower value-adding activities such as programming.
move towards more effective and efficient project
Finally, the average cost of each activity (Table 5)
development and management practices.
potentially provided information for decisions about
which development activities to outsource. As we examined and discussed the significant
variances in detail with project managers, we
Activities and Cost Drivers uncovered many learning opportunities. For example,
we found that the variances for Project Management
The activities identified in this study reflect the project and User Acceptance Testing activities were especially
management structure and the terminology used by high for projects involving external parties (e.g., other
the organization. Other organizations may have a financial institutions, governance bodies, etc.). The
somewhat different set of activities. In general, short-run solution is to revise and increase the effort
because this organization’s set of activities closely estimated for new projects with similar characteristics.
resemble the traditional systems development life However, this approach is merely reactive, and
cycle, comparison of the ABC data on proportion of increasing the time allowed may actually legitimize
project time spent in each activity with findings from inefficient practices. In the long run, the proactive way
other studies provided useful insights to management. of eliminating these variances is to improve on the
processes/ sub-tasks within the activities that
contribute to the variances. In this case, better
The total project time allocation patterns (Table 3, last
coordination procedures with external stakeholders
column) were similar to those reported by Beck and
and technology for coordination should be considered.
Perkins (1983) and Dolado (1997), with the exception
In another example, we found that one of the projects
of detailed design and programming. Time spent in
had extremely low programming time, relative to the
detailed design in the organization studied was
model’s estimates. When interviewing the project
significantly lower than that reported in previously
manager, we found that this was because the use of a
published studies (6% as compared to 27% and 22%
new code generator. The ABC data was thus able to
for Beck & Perkins’ and Dolado’s study, respectively);
provide the organization with an early indicator of
and the time spent in programming was much higher
positive impacts from the use of this programming aid.
(41% as compared to 24% and 34%). Discussion with
Learning can be done on a systematic basis if project
project managers led to the conclusion that design
managers continuously use ABC results to track
effort had been traded-off against programming/ bug
variances across projects adopting new tools or
fixing effort. This highlighted an area of prevailing
technology. Tools that deliver significant benefits can
organizational development practice for managers to
be identified and their use institutionalized. Tracking
consider for change.
such projects over time can also give the managers a
better idea of the learning curve for new technology.
The cost driver for each activity is also contextualized As new development projects are added to the project
to the software development practices and platforms of database, analysis of their development processes
the organization. For example, the use of front-end and resource consumption will result in periodic
program and back-end program as drivers arises from revision of the model.
the organization’s use of a client-server development
platform for its systems.
Implications for Research and Practice
The cost driver rates are of much interest to the This study proposes the application of the ABC
management of the organization. The rate indicates approach to software development and
how much it costs to change requirements (i.e., implementation. Because of the novelty of the
increase the number of a cost driver). This provides approach in this context, the paper provides a
the management with a tool for better management description of how an ABC model can be created and
and control of project time and cost. It also provides a grounds the process and model in empirical data. The
basis for negotiating with users on the cost of the paper also examines the main advantages of applying
project, both at the initial estimation, and as ABC to software development and implementation
subsequent user requests for changes arise. projects. First, the approach formalizes the relationship

68 The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3)
between software development time and IS most of the equations. The second issue is the inability
department costs using a standard organization-wide to model for a variety of differences in project
estimation practice. The ABC approach supplements characteristics. Only type of project (enhancement, in-
the time estimates provided by traditional software house development, or package implementation) was
estimation approaches with detailed costing considered in the analysis. Despite the small sample,
information that highlights the impact of resource mix study provides a rich description of the development of
decisions, and consumption of different development a novel approach to software estimation, and of its
activities. Second, the costing model is contextualized application and informational benefits. Future studies
to the structure and practices of the organization. This can increase the number of projects by using a
is likely to increase accuracy in driver counts by project longitudinal approach, and sampling over a longer time
managers as the contextualized drivers are a better fit frame. This raises challenges of continued access to
with the organization. This, together with the the research site over a longer period of time, or
customized cost driver rates are likely to result in more access to historical projects. The latter is often a
accurate time and cost estimates. Last, and most problem due to lack of good archival records, and
importantly, the integration with detailed costing turnover of the IT personnel involved.
information, contextualization of the model, and
variance analysis results in information that is useful Future research may include additional drivers to
for decision making and organizational learning. increase estimation accuracy of the ABC model.
System complexity and integration were two possible
For the perspective of practitioners, the study also drivers suggested by the senior management at the
suggests a number of recommendations to increase last focus group session. These drivers may account
the benefits received. Activities and cost drivers must for the difference in time required for coding programs
be clearly defined with a list of specific examples in and testing the programs and the system. However,
order to simplify counts and minimize inter-rater the inclusion of additional drivers, while possibly
variances. Strong IS participation in identifying increasing accuracy, will also increase model
activities and cost drivers is necessary to minimize complexity. The problems of increased model
resistance to adoption of the model and to ensure that complexity include user resistance, increased inter-
the model is relevant and useful. The model needs to rater variances among project team members in driver
tbe regularly updated to reflect changes in counts, and delay in estimation.
organizational resource mix and development
platforms and practices. Collection of the data required Finally, research can also study the extension of the
(eg. timesheets) for the model should be automated to ABC approach to other IS domains such as
minimize the effort required to maintain the model. operations. The ABC approach can also be extended
into the vendor organizations domain, where the
The ABC approach also has its limitations. While the costing information may be particularly useful for
approach presented in this paper for the building of costing client projects, and where there may be more
ABC model can be generalized to other organizations’ projects across several client organizations.
software development and implementation, and even
to other functions such as IT operations, the model
itself may require significant customization before it
References
can be used in another organization. The ABC model Adler, P.S. (1989). “When Knowledge is the Critical
parameters (e.g., resource rates and driver rates) Resource, Knowledge Management is the Critical
cannot simply be adopted by another organization. Task,” IEEE Transactions on Engineering
The model (activities and drivers) can be used as a Management, Vol.36, No.2, pp. 87-94.
starting point for other organizations to adapt and use,
but changes have to be made to customize the model Albrecht, A.J. (1979). “Measuring Application
to better serve the needs of the organization. The Development Productivity,” Proceedings of Joint
analysis of resources, costs and utilization in activities SHARE/GUIDE/ IBM Application Development
required to customize the model to an organization Symposium, pp. 34-43.
introduce significant setup costs for this approach. Albrecht, A.J. and Gaffney, J.E. (1983). “Software
Function, Source Line of Codes, and Development
The study itself is subject to the limitation of having Effort Prediction: A Software Science Validation,”
only twenty-two projects. The sample size raises two IEEE Transaction on Software Engineering, Vol.
concerns. One is with the reliability of the regression 19, pp. 639-648.
results. Attempts to mitigate this included testing that Albus, J.S (1981). Brain, Behavior, and Robotics,
the assumption for regression were not violated, and Peterborough: Byte Books, Subsidiary of McGraw-
using only one independent variable (cost driver) in Hill.

The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3) 69
Athey, T. (1998). “Leadership Challenges for the Ebrihimi, N.B. (1999). “How to Improve Calibration of
Future,” IEEE Software, Vol. 15, No. 3, pp. 72-77. Cost Models,” IEEE Transactions on Software
Banker, R.D., Davis, G., and Slaughter, S.A. (1998). Engineering, Vol. 25, No. 1, pp. 136-140.
“Software Development Practices, Software Finnie G.R., Wittig, G.E., and Desharnais, J.M. (1997).
Complexity, and Software Maintenance “A Comparison of Software Effort Estimation
Performance,” Management Science, Vol. 44, No. Techniques: Using Function Points with Neural
4, pp. 433-450. Networks, Case-based Reasoning and Regression
Banker, R.D. and Slaughter, S.A. (1997). “Efficiency of Models,” Journal of Systems Software, Vol. 39, pp.
Complexity Allocation in Software Design: An 281-289.
Empirical Evaluation,” Management Science, Vol. Grandy, R.B., and Caswell, D.L. (1987). Software
43, No. 12, pp. 1709-1725. Metrics: Establishing a Company-Wide Programs,
Baskerville, R. and Pries-Heje, J. (1999). “Knowledge Englewood Cliff, New Jersey: Prentice Hall.
Capability and Maturity in Software Management,” Hall, T., and Fenton, N. (1997). “Implementing
The DATA BASE for Advances in Information Effective Software Metrics Programs,” IEEE
Systems, Vol. 30, No. 2, pp. 26-42. Software, pp. 55-64.
Beck, L.L., and Perkins, E.T. (1983). “A Survey of Helmi, M.A., and Hindi, N. (1996). “Activity-based
Software Engineering Practice: Tools, Methods, Costing in Banking: A Big Challenge,” The Journal
and Results,” IEEE Transactions on Software of Cost and Management Accounting, Vol. 9, Vol.
Engineering, Vol. SE-9, No. 5, pg. 541-561. 2, pp. 5-19.
Benjamin, R.I., Rockart, J.F., Morton, M.S., and Horngren, C.T., Foster, G., Datar, S.M. (1997). Cost
th
Wyman, J. (1984). “Information Technology: A Accounting - A Managerial Emphasis, 9 Edition,
Strategic Opportunity,” Sloan Management New Jersey : Prentice Hall.
Review, Vol.25, No. 3, pp. 3-10. Kaplan, B. and Duchon, D. (1988). “Combining
Boehm, B.W. (1981). Software Engineering Qualitative and Quantitative Methods in
Economics, Englewood Cliffs, N.J.: Prentice-Hall. Information Systems Research: A Case Study,”
Boehm, B.W. (1984). “Software Engineering MIS Quarterly, Vol. 12, No. 4, pp. 571-586.
Economics,” IEEE Transactions of Software Keil, M., Mixon, R., Saarinen, T., and Tuunaiene, V.
Engineering, Vol. SE-10, No. 1, pp. 4-21. (1995). “Understanding Runaway Information
Bradway, B. and Ross, S. (2000). “Measuring Technology Projects: Results from an International
Corporate Customer Profitability: The Role of Research Program based on Escalation Theory,”
Activity-based Cost Analysis,” Corporate Customer Journal of Management Information Systems, Vol.
Management, Vol. 4, Research Brief 6, pp. 1-10, 11, No. 3, pp. 65-85.
(http://Meridien-research.com). Kemerer C.F. and Porter, B.S. (1992). “Improving the
Conte, S., Dunsmore, H., and Shen, V. (1986). Reliability of Function Point Measurement: An
Software Engineering Metrics and Models. Menlo Empirical Study,” IEEE Transactions on Software
Park, California: Benjamin/ Cummings. Engineering, Vol. 18, No. 11, pp. 1011-1024.
Cooper, R. and Kaplan P.S. (1988). “Measure Costs Kemerer, C.F. (1993). “Reliability of Function Points
Right: Make the Right Decisions,” Harvard Measurement,” Communications of the ACM,
Business Review, pp. 96-103. Vol.36, No.2, pp. 85-97.
DeMarco, T. (1982). Controlling Software Projects. Kemerer, C.F. (1987). “An Empirical Validation of
Englewood Cliffs, N.J.: Prentice Hall. Software Cost Estimation Models,”
Communications of the ACM, Vol.30, No.5, pp.
Dodgson, M. (1993). “Organizational Learning: A 416-429.
Review of Some Literature,” Organization Studies,
Vol.14, No. 3, pp. 375-394. Kmenta, J. (1986). Elements of Econometrics, New
York: Macmillan Publishing Company; London:
Dolado, J.J. (1997). “A Study of the Relationship Collier Macmillan Publishers, pp.517-521.
Among Albrecht and MK II Function Points, Lines
of Codes, 4GL and Effort,” The Journal of System Kroll, K.M. (1996). “The ABCs revisited,” Industry
and Software, Vol. 37, No. 2, pp. 161-173. Week, Vol. 254, No. 22, pp. 19-21.
Dugger, R. (1996). “Selling the Decision Maker,” Larsen, T.J. and Levine, L. (1999). “DataBase Special
Datamation, July, pp. 89-90. Issue – Information Systems: Current Issues and
Future Changes,” The DATA BASE for Advances
in Information Systems, Vol. 30, No. 2, pp. 7-11.

70 The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3)
Lederer, A.L. and Prasad, L. (1995). “Perceptual Seaman, C.B. (1999). “Qualitative Methods in
Congruence and Systems Development Cost Empirical Studies of Software Engineering,” IEEE
Estimation,” Information Resource Management Transactions on Software Engineering, Vol. 25,
Journal, pp. 16-27. No. 4, pp. 557-572.
Low, G.C. and Jeffery, D.R. (1990). “Function Points in Srinivasan, K., and Fisher, D. (1995). “Machine
Estimation and Evaluation of the Software Learning Approaches to Estimating Software
Process,” IEEE Transaction on Software Development Effort,” IEEE Transactions on
Engineering, Vol. 16, No. 1, pp. 64-71. Software Engineering, Vol. 21, No.2, pp. 2126-
Mabberley, J. (1998). Activity-based Costing in 2137.
Financial Institutions, Financial Times Watson, I.D. and Marir, F. (1994). “ Case-Based
Management, Pitman Publishing. Reasoning: An Overview,” Knowledge Engineering
Maglitta J. (1991). “It’s Reality Time,” Computerworld, Review Journal, Vol. 9.4, pp. 327-354.
pp. 81-84. Yin, R.K. (1984). Case Study Research: Design and
Miller, J. (1996). Implementing Activity-Based Methods, Beverly Hills, CA: Sage Publications.
Management in Daily Operations, John Wiley &
Sons, Inc. About the Authors
Mukhopadhyay., T., Vicinanza., S.S., and Prietula,
Ginny Ooi is an Assistant Manager in the Finance
M.J. (1992). “Examining the Feasibility of a Case-
based Reasoning Model for Software Effort department of a Singapore banking institution. She
Estimation,” MIS Quarterly, Vol. 15, No. 2, pp. 155- received her Master by Research in Business
(Information Systems) and Bachelor of Accountancy
171.
(Honours) from the Nanyang Business School. Her
Pfleeger, S.L. (1993). “Lessons Learned in Building a research interests include cost and effort estimation for
Corporate Metrics Program,” IEEE Software, pp. software development and implementation, changes in
67-74. business processes and impacts on the organization
Reimann, B.C. and Kaplan, R.S. (1990). “The ABCs of with the use of IT and Internet, and ECommerce
Accounting for Value Creation,” Planning Review, business models.
Vol.18, No. 4, pp. 33-34. Christina Soh is the Head of the Division of
Rubin, H.A. (1991). “Measure for Measure,” Information Technology and Operations Management
Computerworld, pp. 77-79. and Director of the Information Management Research
Rudolph, E.E. (1983). “Productivity in Computer Center (IMARC), Nanyang Business School, Nanyang
Application Development,” Dep. of Management Technological University in Singapore. She received
Studies, University of Auckland, Australia. her Ph.D. in Management from the University of
California, Los Angeles. Her research focuses on ERP
Samson, B., Ellison, D., and Dugard, P. (1997). implementation, ECommerce business models, IT
“Software Cost Estimation Using an Albus business value and national IT policy.
Perceptron (CMAC),” Information Software
Technology, Vol. 39, No. 1, pp. 55-60.

The DATA BASE for Advances in Information Systems - Summer 2003 (Vol. 34, No. 3) 71

You might also like