A Neural Networks-Based Evaluation Of Quality In The SMEs: An Application In Turkish Textile Sector

Transcription

1 Proceedings of 5th International Symposium on Intelligent Manufacturing Systems, May 29-31, 2006: Sakarya University, Department of Industrial Engineering A Neural Networks-Based Evaluation Of Quality In The SMEs: An Application In Turkish Textile Sector Ali Türkyılmaz Dept. of Industrial Engineering, Buyukcekmece Istanbul, Turkey Phone: Fax: aturkyilmaz@fatih.edu.tr Güneş Gençyılmaz Dept. of Management, Kültür University Şirinevler İstanbul Phone: g.gencyilmaz@kultur.edu.tr Selim Zaim Dept. of Management, Fatih University, Buyukcekmece, Istanbul, 34500, Turkey Phone: Fax: szaim@fatih.edu.tr Alexander Nikov Dept. of Industrial Engineering, Buyukcekmece Istanbul, Turkey Phone: Fax: nikov@fatih.edu.tr Abstract: In this study an approach for neural networks evaluation of quality management (QM) is proposed. It is based on a checklist tool using a neural networks-based evaluation model. This approach enables the determination of the most important quality criteria and the generation of recommendations for improving the quality management in textile sector. A neural networks method is illustrated and validated by a study in 156 small- and medium-sized textile companies in Istanbul. Data from company s managers were collected by the checklist tool. Then QM evaluations were calculated using neural networks approach. Based on the approach the most important QM criteria that affect business performance in the textile industry were determined. Relevant recommendations and measures for improving the textile industry were proposed. The advantages, drawbacks and further developments of the study are discussed. Keywords: Quality management, textile industry, business performance, neural networks

2 1 Introduction Quality has become one of the most important drivers of the global competition today. Intensifying global competition and increasing demand for better quality by customers have caused more and more companies to realize that they will have to provide high quality product and/or services in order to successfully compete in the marketplace. To meet the challenge of this global competition, many businesses have invested substantial resources in adapting and implementing total quality management (TQM) strategies. Total quality management (TQM) is an integrated management philosophy and covers many important aspects such as customer satisfaction, meeting customers requirements, reducing rework, reducing wastes, increased employee involvement and team work, process redesign, and supplier management. Any organization, which can be in manufacturing and/or service sector, or profit or non-profit oriented must generate high quality products or services to stay in business. Therefore organizations need to adopt TQM. Usually, successive and effective TQM practices are found in larger and/or multinational organizations (Kanji et al., 2000), (Deming, 1986). Recently, large organizations focus on the core activity and uses extensive practices of outsourcing. Therefore, larger companies usually provide various goods and services from SMEs. Large organizations product and services quality will depend on the quality of suppliers product and services. Thus, large companies encourage the application of TQM approaches among their suppliers, which are most likely to be SMEs. Small and medium - sized enterprises (SMEs) has an important role in the economy of a country such as trade, employment, share of GDP, and job generation potential. SMEs are defined as private enterprises outside the agricultural sector, employing fewer than 250 people. The purpose of this study is to determine the critical factors of total quality management and to measure its effect on business performance in small and medium sized enterprises in Turkey. There are many methods for company quality management evaluation. In this study, we propose a neural-network-based method for quality evaluation of SMEs. Based on a case study, it is found to provide a better understanding of managers perceptions of quality and its improvement (Fornell et al., 1996), (Kaplan et al., 1996), (Kunst et al., 2000), (Jacop et al., 2004), (Raju et al., 2002), (Madu et al., 1962). 2 Description Of The Study Principal purposes of this method are: 1) aggregating the quality management data for evaluations; 2) selecting important QM evaluation criteria for the generation of recommendations and measures for performance improvements. The study consists of five steps. 2.1 Step 1: Knowledge base design For evaluation and design of QM, a knowledge base including evaluation criteria such as role of top management, training of employees and managers, process management, quality data and reporting, employee relations, manager s knowledge, corporate support for quality, past quality performance can be developed. This study is based on the adapted checklist instrument of Saraph, Benson & Schroder, (1989) (cf. Appendix). The checklist tool consists of two parts. The first input part contains 30 questions that identify 1231

3 QM evaluation criteria for role of top management, training of employees and managers, process management, quality data and reporting, employee relations, managers knowledge, corporate support for quality, past quality performance. These questions are asked to managers of the companies for measuring their perceptions of the quality management performance of the company. In the second output part of the checklist tool, a single question regarding manager s overall evaluation of financial performance is asked. For the above criteria, a criteria hierarchy is defined (see Fig. 1). To each criterion, a weight is associated by expert estimates. Each weight represents the relative importance of each sub-criteria within their category. INPUT HIDDEN LAYER OUTPUT LAYER Financial performance Figure 1: Neural network based hierarchy of HSQ evaluation criteria 2.2 Step 2: Data collection During this step the managers evaluations of the single criteria at the lowest first layer of the criteria hierarchy are determined. 2.3 Step 3: Criteria hierarchy learning During this step of our study the initial weights of criteria hierarchy are trained. Because of the strong nonlinear correlation between evaluation criteria we chose a nonlinear evaluation model: the backpropagation (BP) algorithm (Sarkar, 1998). It carries out supervised learning of neural network weights using training data as inputs and known output minimizing the mean square error. In this study, managers perception related to 1232

4 pre-specified QM criterion is paired with overall financial performance evaluation. During the training process, the neural network learns the relationship between output and input criteria. The network arcs connect the processing units -the neurons. With each neuron input a weight is associated, which represents its relative importance in the set of neurons inputs. The inputs of each neuron from other neurons are aggregated. Its net value represents a weighted combination of the neuron inputs. The hierarchy is coded in a hierarchical neural network, where each neuron corresponds to a criterion. The single criteria correspond to the network inputs. The complex criteria correspond to the neurons at hidden layer and to the output neuron. The net function of the neural network is used as an evaluation function. The BP algorithm is a gradient algorithm, which minimizes the average square error between the current output and the target value by modification of the network weights. We consider a neural network with 21 inputs (single criteria), nine hidden neurons (complex criteria) and an output neuron (financial performance) (Maren et al., 1990), (Wasserman et al. 1993), (Rumelhart et al., 1986), (Lippman, 1987) 2.4 Step 4: Lack analysis Lack analysis is accomplished by determining the most important single criteria at input layer whose improvement will lead to the most significant increase of the financial performance. 2.5 Step 5: Improvement/redesign recommendations Based on previous steps, improvement or redesigning recommendations are generated. 3 Case Study Data for this study was collected using a self-administered questionnaire that was distributed to 350 SMEs operating in textile industry within the city of Istanbul in Turkey. This sample was selected randomly from the database of Turkish Small Business Administration (KOSGEB). As of 2005, the KOSGEB database includes a total of 12,270 SMEs in Istanbul, which accounts for nearly 28% of all SMEs registered throughout Turkey. The sampling frame consists of 2,482 SMEs operating in the textile industry in Istanbul. The study focused on the textile industry including textile mill products and apparel, since it has been a leader in implementing progressive quality management practices in Turkey. It was requested that the questionnaire be completed by a senior officer/executive in charge of quality management. The responses indicated that a majority of the respondents completing the questionnaire were in fact members of the top management. Of the 350 questionnaires posted, a total of 180 questionnaires were returned after one follow-up. Twenty-four observations were eliminated due to largely missing values. The overall response rate was thus 45% (156/350), which was considered satisfactory for subsequent analysis. A comparison of the annual sales volume, number of employees and sub-industry variation revealed no significant differences between the responding and non-responding firms (p>0.1). Thus, the responses adequately represented the total sample group. The instrument used in this study was developed by Jayant V. Saraph, P. George Benson, and Roger G. Schroeder (1989) with the purpose of identifying critical factors (areas) of total quality management in a business unit. Business performance is computed as an overall score of financial performance. 1233

5 3.1 Determining critical factors of total quality management using Exploratory Factor Analysis Exploratory factor analysis with varimax rotation was performed on the total quality management criteria in order to extract the dimensions underlying the construct. The factor analysis of the 30 variables yielded four factors explaining 83.93% of total variance. Only eleven of the thirty items loaded on these four factors and, based on the items loading on each factor, the factors were labeled " Role of top management and quality policy (Factor 1), Process management (Factor 2), Quality data and reporting (Factor 3), Employee relations (Factor 4). These eleven items are shown in Table 1. The Cronbach s alpha measures of reliability for the four factors were for factor 1, for factor 2, for factor 3, for factor 4. Since Cronbach s alpha measures for each factor are above the traditionally acceptable value of 0.70, all of the factors were accepted as being reliable for the research. Table 1 Factor analysis of total quality management criteria Variables Factors Extent to which top executives assume responsibility for quality performance Extent to which top management has objectives for quality performance Extent to which top management has developed and communicated a vision for quality as part of a strategic vision of the organization Amount of preventive equipment maintenance Amount of inspection, review, or checking of work Clarity of work or process instructions given to employees Availability of quality data Extent to which quality data are used as tools to manage quality Scope of the quality data includes operational performance Extent to which quality awareness building among employees is on-going Extent to which employees are recognized for superior quality performance 3.2 Lack analysis and generation of improvement/redesign recommendations The current quality management of the companies in textile industry in Istanbul was analyzed according to the steps 4 and 5. Relevant improvement and redesign recommendations were generated. The evaluation of current service quality was carried out in the two stages. At the first stage complex criteria was evaluated and ranked from most important to less important given in Table 1. At the second stage considering the complex criteria, single criterion scores were calculated as shown in Table 2. Subsequently, improvement plans were developed. Table 2 Complex criteria, their evaluations, and relevant areas and ranking Number of factor Complex criteria Evaluations Ranking 1 Role of Top Management and Quality Policy Employee Relations Quality Data and Reporting Process Management

6 Table 3: Single criteria, relevant complex criteria and evaluations Single criteria Extent to which top management has objectives for quality performance Extent to which top management has developed and communicated a vision for quality as part of a strategic vision of the organization Extent to which top executives assume responsibility for quality performance Extent to which employees are recognized for superior quality performance Extent to which quality data are used as tools to manage quality Related complex criteria Role of Top Management and Quality Policy Role of Top Management and Quality Policy Role of Top Management and Quality Policy Evaluations Employee Relations Quality Data and Reporting In the first step, according to the evaluation of the complex criteria, the role of top management and quality policy with a score of 100 was found to be the most important criterion (cf. Table 1). According to Table 2, it is seen that there is no any significant difference between single criterion related to the role of top management and quality policy. In textile industry, successes of TQM applications depend on a strong leadership that must be initiated by the top management. Quality improvement plans proposed by several gurus emphasize primarily the commitment of top management. The top management of the firms determines an appropriate organization culture, vision, and quality policy. Managers of the organizations should determine objectives, and set specific measurable goals to satisfy customer expectations and improve their organizations performance. On the other hand, the top managements must provide adequate resources to the implementation of quality efforts. As it is shown in Table 1, although employee relations, quality data and reporting and process management has almost equal weights with each other the employee relations was slightly different and found to be a second important complex criteria improving firm performance (score ). According to Table 2, extent to which employees are recognized for superior quality performance was found the most important criterion related to the employee relations complex criteria. Companies must develop formal reward and recognition systems to encourage employee involvement, and support teamwork. Quality data and reporting, was found as the third important complex criteria. There are many purposes for gathering data in quality management. Data can be collected to determine defect rate in company to understand current processes. Moreover, data provides inspection, various test results and verification records. Data also are used to analyze the process using various types of statistical process control tools such as control charts, Pareto charts, cause and effect diagrams, check sheet, histograms, scatter diagram, and so on. These traditional quality tools are very useful in monitoring and measuring progress and performance. Management by facts requires that management decisions are based on relevant data and reports. In this model, data and reporting has a very strong correlation with TQM and financial performance of the company. Fourth factor, process management, which includes such criterion as process monitoring, supervision, and preventive equipment maintenance, found the least important complex criteria in the model. After determining important complex and related single criteria one needs to develop an improvement plan to increase total quality management performance and consequently

7 improve the financial performance of the firm. If the performance of the above mentioned single criteria is enhanced, the overall financial performance of the company will be improved. 4 Conclusions An approach for QM evaluation of companies in textile sector is proposed. It is based on a checklist tool using a neural networks-based evaluation model. The most important QM criteria can be determined and recommendations for improving the firm quality and performance can be generated. This is illustrated and validated by a study in 156 smalland medium-sized companies in Istanbul. Based on the approach, the most important QM criteria that affect perfomence of the company in textile industry in Istanbul were determined. Relevant recommendations and measures for improving the firm performance were proposed. While most SMEs in Turkey do not have an established quality department, many of them have invested substantial resources in adapting and implementing TQM programs to improve their performance. It is generally accepted that several SMEs did not achieve any improvement and some only a little. Specifically, due to the presence of a multitude of barriers, many organizations utilize only a partial implementation of TQM, and hence are unable to achieve continuous and systematic improvement. If TQM plan is implemented properly, it produces a variety of benefits such as understanding customers needs, improved customer satisfaction, improved internal communication, better problem solving and fewer errors. The success of a TQM program increases when its implementation is extended to the entire company. This enables the reformation of the corporate culture and the permeation of the new business philosophy into every facet of organization. The philosophy of doing things right must be implemented with enthusiasm and commitment throughout the organization -from top to bottom and the little steps forward the so-called Kaizen by the Japanese- must be viewed as a race without a finish. Consequently, effective implementation of TQM is a valuable asset in a company s resource portfolio, one that can produce important competitive capabilities and be a source of competitive advantage. It should also be acknowledged that the study is subject to some methodological limitations. First, it would be highly suggested that the size and nature of the sample must be enhanced to ensure variability and control for possible extraneous variation. While the sample is restricted to only a single region and a single industry, it would be strongly recommended that data should be gathered from various parts of Turkey including both various manufacturing and service industries. Since the data in this study were collected from top managers of organizations on the basis of their subjective evaluations, objective performance indicators should also be employed in the analysis. Finally, additionel constructs could be utilized in the future studies to gain additional insights in exploring the relationship between TQM and organizational financial performance. 1236

8 5 REFERENCES A. Maren, C. Harston and R. Pap, Handbook of Neural Computing Applications, Academic Press, London, ISBN , 1990 Coşkun Özkan, Selim Zaim, Ali Türkyılmaz "Küçük Ve Orta Büyüklükteki İşletmelerde Toplam Kalite Uygulamalarının, Kurum Performansına Etkileri" 5. Üretim Araştırmaları Sempozyumu, UAS2005, İstanbul Ticaret Üniversitesi, İstanbul, Türkiye, Kasım 2005 Deming WE. Out of the Crisis. MIT Center for Advanced Engineering Study, Cambridge, MA, Jacob R, Madu CN, Tang C. An empirical assessment of the financial performance of Malcolm Baldrige Award winners. International Journal of Quality & Reliability Management 2004; Vol. 21, Kanji GK, Wallace W. Business Excellence through Customer Satisfaction. Total Quality Management 2000; Vol. 11, Kunst P, Lemmink J. Quality Management and Business Performance in Hospitals: A Search for Success Parameters. Total Quality Management 2000; Vol. 11, Lippman, R. P., "An introduction to computing with neural nets," IEEE ASSP Magazine, pp. 4-22, 1987 Madu CN, Kuei CH, Jacob RA. An Empirical Assessment of the Influence of Quality Dimensions on Organizational Performance. Internal Journal of Production Research 1996; Vol. 34, Raju, P.S., & Lonial, S.C. (2002). The Impact of Service Quality and Marketing on Financial Performance in the Hospital Industry: An Emprical Examination. Journal of Retailing and Consumer Services, 9, Rumelhart, D. E., G. E. Hinton, and R. J. Williams, "Learning representations by back-propagating errors," Nature, vol. 323, pp , Saraph, J.V., Benson, G.P. & Schroder, R.G. (1989). An instrument for measuring the critical factors of quality management. Decision Science, 20, Wasserman, P. D., Advanced Methods in Neural Computing, New York: Van Nostrand Reinhold, 1993 Appendix Input 1. Extent to which top executives assume responsibility for quality performance. 2. Acceptance of responsibility for quality by major department heads. 3. Degree to which top management (top executive and major department heads) is evaluated for quality performance. 4. Extent to which top management supports a long term quality improvement process. 5. Extent to which the top management has objectives (Management By Objectives) for quality performance. 6. Importance attached to quality by top management in relation to cost/revenue objectives. 7. Degree to which top management considers quality improvement as a way to increase profits. 8. Degree of comprehensiveness of the quality plan. 9. Extent to which top management has developed and communicated a vision for quality as part of a strategic vision of the organization. 10. Use of statistical control charts to control processes. 11. Amount of preventive equipment maintenance. 12. Amount of inspection, review or checking of work. 13. Importance of inspection, review or checking of work. 14. Stability of work schedules. 15. Clarity of work or process instructions given to employees. 16. Availability of cost of quality data in the company. 17. Availability of quality data. 18. Timeliness of quality data. 19. Extent to which quality data are used as tools to manage quality. 1237

9 20. Extent to which quality data are available to managers and supervisors. 21. Extent to which quality data are used to evaluate supervisor and managerial performance. 22. Extent to which quality data, control charts, etc. are displayed in work areas. 23. Scope of the quality data includes operational performance and process performance. 24. Extent to which employee involvement type programs are implemented in the company. 25. Effectiveness of quality teams or employee involvement type programs in the company. 26. Extent to which the employees are held responsible for error free output. 27. Amount of feedback provided to the employees on their quality performance. 28. Degree of participation in quality decisions by hourly/non-supervisory employees. 29. Extent to which quality awareness-building among employees is on-going. 30. Extent to which employees are recognized for superior quality performance. Output: 1. Financial performance 1238