Software Cost Estimation for Global Software Development

A Systematic Map and Review Study

Manal El Bajta

, Ali Idri

, Jos

e Luis Fern

andez-Alem

, Joaquin Nicolas Ros

and Ambrosio Toval

Software Project Management Research Team, ENSIAS, Mohammed V University, Rabat, Morocco

Software Engineering Research Group, Regional Campus of International Excellence “Campus Mare Nostrum”,

University of Murcia, Murcia, Spain

Keywords:

Systematic Mapping Study, Software Cost Estimation, Global Software Development.

Abstract:

Software cost estimation plays a central role in the success of software project management in the context of

global software development (GSD). The importance of mastering software cost estimation may appear to be

obvious. However, as regards the issue of customer satisfaction, end-users are often unsatisﬁed with software

project management results. In this paper, a systematic mapping study (SMS) is carried out with the aim of

summarising software cost estimation in the context of GSD research by answering nine mapping questions.

A total, of 16 articles were selected and classiﬁed according to nine criteria: publication source, publication

year, research type, research approach, contribution type, software cost estimation techniques, software cost

estimation activity, cost drivers and cost estimation performances for GSD projects. The results show that the

interest in estimating software cost for GSD projects has increased in recent years and reveal that conferences

are the most frequently targeted publications. Most software cost estimation for GSD research has focused on

theory. The dominant contribution type of software cost estimation for GSD research is that of models, while

the predominant activity was identiﬁed as being software development cost. Identifying empirical solutions to

address software cost estimation for GSD is a promising direction for researchers.

1 INTRODUCTION

Global Software Development (GSD) refers to soft-

ware work starting at geographically separated areas

across national boundaries considering synchronous

and asynchronous interaction. GSD has been adopted

by numerous companies. However, these global

projects confront a number of problems, which are

particularly linked to the gap between different par-

ticipants: physical distance between the groups of

developers causing a lack of trust, time-zone differ-

ences, communication problems among teams, effort

estimation problems, cultural differences, and oth-

ers. Current research tends to characterise these prob-

lems, but if success is to be achieved in GSD, compa-

nies must minimise challenges by adjusting their pro-

cesses and rearranging their tools and organisational

structure.

GSD projects can increase requirements as re-

gards development processes, project management

practices, architecture, quality, collaboration tools

and so on. These challenges may exceed the advan-

tages of the lower labour rates in the developing coun-

try since they could lead to substantial overheads in

the day-to-day operations of a GSD project. This rea-

soning shows that it is vital to understand and estimate

the total costs of GSD in order to help evaluate the

comparison with local software development in terms

of efﬁciency.

A large range of software cost estimation tech-

niques had already been developed before the GSD

trend began (Jørgensen, 2004). Early research on the

topic was conducted in 2006 (Keil et al., 2006), when

researchers were able to promote analyses of project

factors in order to gain insights into the comparison

of development costs for distributed software devel-

opment projects and collocated projects. In a study

published in 2012 (Ramasubbu and Balan, 2012),

researchers advance the question of cost estimation

for distributed software projects by identifying chal-

lenges and proposing solutions with which to better

drive estimates. Britto et al. (Britto et al., 2014)

present a systematic literature review on effort esti-

mation in GSD. In their study, only 5 papers were se-

lected, which allowed the extraction of only 10 esti-

mation methods. It is important to note that the study

of Britto et al. did not consider software maintenance

effort/cost estimation; it only concerned software de-

197

El Bajta M., Idri A., Fernández-Alemán J., Nicolas Ros J. and Toval A..

Software Cost Estimation for Global Software Development - A Systematic Map and Review Study.

DOI: 10.5220/0005371501970206

In Proceedings of the 10th International Conference on Evaluation of Novel Approaches to Software Engineering (ENASE-2015), pages 197-206

ISBN: 978-989-758-100-7

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

velopment effort/cost estimation. What is more, the

study did not classify the techniques according to

their contribution type. Considering the importance

of the above limitations, the objective of this study is

to carry out a systematic review which: 1) includes 16

selected papers among them the 5 ones of Britto et al.,

2) considers software development as well as mainte-

nance effort estimation, and 3) discusses effort/cost

estimation performance.

The paper is structured as follows: Sect. 2

presents the research method used in the study. Sect.

3 reports the results and ﬁndings obtained from the

SMS. Sect. 4 outlines threats to validity. Sect. 5 dis-

cusses the main ﬁndings and presents implications for

researchers and practitioners, while our conclusions

and future work are presented in Sect. 6.

2 RESEARCH METHODOLOGY

2.1 Mapping Questions

The SMS was performed to obtain the current

research on software cost estimation for GSD. This

study answers nine mapping questions (MQs). These

questions are presented in Table 1.

Table 1: Mapping questions.

ID Mapping question

MQ1 Which publication sources and channels are

the main targets for software cost estimation

for GSD research?

MQ2 How has the frequency of software cost esti-

mation for GSD research changed over time?

MQ3 What are the research types of software cost

estimation for GSD studies?

MQ4 Which research approaches are used in soft-

ware cost estimation for GSD studies?

MQ5 What are the contribution types of software

cost estimation for GSD research?

MQ6 Which cost estimation techniques are most

frequently used for GSD projects?

MQ7 Which software cost estimation activities have

been addressed by GSD research?

MQ8 Which cost drivers affect GSD projects?

MQ9 Which cost estimation performances have

been obtained from GSD projects?

2.2 Search Strategy and Paper Selection

Criteria

The articles were identiﬁed by consulting the follow-

ing sources: IEEE Xplore digital library, ACM digital

library, ScienceDirect and Google Scholar. The fol-

lowing search string was used in order to perform the

automatic search in the digital libraries selected:

(Software OR system* OR application*) AND

(cost OR effort OR resource) AND (estimat* OR

plan* OR predict* OR measur* OR calcul* OR

manage* OR control*) AND (Global development

OR distributed development OR outsourc* OR Off-

shor* OR Dispersed development). This search string

was applied to the title, abstract and keywords of the

papers to reduce the search results. Each paper was

retrieved by the ﬁrst author and speciﬁc information

of each relevant paper was ﬁlled in an Ms Excel ﬁle.

The aim of the selection process was to identify

the most relevant studies for this mapping study. Each

paper was retrieved and evaluated by one author who

decided whether it should be included by consider-

ing its title, abstract and keywords. The ﬁnal selec-

tion result was reviewed and approved by the remain-

ing authors. The ﬁrst step after the articles had been

identiﬁed was to eliminate duplicate titles, and titles

which were clearly not related to the review (16 se-

lected studies out of 103 relevant studies). The inclu-

sion criteria were limited to those studies that focused

on software cost estimation for GSD projects, and any

studies that met at least one of the following exclusion

criteria (EC) were excluded:

• EC1. Papers that are not published in journals,

conferences or workshops.

• EC2. Papers that are not in English.

2.3 Quality Assessment (QA) Process

The QA in an SMS is a major focus that increases

the depth of a study. In order to enhance our study,

a questionnaire was therefore designed to assess the

quality of candidate papers. The scoring used in this

questionnaire was determined on the basis of previous

studies (Idri et al., 2015), (Ouhbi et al., 2013b) and

(Ouhbi et al., 2013a).

(a) The paper has been published in a recognized and

stable journal or conference. This question was

rated by considering the computer science confer-

ence rankings in the Computing Research and Ed-

ucation (CORE) 2013 Conference Rankings, and

the 2013 Journal Citation Reports. The possible

answers to this question were:

• For conferences: (+2) if it is ranked CORE

A*; (+1.5) if it is ranked CORE A; (+1) if it

is ranked CORE B; (+0.5) if it is ranked CORE

C; (+0) if it is not in CORE ranking.

• For journals: (+2) if it is ranked Q1; (+1,5) if it

is ranked Q2; (+1) if it is ranked Q3; (+0.5) if it

is ranked Q4; (+0) if it is not in JCR ranking.

ENASE2015-10thInternationalConferenceonEvaluationofNovelSoftwareApproachestoSoftwareEngineering

198

(b) The main focus of the paper is software cost es-

timation activities used to deal with GSD chal-

lenges. Yes (+1); Partially (+0.5); No (+0)

obtained. Yes (+1); Partially (+0.5); No (+0)

(d) The study is empirical and presents relevant re-

sults for our SMS. Yes (+1); No (+0)

2.4 Data Extraction Strategy

The data extraction strategy was based on providing

the set of possible answers to the MQs. The strategy

is explained below:

MQ1. In order to answer this question, it is necessary

to identify the publication source and channel for each

paper.

MQ2. In order to discover the publication trend, the

articles should be classiﬁed per publication year.

MQ3. A research type can be classiﬁed into the fol-

lowing categories:

• Evaluation research: existing software cost esti-

mation for GSD approaches are implemented in

practice and an evaluation of them is conducted.

• Solution proposal: a solution for software cost es-

timation for GSD is proposed. This solution may

be a new software cost estimation for a GSD ap-

proach or a signiﬁcant extension of an existing ap-

proach.

• Other, e.g. opinion paper, experience paper.

MQ4. The research approach can be classiﬁed as be-

ing:

• A case study: an empirical inquiry that inves-

tigates a software cost estimation approach for

GSD within its real-life context.

• A survey: a method used to collect quantitative

software cost estimation for GSD information.

• A experiment: an empirical method applied un-

der controlled conditions in order to observe its

effects on software cost estimation in the GSD

context.

• A review: an analysis of software cost estimation

for existing GSD literature.

MQ5. A contribution can be classiﬁed as being:

• A technique: a procedure used to accomplish a

software cost estimation for a GSD task. e.g. a

data mining technique.

• A model: a representation of a system that allows

software cost estimation for GSD properties to be

investigated.

• Other, e.g. process, tool.

MQ6. Several cost estimation techniques for GSD

projects have been used in the last few decades. These

techniques can be classiﬁed as (Wen et al., 2012) and

(Jorgensen and Shepperd, 2007):

• Expert judgment: this involves consulting a group

of experts in order to use their experience to pro-

pose an estimation of a given project (Hughes,

1996).

• machine learning models: Approaches that are

based on soft computing such as artiﬁcial neural

networks, fuzzy logic models and genetic algo-

rithms (Idri et al., 2006).

• non-machine learning models: these provide lin-

ear and non-linear regression models to establish

equations with which to perform software estima-

tion. (Boehm et al., 2000).

MQ7. Software cost estimation activities that were

addressed by GSD research can be categorised as:

• Software development cost: Performed by man-

agers and software system engineers for activities

such as functional design, software requirement,

development code, development tools, integration

of software and ﬁnally the test procedures.

• Software maintenance cost: Related to the con-

trol and the monitoring of the software after it has

been delivered to the ﬁnal user, since there will

always be problems with the software as it gets

older

• Other, e.g. Reengineering cost

MQ8. Cost drivers that affect GSD projects are di-

vided into 4 categories, namely Product, Platform,

Personnel and Project factors (Boehm, 1981).

MQ9. In this paper, we focus on the two main criteria

that affect GSD projects: geographical and temporal

challenges and their inﬂuence on cost performance.

According to the PMBOK Cost Management knowl-

edge area, cost performance is included in three main

outputs of a GSD project:

• Cost Performance Baseline: An authorised time-

phased budget at completion used to measure,

monitor, and control overall cost performance on

the project.

• Work performance measurements: The calculated

cost variance for work packages and control ac-

counts

• Basis of estimates: The amount and type of ad-

ditional details supporting the cost estimate vary

according to the application area.

SoftwareCostEstimationforGlobalSoftwareDevelopment-ASystematicMapandReviewStudy

199

3 RESULTS

3.1 Quality Assessment

SMSs generally emphasise the quality of selected

studies. This QA is usually carried out to discover

the general view of the paper’s implication in the sub-

ject. However, Kitchenham et al. (Kitchenham and

Charters, 2007) specify that even if some researchers

use QA as a selection criterion in their systematic re-

view, this assessment is not mandatory for an SMS.

Table 2 provides information about the total score of

the selected studies. The majority of the selected pa-

pers (66.25%) have at least a medium score for qual-

ity, which shows that they contain useful information,

particularly as regards information on software cost

estimation and the impact of GSD projects on the cost

estimates. No studies were discarded from these in-

puts during the QA process.

Table 2: Quality levels of relevant studies.

Quality level Papers Percent (%)

Very high ( 4 < score 6 5) 1 6,25

High ( 3 < score 6 4) 1 6,25

Medium ( 2 < score 6 3) 7 43,75

Low ( 1 < score 6 2) 5 31,25

Very low ( 0 < score 6 1) 2 12,50

Table 3 shows the number of articles based on the

ranking of the conference or journal at/in which they

were published.

Table 3: Articles by their journal or conference rank.

Journals Number Conferences Number

Q1 1 CORE A* 1

Q2 2 CORE A 1

Q3 0 CORE B 0

Q4 1 CORE C 6

3.2 MQ.1: Source and Channel of

Publications

Table 4 provides a schematic representation of publi-

cation channels and the number of articles per publi-

cation source. Table 5 presents the journals and con-

ferences at which the papers selected for this SMS

were published. This result clearly shows that the

International Conference on Global Software Engi-

neering (ICGSE) is the main publication source for

our topic. With regard to journals, systems, software

and computer science journals are the targets of re-

searchers in the ﬁeld.

Table 4: Publication channel.

Publication channels Selected papers Percent

Conference 11 68,75%

Journal 5 31,25%

Total 16 100%

Table 5: Journal (J) and Conferences (C) of selected studies.

Publication channels Type Total

International Conference on Global Soft-

ware Engineering (ICGSE)

C 4

Computer Science and Information Tech-

nology (CSIT)

C 2

Software Engineering International Con-

ference (ICSE)

C 1

SRII Global Conference (SRII) C 1

Software Engineering, Artiﬁcial In-

telligence, Networking and Paral-

lel/Distributed Computing (ACIS)

C 1

Services Computing, IEEE International

Conference (SCC)

C 1

Innovations in Information Technology

Conference

C 1

International workshop on Economics

driven software engineering research

(EDSER)

C 1

IEEE Software J 1

Advances in Software Engineering J 1

Systems and Software J 1

European Journal of Scientiﬁc Research J 1

3.3 MQ.2: Publication Distribution Per

Year

Fig. 1 shows the number of publications per year. The

amount of publications interested in software cost es-

timation for GSD projects has increased since 2006.

This year corresponds to the outset of the increasing

concern about the effect of globalization on the soft-

ware industry in general (da Silva et al., 2010) and

is also the year in which the ﬁrst ICGSE conference

took place.

Figure 1: Publication per year.

3.4 MQ.3: Research Type

Sixty two percent of the selected articles are evalua-

tion research, while 25% of the selected papers are so-

ENASE2015-10thInternationalConferenceonEvaluationofNovelSoftwareApproachestoSoftwareEngineering

200

Figure 2: Research Type.

lution proposals. 12,50% are contained in the ”Other”

category, which comprises theoretical papers and ex-

perience papers. Fig. 2 divides the selected articles

by their publication date. The ﬁrst column shows arti-

cles published before the year 2009, the second shows

those published between 2009 and 2011 and the last

shows those published since 2012.

According to the data shown in Fig. 2, the number

of evaluation research papers is low in comparison to

the number of solution proposals up until 2009. In

this period, estimation cost for GSD projects was a

relatively new subject that needed more investigation

and exploration. From 2012 on, the focus shifted to

the validation and evaluation of existing software cost

estimation methods for GSD projects.

3.5 MQ.4: Research Approach

Four of the selected papers are solution proposals.

Two of them were validated using experiments, while

the other two were not empirically validated. Five

out of 10 of the selected evaluation research studies

were based on theoretical approaches and 4 out of 10

were based on case studies, while not a single arti-

cle was based on an experiment. This situation may

result from the difﬁculty involved in observing the ef-

fects of methods on software cost estimation under

controlled conditions, particularly in the case of dis-

tributed projects in the GSD context. More details on

MQ.3 and MQ.4 are provided in Table 6.

Table 6: Research Types and Approaches.

Case

study

Experiment Survey Theory

Evaluation 4 0 1 5

Solution 1 1 0 2

Other 1 0 0 1

3.6 MQ.5: Contribution Types

Fig. 3 presents the distribution of the selected studies’

contribution types. Thirty eight percent of the cate-

gories are models. Techniques (Lotlikar et al., 2008)

Figure 3: Contribution types.

for the cost estimation of GSD represent 50% of the

selected studies’ types. These models gather data

mining techniques used to help researchers and soft-

ware companies establish data by providing a number

of algorithms and methods with which to deal with

software cost and effort challenges. Each of these data

mining techniques analyses data and provides results

with which to best match the software cost of a GSD

project. The third partition ”Other” represents only

twelve percent split between processes and tools.

3.7 MQ.6: Software Cost Estimation

Techniques

Several software estimation techniques can success-

fully be used to estimate costs. non-machine learn-

ing models (7 out of 16 selected studies) include CO-

COMO (II) (Keil et al., 2006), SLIM (Muhairat et al.,

2010), Use Case Points (Azzeh, 2013) and Function

Points (Peixoto et al., 2010). The use of expert judg-

ment (3 out of 16 selected studies) consists of asking

the opinion of multiple experts who use their experi-

ence and knowledge of the project to provide an es-

timation of the cost. An objective estimation is se-

cured by obtaining as many values as possible from

different experts. Indeed, the objective of the Delphi

technique (Peixoto et al., 2010) is to repeat the estima-

tion process until an agreement is established. Table 7

presents details of the cost estimation techniques used

for GSD projects.

3.8 MQ.7: Software Cost Estimation

Activities

About 80% of the selected studies discuss the soft-

ware development costs of GSD projects. These stud-

ies show the strong link between the software life cost

and its development phase. The second most frequent

topic after development cost is contained in the few

SoftwareCostEstimationforGlobalSoftwareDevelopment-ASystematicMapandReviewStudy

201

Table 7: Estimation techniques for GSD projects.

Type technique Estimation techniques Papers Percentage

non-machine

learning models

COCOMO(2) (Nassif et al., 2012), (Ramasubbu and Balan, 2012), (Azzeh,

2013), (Muhairat et al., 2010), (Keil et al., 2006), (Lamers-

dorf et al., 2010)

85,7%

SLIM (Nassif et al., 2012), (Muhairat et al., 2010) 28,5%

Function points (Nassif et al., 2012), (Peixoto et al., 2010) 28,5%

Use case points (Nassif et al., 2012), (Azzeh, 2013), (Peixoto et al., 2010) 42,8%

Multiple Linear Re-

gression

(Nassif et al., 2012) 14,2%

Expert judgment Delphi (Nassif et al., 2012), (Peixoto et al., 2010) 28,5%

ISBSG (Nassif et al., 2012), (Muhairat et al., 2010) 28,5%

Planning Pocker (Peixoto et al., 2010) 14,2%

Epert Judgement (Peixoto et al., 2010) 14,2%

machine Artiﬁcial Intelligence (Humayun and Gang, 2012) 14,2%

learning models Case-based reasoning (Hamdan et al., 2006), (Ramasubbu and Balan, 2012) 28,5%

Regression trees (Humayun and Gang, 2012) 14,2%

Neural Network (Nassif et al., 2012) 14,2%

Genetic Algorithm (Humayun and Gang, 2012) 14,2%

studies that concentrate on the maintenance phase.

This phase focuses principally on the extraction and

consideration of factors that affect software mainte-

nance. If the software maintenance cost is to be prop-

erly applied, it is essential to estimate the cost and

reduce it by controlling certain factors.

Table 8 summarises the elements that affect soft-

ware cost development and maintenance in the GSD

context. Software development and maintenance are

the major issues to affect a GSD project. The study

is based on the analysis of data collected from se-

lected papers. Software activities have been shown

to have signiﬁcant costs. The development cost re-

sulting from the overall estimate and the estimation

of the beneﬁts of strategies and the networking re-

main highly uncertain and open to improvement, as

do the costs incurred as the results of maintenance,

particularly modiﬁcation, improvement or enhance-

ment along with reengineering costs. These costs

have been known to erode whatever beneﬁts the GSD

model may provide.

3.9 MQ.8: Software Cost Drivers

There are 16 cost drivers, which are divided into the

four categories depicted in Table 9: Product, Plat-

form, Personnel, and Project Factors. As this table,

shows some cost drivers are common to all types of

software projects while others are speciﬁc to GSD

projects. The majority of cost drivers that impact

on GSD projects are related to factors in distributed

software projects: time zone, language (communica-

tion) and cultural differences (team culture). Note that

the most frequently used cost drivers are project ef-

fort (44,4%), process model (33,3%) and time zone

(22,2%).

In order to establish trust in distributed projects,

researchers recommend bringing about cultural un-

derstanding, creditability, capabilities, pilot project

performance, personal visits and investments in the

ﬁeld of GSD. These studies also suggest cultural un-

derstanding, capabilities, contract conformance, qual-

ity, timely delivery, development processes, manag-

ing expectations, personal relationships and perfor-

mance as the key factors for better achievement par-

ticularly as regards good communication.

3.10 MQ.9: Software Cost

Performances

The main reason for studying cost performances in

the GSD context is to reduce costs. Five different cost

performance variables were included to quantitatively

characterise GSD projects: Distributed work, client

control and behavior, project team, project methodol-

ogy and technology variables (Ramasubbu and Balan,

2012).

Cost performance is principally evaluated in three

ways, as can be seen in Table 10 (Ramasubbu and

Balan, 2012). The direction of distributed develop-

ment as regards cost performance is decided by the

direction of the methods in a statistical test model cre-

ated using quantitative data or grounded conclusions

from qualitative data obtained from primary studies.

In summary, only 3 studies provide estimates derived

from empirical data obtained from cost performance

methods applied in different projects of different com-

panies.

4 THREATS TO VALIDITY

The results of this SMS may have been inﬂuenced by

ENASE2015-10thInternationalConferenceonEvaluationofNovelSoftwareApproachestoSoftwareEngineering

202

Table 8: Software cost estimation activities.

Activities Elements Descriptions Papers

Software

development

cost

Development

budget

Costs resulting from the overall estimate

to software development

(Nassif et al., 2012), (Azzeh, 2013)

Software

improvement

Costs resulting from estimating the bene-

ﬁts of strategies such as tools, reuse, and

process maturity

(Ramasubbu and Balan, 2012)

Project plan-

ning and

control

cost of schedule and control breakdowns

by component and activity

(Lamersdorf et al., 2010), (Narendra et al.,

2012)

Project con-

straints

Costs resulting from the networking, com-

munications, delay in Response and dif-

ferent Time Zone

(Peixoto et al., 2010), (Muhairat et al., 2010)

Software

maintenance

cost

Corrective

maintenance

Costs resulting from the modiﬁcation of

software into correct issues detected after

initial deployment

(Ramasubbu and Balan, 2012)

Adaptive

maintenance

Costs resulting from the modiﬁcation of a

software solution to help it stay effective

in a changing business environment

(Ramasubbu and Balan, 2012)

Perfective

maintenance

Costs resulting from the improvement or

the enhancement of a software solution to

improve overall performance

(Ramasubbu and Balan, 2012)

Reengineering

cost

Enhancements Costs resulting from the sequence innova-

tions

(Forbath et al., 2008)

Table 9: Software cost drivers.

Category Cost drivers Papers

Product Code size (Muhairat et al., 2010)

Reuse (Ramasubbu and Balan, 2012)

Product complexity (Keil et al., 2006)

Platform Design and technology newness (Ramasubbu and Balan, 2012), (Forbath et al., 2008)

Time zone (Narendra et al., 2012), (Lamersdorf et al., 2010)

Platform volatility (Keil et al., 2006)

Personnel Team size (Ramasubbu and Balan, 2012)

Team culture (Azzeh, 2013)

Team trust (Azzeh, 2013), (Muhairat et al., 2010)

Communication (Keil et al., 2006), (Narendra et al., 2012)

Development productivity (Ramasubbu and Balan, 2012), (Lamersdorf et al., 2010)

Project Project effort (Ramasubbu and Balan, 2012), (Peixoto et al., 2010),

(Lamersdorf et al., 2010), (Muhairat et al., 2010)

Project management effort (Ramasubbu and Balan, 2012), (Peixoto et al., 2010),

(Lamersdorf et al., 2010), (Nassif et al., 2012)

Process model (Azzeh, 2013), (Muhairat et al., 2010)

Task allocation (Lamersdorf et al., 2010), (Narendra et al., 2012)

Work Pressure (Muhairat et al., 2010)

Client involvement (Ramasubbu and Balan, 2012)

Work dispersion (Keil et al., 2006), (Ramasubbu and Balan, 2012)

Table 10: Software cost performances.

Evaluation

type

Cost performance

method

papers

Baseline

comparison

Historical project

databases

(Muhairat et al.,

2010)

Variation re-

duction

MRE , Prediction

level

(Ramasubbu and

Balan, 2012)

sensitivity

analysis

CCNN (Nassif et al.,

2012)

the coverage of the study search, bias in study selec-

tion, and inaccuracy in study data extraction. Four

types of threats to the validity (Easterbrook et al.,

2008) of the study results are therefore discussed in

the following subsections.

Construct validity is concerned with the exacti-

tude of the interpretation of the concepts studied and

the completeness of the relevant studies collected. In

this mapping study, the key concepts under consider-

ation are contributions towards software cost estima-

tion for GSD projects. To ensure the correct interpre-

tation of these key concepts, we veriﬁed the deﬁni-

tions of the concepts in related literature and all the

SoftwareCostEstimationforGlobalSoftwareDevelopment-ASystematicMapandReviewStudy

203

authors discussed these deﬁnitions in order to reach a

consensus as to their understanding of them.

Internal validity is concerned with the analysis of

the data extracted. The threats to internal validity

are minimal considering that only descriptive statis-

tics were used during the data analysis in this SMS.

Conclusion validity is concerned with the search

terms used in the automatic search and the search

sources are presented in order to make the results of

this mapping study reproducible.

External validity is concerned with the represen-

tativeness of the selected studies as regard the overall

goal of the mapping study. The results of this map-

ping study were considered with regard to the soft-

ware cost estimation for distributed projects. These

results and representative venues can serve as a start-

ing point for researchers and practitioners working in

this ﬁeld.

5 DISCUSSION

This mapping study indicates that the application of

software cost estimation techniques for GSD projects

is a fairly immature area in both research and prac-

tice. First, about two thirds of the selected studies (11

studies out of 16) were published at conferences and

workshops, while only 31.2% (5 out of 16) of the se-

lected studies attained the maturity needed to be pub-

lished in a journal. Furthermore, only one of the se-

lected studies (Ramasubbu and Balan, 2012) attained

a very high quality level (i.e., evidence obtained from

QA).

The fact that the number of selected studies in-

creased over the last decade shows that the application

of software cost estimation knowledge is receiving in-

creasing attention from the software research commu-

nity. The selected studies were published at 12 differ-

ent venues, indicating that extensive attention is being

paid to this study topic by researchers with a broad

range of different research interests in software cost

estimation. All of the above indicates that this study

topic is likely to remain attractive. However we would

urge the research community to strive for high-level

evidence in future studies. The results of this SMS

also highlight a number of implications for further re-

search in the ﬁeld:

(1) Challenges associated with software cost and ef-

fort estimation in GSD are not new. One of the main

reasons for the growth in GSD is the cost of reduc-

ing software development, and effort estimation is a

key component of this cost. Good effort estimation

is thus important for the success of any GSD project.

The results of this mapping study show the need for

more research into techniques that can be used to im-

prove software cost estimation analysis. An adapta-

tion of techniques and models that takes into account

the challenges and factors associated with GSD must

also be investigated.

(2) This mapping study also shows that the applica-

tion of the knowledge recovery approach in various

forms needs to be explored seriously. In many soft-

ware cost estimation cases, practitioners need to re-

cover the knowledge about software characteristics,

especially when developing or maintaining a global

software project that is not well described and docu-

mented. But little work has been done on the applica-

tion of knowledge recovery in software cost estima-

tion activities for GSD.

(3) The quantiﬁcation of the cost drivers’ impact on

productivity implies a high degree of objectivity and

accuracy. However, concepts such as the impact of

communication or team trust and team culture on pro-

ductivity are very difﬁcult to quantify, and the re-

sults should be treated with care. This is owing to

the complexity and unpredictability of personnel be-

haviour which has the greatest impact on estimation

costs, particularly in distributed development.

6 CONCLUSIONS AND FUTURE

WORK

This paper presents the outcomes of an SMS of cost

estimation in the context of GSD projects, in order

to serve both research and practice. This SMS has

shown a wide spectrum of software estimation tech-

niques, activities and cost drivers for GSD projects.

Most of the selected studies present cost contribu-

tion as regards cultural, language and time zone dif-

ferences, which are directly related to making the

achievement of globally performed software projects

more stimulating.

Upon considering the lack of primary studies

identiﬁed in this SMS, we believe that further research

is required into the approaches used in the GSD con-

text. We are also of the opinion that the adaptation

of those techniques based on the speciﬁc aspects of

GSD, in addition to the inherent uncertainty of the

data, could provide more faithful estimates of effort.

The globally distributed environment implies many

challenges and elements. The GSD sourcing strat-

egy and cost estimation process topology could have

a great inﬂuence on cost estimates. Future research

should therefore be carried out to explore how these

challenges and factors affect cost estimation tech-

niques.

ENASE2015-10thInternationalConferenceonEvaluationofNovelSoftwareApproachestoSoftwareEngineering

204

ACKNOWLEDGEMENTS

This work has been supported by the Erasmus

Mundus program Action 2: EU Mare Nostrum UE-

MARE NOSTRUM (204195-EM-1-2011-1-ES-ERA

MUNDUS-EMA21) and the project ‘DataMining for

Software Project Management’ ﬁnanced by the Uni-

versity Mohammed V of Rabat. This work is also par-

tially supported by the GEODAS-REQ project (Span-

ish Ministry of Economy and Competitiveness and

European Fund for Regional Development, TIN2012-

37493-C03-01)

REFERENCES

Azzeh, M. (2013). Software cost estimation based on use

case points for global software development. In Com-

puter Science and Information Technology of the 5th

International Conference, pages 214–218.

Boehm, B., Abts, C., and Chulani, S. (2000). Software de-

velopment cost estimation approachesa survey. An-

nals of Software Engineering, 10(1-4):177–205.

Boehm, B. W. (1981). Software engineering economics.

Upper Saddle River, NJ: Prentice-Hall.

Britto, R., Freitas, V., Mendes, E., and Usman, M. (2014).

Effort estimation in global software development: A

systematic literature review. In Proceedings of the 9th

IEEE International Conference on Global Software

Engineering (ICGSE), pages 135–144.

da Silva, F. Q., Costa, C., Franc¸a, A. C. C., and Prikla-

dinicki, R. (2010). Challenges and solutions in dis-

tributed software development project management:

A systematic literature review. In Proceedings of the

5th IEEE International Conference on Global Soft-

ware Engineering (ICGSE), pages 87–96. IEEE.

Easterbrook, S., Singer, J., Storey, M.-A., and Damian, D.

(2008). Selecting empirical methods for software en-

gineering research. In Guide to advanced empirical

software engineering, pages 285–311.

Forbath, T., Brooks, P., and Dass, A. (2008). Beyond

cost reduction: Using collaboration to increase inno-

vation in global software development projects. In

Proceedings of the 3rd IEEE International Conference

on Global Software Engineering (ICGSE), pages 205–

209.

Hamdan, K., El Khatib, H., Moses, J., and Smith, P. (2006).

A software cost ontology system for assisting estima-

tion of software project effort for use with case-based

reasoning. In Innovations in Information Technology,

2006, pages 1–5.

Hughes, R. T. (1996). Expert judgement as an estimat-

ing method. Information and Software Technology,

38(2):67–75.

Humayun, M. and Gang, C. (2012). Estimating effort in

global software development projects using machine

learning techniques. International Journal of Infor-

mation and Education Technology, 2(3):208–211.

Idri, A., Azzahra Amazal, F., and Abran, A. (2015).

Analogy-based software development effort estima-

tion: A systematic mapping and review. Information

and Software Technology, 58(0):206–230.

Idri, A., Zahi, A., and Abran, A. (2006). Software cost es-

timation by fuzzy analogy for web hypermedia appli-

cations. In Proceedings of the International Confer-

ence on Software Process and Product Measurement,

Cadiz, Spain, pages 53–62. Citeseer.

Jørgensen, M. (2004). A review of studies on expert es-

timation of software development effort. Journal of

Systems and Software, 70(1):37–60.

Jorgensen, M. and Shepperd, M. (2007). A systematic re-

view of software development cost estimation stud-

ies. Software Engineering, IEEE Transactions on,

33(1):33–53.

Keil, P., Paulish, D. J., and Sangwan, R. S. (2006).

Cost estimation for global software development. In

Proceedings of the 2006 International Workshop on

Economics Driven Software Engineering Research

(EDSER), pages 7–10.

Kitchenham, B. A. and Charters, S. (2007). Guidelines for

performing systematic literature reviews in software

engineering. Technical report, Software Engineering

Group, Keele University and Department of Computer

Science University of Durham.

Lamersdorf, A., Munch, J., Torre, A. F.-d. V., Sanchez,

C. R., and Rombach, D. (2010). Estimating the ef-

fort overhead in global software development. In Pro-

ceedings of the 5th IEEE International Conference on

Global Software Engineering (ICGSE), pages 267–

276.

Lotlikar, R. M., Polavarapu, R., Sharma, S., and Srivastava,

B. (2008). Towards effective project management

across multiple projects with distributed performing

centers. In Proceedings of the IEEE International

Conference on Services Computing (SCC), volume 1,

pages 33–40.

Muhairat, M., Aldaajeh, S., and Al-Qutaish, R. E. (2010).

The impact of global software development factors on

effort estimation methods. European Journal of Sci-

entiﬁc Research, 46(2):221–232.

Narendra, N. C., Ponnalagu, K., Zhou, N., and Gifford,

W. M. (2012). Towards a formal model for opti-

mal task-site allocation and effort estimation in global

software development. In Proceedings of the Annual

SRII Global Conference, pages 470–477.

Nassif, A. B., Capretz, L. F., and Ho, D. (2012). Software

effort estimation in the early stages of the software

life cycle using a cascade correlation neural network

model. In Proceedings of the 13th International Con-

ference on Software Engineering (ACIS), Artiﬁcial In-

telligence, Networking and Parallel/Distributed Com-

puting, pages 589–594.

Ouhbi, S., Idri, A., Fern

andez-Alem

an, J. L., and Toval, A.

(2013a). Requirements engineering education: a sys-

tematic mapping study. Requirements Engineering,

pages 1–20.

Ouhbi, S., Idri, A., Fern

andez-Alem

an, J. L., and Toval, A.

(2013b). Software quality requirements: a system-

atic mapping study. In Software Engineering Con-

SoftwareCostEstimationforGlobalSoftwareDevelopment-ASystematicMapandReviewStudy

205

ference (APSEC, 2013 20th Asia-Paciﬁc, volume 1,

pages 231–238. IEEE.

Peixoto, C. E. L., Audy, J. L. N., and Prikladnicki, R.

(2010). Effort estimation in global software develop-

ment projects: Preliminary results from a survey. In

Proceedings of the 5th IEEE International Conference

on Global Software Engineering (ICGSE), pages 123–

127.

Ramasubbu, N. and Balan, R. K. (2012). Overcoming the

challenges in cost estimation for distributed software

projects. In Proceedings of the 34th International

Conference on Software Engineering, pages 91–101.

Wen, J., Li, S., Lin, Z., Hu, Y., and Huang, C. (2012). Sys-

tematic literature review of machine learning based

software development effort estimation models. In-

formation and Software Technology, 54(1):41–59.

ENASE2015-10thInternationalConferenceonEvaluationofNovelSoftwareApproachestoSoftwareEngineering

206