CRITICAL SUCCESS FACTORS OF OFFSHORE SOFTWARE DEVELOPMENT PROJECTS

Index

Objectives
Methodology
Research Process

In the following sections, we first present the research objectives of the chapter at hand. Next, we introduce the research methodology applied and outline the research process implemented in order to reach the defined goals.

The majority of CSF studies are limited to the identification and the description of CSF (Esteves, 2004). Within the conducted literature review on CSF research in the field of IT outsourcing (compare section 5.2), we were able to identify only one CSF research paper (Adelakun and Jennex, 2003) which also analyzes the relevance of the presented CSF. In contrast, with regard to a more advanced CSF analysis, we do not know of any studies. The research project at hand aims to close this gap. Therefore, the main objective of this chapter is to create a deeper understanding in regard to the relevance, the phase specificity, and the level of influence of the identified 29 CSF for OSD projects (compare chapter 5). In an effort to reach this goal, particularly the following research questions will be addressed: “Which of the identified CSF are particularly relevant in which OSD project contexts, which of these CSF are specific for which OSD project phases, and in which time frame can these CSF be influenced?”

According to Esteves (2004), researchers apply both qualitative and quantitative research methods, and here, primarily case studies and surveys (both based on interviews), to analyze the relevance of CSF. In this context, the interview partners are usually asked to compile a list of the most relevant CSF or to assess the relevance of each individual CSF by means of a Likert scale.

In an effort to analyze the identified CSF in more detail, we decided to apply a predominantly quantitative research design. Crucial for this decision was the fact that we aimed to numerically measure the collected data and to employ statistical procedures (Creswell, 2003).

As a quantitative research method, we decided to conduct an online survey (also referred to as web survey). The reasons for using an online survey were that it was the easiest, fastest, and least expensive way to access experts in the field of OSD (Esteves, 2004). However, beside these advantages of an online survey compared to a classical survey (using a “paper-pencil-approach”), the technique possesses the disadvantages that a systematic selection of the participants is not or rather conditionally possible and statements on “non-participants” cannot be made (Hauptmanns, 1999).

In the process of designing the questionnaire, we further decided to exclusively formulate close-ended questions. This can be reasoned by the fact that we did not aim to identify additional CSF, but focused on the quantitative examination of the already identified CSF.

In literature, several different guidelines on how to design an online questionnaire are discussed. However, according to Krcmar, Leimeister, and Sidiras (2004), all of these guidelines (e. g., Büning, Haedrich, Kleinert, and Kuß, 1981) include the following three basic principles: Accuracy, neutrality, and simplicity. These principles also served as design criteria for the preparation of our online questionnaire. In an effort to comply with these criteria, we paid particular attention to the formulation of clear and non-redundant questions (accuracy), the ensuring of the impartiality of the participants, e. g., by listing the CSF in alphabetical order, (neutrality), and the logical structuring of the questions (simplicity).

Regarding the research process, the following table outlines the four research steps implemented.

Table: Description of research steps

In the following, the research steps mentioned above will be described in more detail.

Literature Review

On the basis of a comprehensive literature review, we were able to identify 15 CSF studies in the field of IT outsourcing. The analysis of these studies provided us with first insights in regard to the relevance, the phase specificity, and the level of influence of our list of 29 CSF. These insights particularly influenced the determination of the concrete analysis aspects as well as the formulation of the survey questions.

Preparation of the Online Survey

In the style of the interview guidelines used for the identification of the 29 CSF , the online questionnaire was divided into two parts: While the first part focused on the quantitative analysis of the identified CSF, the second part aimed to collect general information on the participants. With regard to the second part, the implementation of an Internet-based survey enabled an automatical adaptation of questions depending on the information provided by the participants. For instance, dependent on the specified company perspective (client, provider, or consultancy), the number and the formulation of the subsequent questions varied.

In regard to the formulation of both parts of the questionnaire, we first proposed questions and answer options, and refined these in an iterative process. Second, we tested the online survey with 20 test persons. Based on the feedback of these persons, we further refined the survey. Here, we paid particular attention to the adequate usage of mandatory questions.

Within the test phase, we also examined the capability of the underlying hard- and software infrastructure. In addition, we already tested the data import to MS Excel and the data evaluation with SPSS in this phase. The design of the online survey was carried out with support of the software tool “PHP Surveyor”, which was already successfully deployed in numerous research projects by our department.

After having completed the test phase, we announced the online survey within the monthly online newsletter of the German Association of Information Economy, Telecommunication, and New Media (BITKOM). Further, we published a press release in different media, hereby requesting representatives of German-speaking companies to participate in the research project.

Data Collection

Our research focus on OSD projects required a careful selection of potential participants. In this context, we made contact with company representatives, possessing OSD experience, via the Internet communication platform openBC (http://www.openbc.com). By means of this platform, we were able to identify as well as directly contact 247 persons, working in the field of OSD. In addition, we sent e-mails to 813 German-speaking companies, which were listed in various company directories (e. g., http://www.firmenregister.de), and got in touch with 161 medium-sized and large-scale enterprises via phone and e-mail.

All contacted companies and persons were invited to participate in the online survey, provided that these companies or persons had already implemented OSD projects. For this purpose, the online survey was made available on the Internet pages of the University of Erlangen-Nuremberg over a period of two months (October and November 2005).

Data Analysis

In total, 103 persons participated in the online survey. Due to the comprehensive usage of mandatory questions within the online survey, all of the 103 data records were able to be included within the data analysis process. This process was carried out with the help of SPSS.

The following table gives an overview of the three fundamental analysis aspects of the online survey as well as the corresponding assessment values and assignment classes respectively.

Table: Overview of analysis aspects

Regarding the analysis of the relevance of the identified CSF, we compiled an overall CSF ranking on the basis of the calculation of the arithmetic mean for each individual CSF. Corresponding CSF rankings were developed for both the entirety of the participants and individual participant groups (e. g., participants working for OSD clients). In addition, we analyzed the significance of group-specific assessment differences within different analysis dimensions (e. g., differences between participants working for OSD clients, providers, and consultancies within the dimension “company perspective”).

In an effort to examine the statistical significance of identified assessment differences, dependent on the number of dimension values of a specific dimension, we either conducted a t-test (in the case of two values) or an ANOVA-test (in the case of three or more values) with an alpha value of 0.05. That means that an alpha error (that is, the identification of a difference within the sample which does not exist in reality) occurs with a probability of 5 percent. To perform the mentioned tests, we defined the following difference hypothesis: “The arithmetic means of the CSF differ between the individual dimension values.” The corresponding null hypothesis reads as follows: “The arithmetic means of the CSF do not differ between the individual dimension values.”

According to Kähler (2004), parametric test procedures (e. g., variance analyses) assume normal variables. Therefore, before implementing the t-test and the ANOVA-test respectively, we verified the normal distribution of the variables by means of histograms.

With regard to the analysis of the phase specificity and the level of influence of the 29 CSF, we were not able to calculate the arithmetic means due to the usage of nominal assessment scales. Instead, we compiled frequency tables which indicated how often each CSF was assigned to a specific project phase or level of influence respectively. Corresponding frequency tables were again created for both the entirety of the participants and individual participant groups.

References