UNCORRECTED PROOFS. A State-of-the-Practice Study on Communication and Coordination between Chinese Software Suppliers and Their Global Outsourcers

Transcription

1 AQ SOFTWARE PROCESS IMPROVEMENT AND PRACTICE Softw. Process Improve. Pract. ; : Published online in Wiley InterScience ( DOI:./spip. A State-of-the-Practice Study on Communication and Coordination between Chinese Software Suppliers and Their Global Outsourcers Jianqiang Ma, * Jingyue Li, Weibing Chen, Reidar Conradi, Junzhong Ji and Chunnian Liu Department of Computer and Information Science, Norwegian University of Science and Technology (NTNU), NO- Trondheim, Norway Beijing Municipal Key Laboratory of Multimedia and Intelligent Software Technology, College of Computer Science and Technology, Beijing University of Technology, Beijing, China SPIP Research Section Most studies of software outsourcing focus on how to manage outsourcing from the perspective of the outsourcer. There are few studies of outsourcing presented from the viewpoint of the supplier. Given that more and more Chinese software companies are getting software outsourcing subcontracts from all over the world, it is important to understand how software outsourcing projects are actually performed in China, and to identify ways in which performance might be improved. Our study first collected data through a questionnaire-based survey from finished projects in Chinese software suppliers. A follow-up study was performed by telephone interviews with participants, to find explanations for phenomena revealed by the preceding survey. The results show that: (a) differences in natural languages are not a real obstacle for Chinese software suppliers; (b) is most used to discuss issues related to development, while face-to-face meetings are mainly used to discuss issues related to management and requirements; and (c) the two main reasons for overtime work at the Chinese suppliers are changes to the design or requirements that are proposed by the outsourcers and the suppliers initial underestimation of the effort required. Copyright John Wiley & Sons, Ltd. KEY WORDS: software outsourcing; empirical study; software process improvement Correspondence to: Jianqiang Ma, Department of Computer and Information Science, Norwegian University of Science and Technology (NTNU), NO- Trondheim, Norway majian@idi.ntnu.no Copyright John Wiley & Sons, Ltd.. INTRODUCTION Software outsourcing is becoming increasingly attractive to software development companies, because it helps them to save costs and to focus on their core businesses (Donahoe and Pecht, Ahmed ). Along with her policy of opening to

2 J. Ma et al. the outside world, China is becoming one of the biggest suppliers of software, together with India, Ireland, and Russia, among others. Although the volume of Chinese software exports in was $ million, which is a fivefold increase over the value, this is still only one sixth of India s and.% of the total volume of Chinese exports (Ministry of Information of the People s Republic of China & Chinese Software Industry Association, Wong ). Compared to in-house software development, there are several factors that may affect the effectiveness of software outsourcing, such as infrequent and inefficient communication between decentralized teams (Ferguson et al. ), differences in time zones (Carmel and Agarwal ), and cultural differences (Walsham, Kankanhalli et al., Krishna et al. ). Many studies have been performed on these issues, but mainly from the outsourcers perspective. Since China is emerging as a major player in software outsourcing, it is important to study such issues and develop guidelines for outsourcing, in order to help Chinese software suppliers to improve their businesses. We conducted a survey to investigate three software engineering issues among Chinese software suppliers: differences in natural language, efficient communication, and overtime work. We used membership lists from a national Chinese Software Organization (henceforth, CSO )toachievearepresentative subset of software companies. The survey was performed in four steps. First, we designed a preliminary questionnaire, using a review of the literature as a basis. Second, we conducted a prestudy that included individual interviews and a workshop discussion to examine the questionnaire. Third, we used a questionnaire-based survey to investigate the three issues. Data from finished projects of Chinese suppliers were collected. Fourth, we performed a follow-up study by calling back participants to identify the reasons for the phenomena that we discovered in our analysis. The study described here is novel in that it is the first state-of-the-practice study of issues pertaining to communication and coordination from the suppliers perspective in China. The results reveal how Chinese software suppliers communicate and coordinate with their global outsourcers, and indicate The name of this organization was omitted for reasons of confidentiality. areas in which these processes could be improved. Using the results of this study as a basis, we propose three strategic recommendations to help Chinese software suppliers to facilitate their collaborations with global outsourcers. The remainder of this article is structured as follows: Section presents the related work and research questions. Section describes the research design. Section shows the profile of our study participants. Section presents results and discussions. Section contains a general discussion. Conclusions and ideas for future work are presented in Section.. RELATED WORK AND RESEARCH QUESTIONS There are two main participants in software outsourcing: the outsourcer and the supplier. Software outsourcers are the organizations that award software development subcontracts to other organizations, which are called the suppliers. Most previous studies focus on managing the software outsourcing from the outsourcers point of view, such as how to manage cultural differences (Walsham, Kankanhalli et al., Krishna et al. ), how to evaluate and select capable suppliers (Lacity ), and how to manage issues related to contracts (Aubert et al. ). However, only a few studies (Smite, Taylor ) have focused on facilitating the efficiency of software outsourcing from the suppliers perspective, such as how to communicate and cooperate with outsourcers. Given that China is one of the largest and fastestgrowing software suppliers in the world, it is a matter of great importance to investigate how Chinese software suppliers can improve their practices to better communicate and cooperate with software outsourcers all over the world. The possible lack of skills in English may cause problems for Chinese software suppliers when they communicate with outsourcers, from other countries, who use English as their predominant means of international communication (Kankanhalli et al. ). Given that it is much more difficult to follow the plans in projects across organizational and geographic boundaries than in projects within the same place and organization (Herbsleb et al. ), the boundaries between China and other countries may also introduce addi-tional project overruns (Herbsleb and Grinter, Smite ). Copyright John Wiley & Sons, Ltd. Softw. Process Improve. Pract., ; : DOI:./spip

3 Chinese Software Suppliers and Global Outsourcers.. Differences in Natural Languages In software outsourcing projects, the natural language that is used in the documents is an important factor, because language differences may cause misunderstandings. For instance, the word more than in Japanese means equal to and more than (> =), but in Chinese it means strictly more than (>). Such misunderstandings in requirements specifications or other development documents may, in worst-case scenarios, lead to project overruns. In a previous study, Carmel and Agarwal () proposed that the language factor is one of the reasons for the success of software outsourcing in countries with strong English language capabilities, such as the Philippines and Singapore. Kshetri () concluded that the major Chinese market for software outsourcing is Japan, because the cultures of the two countries are similar. By contrast, India s major markets are the US and UK because of a shared English language. So, the biggest barrier to Chinese companies supplying software to Western companies seems to be the English language. However, Japan and China also speak different natural languages, but the outsourcing projects between them seem to work well. So our first research question is as follows: RQ: Are differences in natural languages a barrier to Chinese software suppliers?.. Efficient Communication Compared with in-house software development, the communication in software outsourcing is affected by both spatial and temporal distance. Carmel and Agarwal () concluded that communicating synchronously, such as by holding telephone meetings or video conferences, can resolve misunderstandings and small problems before they get out of hand. By contrast, asynchronous communication, such as , often delays or complicates the resolution of problems. However, the results of another study by Smite () show that is used much more than synchronous channels in a Latvian subcontractor company. To improve suppliers current practices, it is important to know which communication channels are usually used between foreign outsourcers and Chinese suppliers and what subjects are discussed. So our second research question is as follows: RQ: What are the most common channels of communication between foreign outsourcers and Chinese suppliers?.. Overtime Work According to a survey by Moløkken and Jørgensen (), around % software development encounters effort and/or schedule overruns. The average effort overrun is about %. Schedule and budget risks are the most highlighted risk factors in software outsourcing (Taylor ). Compared with in-house development, factors such as time-zone differences and organizational differences may also cause project overruns (Smite ). In order to resolve the problem of overrun and to keep the project schedule, overtime work is often used as a remedy. However, most Chinese software suppliers get their contracts from Japan, which has only a one-hour time difference with China. Hence, it would seem that time-zone difference may not be the major reason for Chinese suppliers working overtime. We wonder whether there are other common causes of project overrun. So, our third research question is: RQ: Do Chinese suppliers work overtime and, if so, what are the main reasons for that?. RESEARCH DESIGN To answer the research questions, we used a questionnaire-based survey that was followed up by telephone interviews... Step : The Preliminary Questionnaire The preliminary questionnaire was designed to study seven issues of software outsourcing from the suppliers perspective and had ten sections. Two sections contained questions to collect background information of projects, companies, and respondents. Questions in three sections are related to the research questions in this article. The remaining sections investigate four issues, which are related to technology management (e.g. how is the source code integrated and updated) and business relationship management (e.g. how did the outsourcer and supplier get to know each other), which are not reported in this article. Copyright John Wiley & Sons, Ltd. Softw. Process Improve. Pract., ; : DOI:./spip

4 J. Ma et al... Step : The Pre-study to Verify and Refine the Preliminary Questionnaire The pre-study included two substeps.... Individual Interviews We first interviewed five project managers from five different software suppliers. All the interviewees had solid experiences in software outsourcing. Each interview was conducted by two authors of this article. One was responsible for conducting the interview and the other recorded answers and asked for clarification if needed. The interviews were taped for later use.... A Workshop Discussion After the individual interviews, we revised most of the open questions in the preliminary questionnaire into closed questions and made a second version of the preliminary questionnaire. We then organized a workshop with more than industrial experts to comment on the second version of the questionnaire. Using the input from the workshop as a basis, we revised the questionnaire into a final version. The final questionnaire contained questions and took about minutes to complete... Step : The Main Study to Collect Data In the main study, the data was collected by cooperating with the CSO. As mentioned above, we received completed questionnaires from companies. We selected the sample and collected data based on the following.... Establish the Target Population We selected companies at random from a database of the CSO, which included about Chinese software companies.... Send Invitation Letters by to Obtain Possible Participants We sent invitation letters by to the selected companies. The invitation letter introduced the survey and specified that participants in the survey would be rewarded with either a copy of the final report of the survey or a one-year membership of the CSO worth Chinese Yuan. We received responses from software outsourcing companies. These companies were later used as the original contact list.... Send Questionnaires by to Possible Participants The unit of our study is defined as a finished software outsourcing project. We sent questionnaires (as MS Word files) by to the companies and asked them to select one or more projects to complete the questionnaire. It was not possible to obtain a full list of relevant projects in each company; hence, the respondents themselves selected the projects from their company s portfolio, i.e. a convenience sample.... Collect Completed Questionnaires with Follow-Up From the companies, we received completed questionnaires. To ensure the quality of the data, we excluded questionnaires, which had been completed by respondents who had less than three years work experience. For the remaining questionnaires, we contacted the respondents again by telephone to clarify possible misunderstandings and to fill in the missing data. At the same time, we contacted the remaining companies by telephone to try to persuade them to complete the questionnaire. By doing this, we got a further completed questionnaires... Step : The Follow-up Study We analyzed the data and designed seven followup research questions. To obtain answers to these questions, we conducted telephone interviews by calling selected participants of both the pre-study and the main study. All five project managers who participated in the pre-study interviews and participants of the main study had provided their telephone numbers. These persons were used as the initial contact lists. We sent the results of the main study to these potential respondents before the telephone interviews, in order to prepare them for answering further questions. Then, the first author of this article called each of these persons and asked follow-up questions. Each telephone interview lasted minutes, on average.. PARTICIPANTS OF THE STUDY We now present an overview of the participants in the main study and in the follow-up study. AQ Copyright John Wiley & Sons, Ltd. Softw. Process Improve. Pract., ; : DOI:./spip

5 Chinese Software Suppliers and Global Outsourcers % % % % % % % % % % % % Figure. The distribution of companies %.% %.% % Small (<=) Medium (-) Large (-) Super lage (>) Software companies in China.. Overview of Collected Participants in the Main Study... Participating Companies Using the number of employees as the basis for classification, the participating companies includes small, medium-sized, large, and super-large software companies, as shown in Figure. Comparisons with the average number of employees in companies in the Chinese software industry (Ministry of Information of the People s Republic of China & Chinese Software Industry Association ) reveal that most of the companies that participated in our survey were medium and large ones.... Respondents Most of the respondents are IT managers, project managers, or software architects. More than % of them have at least five years experience in software development. All of them have at least a Bachelor s degree in computer science or telecommunications.... Participating Projects In total, our survey collected data from finished projects. The mean duration of these projects (before the first delivery) was days, and the mean project effort was persondays. Twenty-one projects were outsourced from US or European companies. One of these projects used French in the documents, while the other used English. Thirty-two projects were outsourced from Japanese companies. One Outsourcing companies we studied of the projects used Chinese in the documents, eight of them used both English and Japanese, and the other projects used Japanese... Overview of Collected Participants in the Follow-up Study In the follow-up study, we interviewed participants by telephone. Three of them were from small companies, five from medium-sized companies, two from large companies, and one from a super-large company. Six of them were project mangers, three were IT managers, and two were senior developers.. RESULTS AND DISCUSSIONS We now present the results of each research question and the corresponding follow-up research questions... Investigating RQ: Are Differences in Natural Languages a Barrier to Chinese Software Suppliers?... Results on RQ from the Main Study We investigated how well the suppliers employees understand the non-native natural languages used in project documents provided by the outsourcers (e.g. business agreements, requirements specifications, and design specifications). The scale Copyright John Wiley & Sons, Ltd. Softw. Process Improve. Pract., ; : DOI:./spip

6 J. Ma et al. (a) (b)..... (c) Master Fluent Competent Read and write Unfamiliar Do no know/ Not answered English skill Japanese skill Master Fluent Competent Read and write Unfamiliar Do no know/ Not answered English skill Japanese skill Master Fluent Competent Read and write Unfamiliar Do no know/ Not answered English skill Japanese skill Figure. The respondents skills in the different natural languages used in documents. (a) Twenty projects outsourced from Western companies using English in the documents. (b) Twenty-three projects outsourced from Japanese companies using Japanese in the documents. (c) Eight projects outsourced from Japanese companies using both English and Japanese in the documents used to measure people s language skills is as follows: master (near native), fluent, competent (can speak, read, and write, but not fluently), read and write only, unfamiliar, and do not know. The results of the RQ are presented in Figures and. Copyright John Wiley & Sons, Ltd. Softw. Process Improve. Pract., ; : DOI:./spip

7 Chinese Software Suppliers and Global Outsourcers AQ Master Fluent Competent Read and write Unfamiliar Do no know/ Not answered English skill Figure. The companies average skills in the natural languages Figure (a) shows that the respondents who worked on projects that were outsourced from American and European companies (with only English in the documents) have a good knowledge of English. However, their Japanese language skills are poor. Figure (b) shows that the respondents who worked on projects that were outsourced from Japanese companies with only Japanese in the documents are good at Japanese. Some of them have solid English skills, while the others have poor English skills. Figure (c) shows that the respondents who worked on projects that were outsourced from Japanese companies with both English and Japanese in the documents have much better skills in English than in Japanese. Figure shows that, on average, the suppliers employees have better language skills in English than in Japanese.... Insight on RQ from the Main Study Kshetri () concluded that The US and the UK are India s major markets, thanks to its English competence. On the other hand, China s cultural similarity to Japan and South Korea has facilitated its software export. However, our results in Figures (c) and show that the employees of the Chinese software supplier Japanese skill companies have, on average, better skills in English than in Japanese. Moreover, the results presented in Figure (b) show that some respondents who worked with documents written in Japanese also have good English language skills. One possible reason is that most employees in the Chinese software supplier companies that we investigated, have a Bachelor s or higher degree. In order to get such degrees, they must pass an English certification called CET- (College English Test band ) or CET- to demonstrate their competence in English. Given that Chinese software suppliers have successfully developed large software systems for Japanese outsourcers, their equivalent or better English skills indicate a big potential for developing software for outsourcers from English-speaking countries as well.... Follow-up Study for RQ Our discussions with the interviewees in the prestudy revealed that some companies are using a steering group, in which language-competent people act as a buffer between developers and outsourcers (as shown in Figure ). The Chinese software suppliers that use steering groups managed to resolve the confusions caused by the possibly poor language skills of the developers. To deepen our understanding of the usage of steering groups, three follow-up questions RQ.f to RQ.f were designed as follows: RQ.f: When should a steering group be organized? RQ.f: How should a steering group be organized? RQ.f: What are the advantages and disadvantages of using a steering group? With respect to RQ.f, the interviews with the follow-up study participants illustrate that Copyright John Wiley & Sons, Ltd. Softw. Process Improve. Pract., ; : DOI:./spip

8 J. Ma et al. Supplier Outsourcer Send documents Steering group Send documents Developers Discuss confusions Discuss confusions Color Figure - Online only - The steering group members get the documents from the outsourcer and try to figure out confusions or possible mistakes with the outsourcer. - The developers get documents from the outsourcer via steering group members - If the developers get confused with the documents from the outsourcer, they will discuss with the steering group members first. - In case developers language confusions cannot be clarified by the steering group members, the steering group members will clarify the matters with the outsourcer on behalf of the developers. Figure. The steering group process Four out of participants in the telephone interviews stated that there is a formally organized steering group in their companies. In three of the four companies (named Group A herein), the steering group is used when a project includes many junior developers and the time-to-market is not critical. When the time-to-market pressure of the project is high, the steering group is used and the project includes only senior developers with sufficient language skills. The remaining company (named Group B herein) is a newcomer to the outsourcing market and few of its developers have sufficient Japanese language skills. Thus, a translator has to be involved with the steering group to translate documents. Six participants stated that there is no formally organized steering group in their companies. Project managers or team leaders in their companies perform tasks that are similar to those of personnel in the steering group. One other participant stated that all employees in his company have sufficient language skills. Thus, the employees of this company usually contact outsourcers directly, without any help from their colleagues. With respect to RQ.f, thefourcompaniesused a slightly different process to organize the steering group. The personnel in the steering groups of Group A companies consist of senior developers with good language skills. When such personnel receive documents from outsourcers, they first analyze and discuss the documents themselves. If there is any confusion as to what is required, they discuss the matter with the outsourcer. After all confusions have been clarified, they translate the documents and send them to the internal developers. To improve the efficiency of the project, documents are translated and software is developedin parallel. That is, as soon as parts of the documents have been translated, those parts are transferred to a developer for technical work. At the same time, the personnel in the steering group will continue to translate the remaining documents. The personnel in the steering group of the Group B company include a professional translator who has little IT knowledge and a senior developer who does not have the language skills to do the work on his own. The translator translates Copyright John Wiley & Sons, Ltd. Softw. Process Improve. Pract., ; : DOI:./spip

9 Chinese Software Suppliers and Global Outsourcers AQ the document first and then gives it to the senior developer for verification. If the translator has misunderstood anything or has translated incorrectly, the senior developer will inform the translator, who will then correct the errors or communicate with the outsourcer. The problem, recognized by the company employees, with using such a process, is that the senior developer or other developers in the company often have to wait for translations from the translator. Further, the translator may sometimes make mistakes when translating technical terms and misleads the developers. The progress of the project is therefore greatly affected. For question RQ.f, all the companies that have a formal steering group gave similar answers. They stated that the advantage of using such a steering group is that the language barriers can, to a certain extent, be eliminated, especially when the projects include many junior developers. Such junior developers usually do not have sufficient language skills or are not familiar with specific terms used by outsourcers. The stated disadvantage of using a steering group is that it may affect the efficiency of the project; especially if it is not well organized.... Conclusion on RQ Differences in natural languages may not be a barrier to Chinese software suppliers for developing software for the English-speaking outsourcers. Having a steering group whose personnel have good English language skills can help to reduce the possible negative effect of supplying clients who speak a different language. Personnel in the steering group should have both sufficient language skills and relevant technical experience. Table. Means of communication and issues for discussion (N =, multiple choice) Asynchronous However, using a steering group may increase project costs... Investigating RQ: What are the Most Common Channels of Communication Between Foreign Outsourcers and Chinese Suppliers?... Results for RQ from the Main Study We listed several possible channels of communication, such as , video conferencing, and faceto-face meetings, which outsourcers and suppliers could use to discuss issues. For each communication channel, we asked whether it had been the principal channel used for specifying, clarifying, or negotiating certain issues, such as requirements, business, or schedule. If the communication channel had been used for more than one issue, the respondents could mark several alternatives. We calculated the number of projects that used a certain communication channel for a specific issue versus the total number of projects (i.e. ). For example, projects mainly used to discuss issues related to requirements. Thus, the result is / or %. A summary of all the results is given in Table. Since the maximum value of a number in Table is %, we classified the values into four categories to simplify the presentation. We excluded values less than %. Values between % and % were replaced with one plus. Values between % and % were replaced with two pluses. Values of more than % were replaced with three pluses. The data in Table show that correspondence was used for almost all issues, such as discussing requirements, harmonizing the schedule, handling changes, development, and testing. Among synchronous communication channels, telephone calls and face-to-face meetings were popular. Synchronous Net meeting Phone meeting Video conf. Face-to-face meeting Requirements Development Testing ++ Maintenance Staff arrangement + Harmonizing schedule Handling changes Business related +++ Copyright John Wiley & Sons, Ltd. Softw. Process Improve. Pract., ; : DOI:./spip

10 J. Ma et al. Face-to-face meetings were mostly used to discuss topics related to requirements and business. Net meetings and video conferences were less used than the other channels.... Insight on RQ from the Main Study Carmel and Agarwal () proposed that synchronous means of communication are more suitable for use in outsourcing contexts than asynchronous means. However, few empirical studies have investigated how synchronous and asynchronous communication channels are used in practice. Our results show that both synchronous (such as telephone calls and face-to-face meetings) and asynchronous (such as ) communication channels have been used by all companies. However, some communication channels were used more than others for resolving particular kinds of issue. The advantage of is that it is much cheaper than synchronous communication channels and is unaffected by separation of space or time. The Chinese software suppliers used to discuss almost everything with the outsourcers. By contrast, face-to-face meetings were used mainly when discussing issues related to requirements and business. Smite performed a case study on one Latvian software supplier (Smite ), which shows that is used much more than other communication channels, while net meeting and video conferencing are almost never used. Our results for Chinese software suppliers show the same trend. However, suppliers and outsourcers may well be based in different time zones and observe different public holidays and these differences may affect their available communication channels. We determined whether or not differences in time zone had any significant effect by dividing the projects into two groups: those with and those without time zone problems (i.e. Japanese versus Western). The two groups show similar results, as shown in Table. Thus, the fact that suppliers and outsourcers may well be based in different time zones and observe different public holidays, does not affect our findings.... Follow-up Study for RQ Although we studied whether or not a given communication channel was used to discuss a specific issue, we received only yes or no answers. A more important issue is how often a certain means of communication was used and how efficient it was. In addition, the data show that both and face-to-face meetings were used to resolve issues related to requirements. A further issue is which communication channel is more cost-efficient, especially for discussing changes to requirements. Thus, we asked follow-up questions RQ.f and RQ.f: RQ.f: What are the most cost-efficient communication channels and how often are they used? RQ.f: What are the most cost-efficient communication channels for discussing changes to requirements? To answer RQ.f, we asked the telephone interviewees which three communication channels they found to be the most efficient and how often they used them. The data are shown in Table. Data in Table show that all companies listed telephone calls as one of the three most efficient communication channels and that they used this means of communication once or twice per week, and Ten out of companies listed as one of the three most efficient communication channels and used it anytime. Six out of companies regard video conferencing as an efficient communication channel. However, only one used it often (twice a week). Four considered face-to-face meetings to be an efficient communication channel. However, none of them used it often. None of our participants regarded net meeting as an efficient channel. With respect to RQ.f, we asked respondents to list the most efficient communication channel for discussing changes to requirements with outsourcers. Six out of respondents considered both s and telephone meetings to be efficient. Three preferred face-to-face meetings and two preferred video conferencing.... Conclusion on RQ is used frequently and is experienced as a cost-efficient means of communication. It is used mainly for handling requirements, all kinds of changes, and for coordinating schedules. Telephone calls are made once or twice per week, and this is also regarded as an efficient communication channel. However, it is rarely used to discuss certain issues, such as testing, maintenance, and staff arrangements. Although face-to-face meetings Copyright John Wiley & Sons, Ltd. Softw. Process Improve. Pract., ; : DOI:./spip

11 Chinese Software Suppliers and Global Outsourcers Table. The three most cost-efficient communication channels and their frequency of usage Companies of the follow-up study Three most cost-efficient communication channels and their frequency of usage AQ (Anytime) Video conferencing Telephone calls (not often) (two times per week) Telephone calls (once per week) (Anytime) Face-to-face meetings (very few) (Anytime) Telephone calls (once per week) Video conf. (very few) Telephone calls (Anytime) (once or twice per week) Face-to-face meetings (very few) Video conferencing (very few) Telephone calls (onceortwiceperweek) Telephone calls (Anytime) Video conferencing (very few) (two to three times per week) Telephone calls Video conferencing (very few) (Anytime) (two to three times per week) Face-to-face meetings (very few) (Anytime) Telephone calls (two times per week) (Anytime) Telephone calls (once per week) Face-to-face meetings (very few) (Anytime) Telephone calls (two times per week Video conferencing (very few) Telephone calls (Anytime) Video conferencing (very few) (once or twice per week) are used for management and requirements issues and are regarded as efficient, they are rarely used. Net meeting and video conferencing are rarely used... Investigating RQ: Do Chinese Suppliers Work Overtime and, if so, What are the Main Reasons for that?... Results on RQ from the Main Study We asked whether there was overtime work in the Chinese software supplier companies. Results show that out of projects had overtime work, usually unpaid. We also asked the respondents about their average number of working hours per week, the percentage of employees in their companies that do overtime, and their outputs during overtime versus their total outputs. The results show that the average numbers of work hours per week are about, which means that the average overtime is hours per week (% more). Since % of the fulltime employees do overtime, the expected relevant outputs from the overtime hours should be % (% of %) of that from normal hours. However, the results show that % of the project outputs come from the overtime working hours and % of the project outputs come from the normal working hours. The ratio between them is :, equivalent to %. It reveals that overtime is three times as efficient as work done during normal working hours. To determine the reasons for overtime work, we listed several possible alternatives from the literature (Taylor ) and from the feedback of the pre-study. The respondents were asked to select one or more of the alternatives. More than % of the respondents pointed out that changes to design or requirements are the major reason for working overtime (see Table ). In addition, nearly % of the respondents think that the necessity to do overtime is due to an initial underestimation of project effort and duration.... Insight on RQ from the Main Study We found that the major reason for overtime work at the Chinese software suppliers is the change to design or requirements, rather than time-zone Table. Reasonsforovertimework(N =, multiple choice) Number of Percentage projects (%) Different time zones Different holiday periods Changes to design or requirements Required by the outsourcers Insufficient local competence Overoptimism about schedule or effort Resolve some specific problems Copyright John Wiley & Sons, Ltd. Softw. Process Improve. Pract., ; : DOI:./spip

12 J. Ma et al. AQ differences. Furthermore, an initial underestimation of effort and duration is also an important reason for overtime work. Some of the projects were underestimated because of lack of experience, but many might be also due to tactical bidding (Taylor ). That is, to get a contract, a supplier indicates an unrealistic time schedule or effort to the outsourcer. This then leads to later overtime work to keep to the deadline or budget. If the overtime work (% more) in our investigated projects were paid at the usual rate of % more than normal payment, the real effort budgets of our investigated software outsourcing projects should be increased by %. This means that software outsourcing projects suffer from under-budgeting ( %) (Moløkken and Jørgensen ) similar to that in other software projects. In our pre-study, some of the respondents attributed the frequent and unpaid overtime work in their companies to their companies encourage- ment to work overtime. As a result, the employees may not have tried as hard as they could to work as effectively during normal hours and may have postponed certain work until the overtime hours. This may explain our finding that the overtime work is much more productive than work done in normal working hours. Less interruption from other employees could be another explanation as towhyovertimeworkismoreproductive.however, given that our results show that about % full-time employees worked overtime, it is difficult to conclude that there would be fewer interruptions during overtime hours than in normal working hours.... Follow-up Study for RQ Our data show that changes to requirements is one of the two major reasons for overtime work, while time-zone difference is a minor one. That being so, we formulated the following follow-up research questions RQ.f and RQ.f: RQ.f: What are Chinese software suppliers current practices of solving requirement changes and what are the difficulties? RQ.f: How do Chinese suppliers deal with time-zone differences and differences in the occurrence of public holidays? To answer RQ.f, we asked respondents to tell us, in detail, how they address proposed changes to requirements. The answers from all companies were similar. When facing requirement changes initiated by the outsourcer, first, the suppliers hold an internal meeting to discuss the effect, the risk, and the possible cost due to changes. If the changes are acceptable, the supplier performs the changes at no extra cost. If a great deal of work has to be done to perform the changes, the estimated extra charge and time are reported to the outsourcer for discussion. The changes are negotiated until an agreement is reached. The agreement is signed by both partners before the proposed changes are sent to the developers for implementation. However, in most cases, the Chinese suppliers have to compromise and make changes without being paid extra or being given extra time in which to make the changes. Thus, the developers have to work overtime to fulfill changed requirements. With respectto RQ.f, the respondents answers show that working overtime is still the only way in which problems caused by different time zones and holiday periods are addressed. Normally, employees of the Chinese supplier companies mayhavetowork hoursmorethannormal working hours to attend telephone meetings with outsourcers. This kind of overtime work is usually unpaid. If the suppliers employees have to work overtime during holidays, they are sometimes paid double or treble the normal rate. The average wage level in China is not very high; hence, paying overtime wages for work done during holiday periods usually does not increase the project costs unduly. That is why project managers currently accept the simple overtime strategy without looking for other solutions.... Conclusion on RQ The main reasons for overtime work at Chinese software supplier companies are changes to design or requirements and initial overoptimism about the budget. Another reason for overtime work is the cultural encouragement given for overtime work, which may also reduce efficiency during normal work hours.. FINAL DISCUSSION Using the results of RQ to RQ as a basis, we first suggest three strategies for Chinese software suppliers to improve their communication and coor-dination with their outsourcers. We then discuss the possible threats to the study. AQ AQ Copyright John Wiley & Sons, Ltd. Softw. Process Improve. Pract., ; : DOI:./spip

13 Chinese Software Suppliers and Global Outsourcers.. Recommendations... Manage the Possible Language Differences with a Steering Group An absence of problems with language is one reason for successful software outsourcing. The success of the software industries in India, the Philippines, and Singapore are accepted examples (Carmel and Agarwal, Kshetri ). Before our study, few studies had systematically investigated the language skills of the employees of Chinese software suppliers. Outsourcers from English-speaking countries may still be sceptical about the English language skills of their Chinese suppliers. However, our results show that differences in natural languages are not a barrier to Chinese software suppliers. For Chinese suppliers, one possible way of managing the effects of language is to build a steering group that comprises language- and ITcompetent project managers and senior developers. The members of such steering groups can work as a bridge between the external outsourcers and internal developers. However, using such a steering group may delay the progress of a project, if the personnel in the group are not well-selected and organized.... Do Not Overuse Synchronous Communication Methods A previous study (Carmel and Agarwal ) shows that synchronous communication channels usually have more advantages than asynchronous ones. However, our results and another previous study by Smite () show that is still used frequently and is regarded as efficient. Calling on the telephone is regarded as another efficient communication channel between outsourcers and suppliers. However, it is not used as frequently as and is not often used to discuss issues that pertain to testing, maintenance, and scheduling. Other synchronous channels, such as net meeting and video conferencing, are used rarely. To improve the efficiency of communication and to reduce unnecessary communication costs, we recommend that suppliers avoid excessive synchronous communication. Face-to-face meetings should be used mainly to discuss issues related to business and requirements. For issues related to design, development and testing, is normally an efficient means of communication.... Manage the Requirements Changes and Avoid Unnecessary Encouragement on Overtime Work In software development, changes in requirements can be substantial as the project progresses, e.g. % per month or % over three years (Madhavji et al. ). Our results show that the main reason for overtime work in Chinese software supplier companies is the change to design or requirements initiated by the outsourcer. Another reason is the initial underestimation of effort, which is often caused by tactical bidding. We recommend that Chinese software suppliers appoint a change management group to discuss and negotiate changes with the outsourcers. An overall remedy to most of these issues is incremental development to reduce the risks for overruns, but this may be difficult to combine with the existing outsourcer supplier contracting process. With respect to differences in time zone and holiday periods, the current solution of Chinese suppliers is simply to ask their employees to work unpaid overtime. Other solutions to solve problems due to time-zone differences, for example, adjusting the local working time to harmonize with the outsourcer s working time, should supplement current overtime strategies. In addition, we observed that there is an unnecessary, and perhaps cultural, encouragement for overtime work, which may cause employees to postpone their work from normal work hours to overtime hours. Chinese suppliers need to improve their work efficiency in normal hours and try to avoid tactical bidding... Threats to Validity... Construct Validity In this study, most variables and alternatives were taken directly, or with little modification, from the existing literature. We did a pre-study to ensure that the quality of the questionnaire, and nearly % of the questions and alternatives in the final questionnaire were revised on the basis of the pre-study.... Internal Validity We promised the participants in the study that they would receive a final report or a one-year membership of the CSO (worth Chinese Yuan). Most respondents participated as volunteers and selected the report as the reward. Given their lack of financial motivation, we believe that, in general, the respondents answered the questionnaire truthfully. Copyright John Wiley & Sons, Ltd. Softw. Process Improve. Pract., ; : DOI:./spip

14 J. Ma et al. With respect to RQ, we tried to use quantitative metrics, such as lines of code, to measure project outputs in the overtime hours. However, the results of the pre-study showed that most companies did not record such data. The subjective measures that we currently use may threaten the conclusions of this research question. Another possible threat is that the respondents might have failed to recall past projects accurately, because our unit of study was a finished project.... External Validity There were more than, software companies registered in China in (Ministry of Information of the People s Republic of China & Chinese Software Industry Association ). However, the initial database we used contained only about half of them. Although we put in a lot of effort into collecting the data, we only received responses from out of companies selected at random. We do not know why the remaining companies did not participate. The respondents completed the questionnaires by recalling finished projects, which were selected on the basis of convenience. In addition, the participants in the follow-up interviews comprised only % of all participants of the survey. All the above issues may threaten the external validity of the study.. CONCLUSIONS AND FUTURE WORK We have the presented results of a state-of-thepractice survey on Chinese software outsourcing in industrial projects. Our main conclusions are the following: Differences in natural languages are not a critical factor for the success of Chinese software suppliers. Chinese suppliers English skills are good enough and measures (e.g. steering groups) can be taken to resolve language differences to a sufficient degree. is used to discuss all issues and is used more frequently to solve issues regarding development than any other means of communication, while face-to-face meetings are popularly used for issues that pertain to management and requirements. Net meeting and video conferencing are not much used. The main reasons for overtime work at the Chinese software suppliers are changes in the design or requirements and initial overoptimism about the effort or duration of a project. The overoptimism is often caused by tactical bidding. However, the current solution, i.e. unpaid overtime work, may reduce efficiency in normal working hours. Finally, given that China has no comprehensive, national database of IT companies, it is difficult to select a random sample of participants in such surveys. The results brought to light several important issues that will be investigated in the future: Although we find that overtime work is more efficient than work during normal hours, a further case study, using more reliable and quantitative metrics, is needed to confirm our conclusion. Our results show that it is important to control changes to requirements to avoid overruns of the outsourcing projects. Ways in which alternative processes, such as incremental or agile ones, can be used effectively in projects across geographic and company boundaries, need to be investigated. ACKNOWLEDGEMENTS This study was a joint research effort between BJUT and NTNU, partially funded by the Norwegian SEVO project with grant /V. We would like to thank the CSO for the data sampling and questionnaire collection. We also thank all participants in the study. REFERENCES Ahmed RE.. Software maintenance outsourcing: issues and strategies. Computers & Electrical Engineering ():, DOI:./j.compeleceng... Aubert BA, Rivard S, Patry M.. A transaction cost model of IT outsourcing. Information & Management ():, DOI:./j.im... Carmel E, Agarwal R.. Tactical approaches for alleviating distance in global software development. IEEE Software ():, DOI:./.. Donahoe DN, Pecht M.. Are U.S. jobs moving to China? IEEE Transactions on Components and Packaging Copyright John Wiley & Sons, Ltd. Softw. Process Improve. Pract., ; : DOI:./spip

15 Chinese Software Suppliers and Global Outsourcers Technologies ():, DOI:./TCAPT... Ferguson E, Kussmaul C, McCracken DD, Robbert MA.. Offshore outsourcing: current conditions and diagnosis. ACM SIGCSE Bulletin ():, DOI: Herbsleb JD, Grinter RE.. Splitting the organization and integrating the code: Conway s law revisited. Proceedings of the st International Conference on Software Engineering, Los Angeles,. Herbsleb JD, Paulish DJ, Bass M.. Global software development at siemens: experience from nine projects. Proceedings of the th International Conference on Software Engineering, St. Louis,. Kankanhalli A, Tan BCY, Wei K, Holmes MC.. Cross-cultural differences and information systems developer values. Decision Support Systems ():, DOI:./S-()-. Krishna S, Sahay S, Walsham G.. Managing cross-cultural issues in global software outsourcing. Communications of the ACM ():, DOI: doi.acm.org/./.. Kshetri N.. Structural shifts in the Chinese software industry. IEEE Software ():, DOI:./MS... Lacity M.. Lessons in global information technology sourcing. IEEE Computer ():, DOI:./MC... Madhavji NH, Fernandez-Ramil J, Perry D.. Software Evolution and Feedback: Theory and Practice. Wiley: New York. Ministry of Information of the People s Republic of China & Chinese Software Industry Association.. Annual Report on the Chinese Software Industry, Moløkken K, Jørgensen M.. A review of surveys on software effort estimation. Proceedings of the International Symposium on Empirical Software Engineering, Rome,. Smite D.. Global software development projects in one of the biggest companies in Latvia: is geographical distribution a problem? Software Process Improvement and Practice ():, DOI:./spip.. Taylor H.. Critical risks in outsourced IT projects: the intractable and the Unforeseen. Communications of the ACM ():, DOI: Walsham G.. Cross-cultural software production and use: a structurational analysis. MIS Quarterly ():. Wong J.. China s economy in : at a new turning point and need to fix its development problems. China and World Economy ():, DOI:./j.- X...x. Copyright John Wiley & Sons, Ltd. Softw. Process Improve. Pract., ; : DOI:./spip