Wanted Combinations of Supplier Attributes for PRD manufacturers
Billy T.W. Yu, Peter, K.C. Lee and W.M. To
Macao Polytechnic Institute, Macao SAR
Abstract (No.003-0397)
For the sustainable and mutually beneficial development for all provisional governments and their people concerned around the Pearl River Delta (PRD), trade barriers were lowered recently especially in light of the launching of the Closer Economic Partnership Arrangement (CEPA) between the mainland, Hong Kong and Macao. With the increase in the number of potential customers, suppliers would better profile themselves precisely. This paper adopted a data-mining approach and utilized a classification model to help suppliers predict the combinations of supplier-attributes that specific type of prospective customers (or buyers) are in quest of. A number of attributes concerning the customers and the suppliers, like the purchasing capability of the customers, trust, communications, and the degree of alliance integration, were critically reviewed.
Keywords
Data-mining, supplier evaluation, trust, communications, alliance integration, predictive modeling
- Introduction
Supply chain management is process-oriented. It integrates material procurement, production and distribution, financial arrangement, and the exchange of information between all parties concerned (Handfield and Nichols 1999). It requires the cooperation between suppliers, manufacturers, wholesalers, retailers and consumers. In order to continually improve supply chain management for a particularindustry, information on supply chain practices between suppliers and buyers (or the expectations and satisfaction levels of buyers on their suppliers) have to be gathered and analyzed periodically. One of the best ways to obtain information from buyers is to conduct a questionnaire survey with a well-developed instrument. However, there are numerous ways to analyze the data. Some researchers (Tan et al. 1998, Hong and Cheng 2004) analyzed the data using simple statistical techniques such as correlation and linear regression. Others (Fearne et al. 2004, Spekman et al. 2002, Bagchi 2004) preferred advanced statistical techniques such as analysis of variance, multiple regression, etc. Power (2002) proposed a structural equation model and analyzed supply chain management for the manufacturing and retailing industries in Australia. Most of these researchers would have undertaken a mental filtering process before analyzing their data sets, resulting in unobvious missing connections to be explored in their conclusions and mathematical models. To alleviate the mental manipulation that might have taken placein numerous incidents, Adriaans and Zantinge (1996) defined data mining as the use of sophisticated data analysis tools to discover previously unknown, valid patterns and relationships in large data sets. In this paper, we propose to use a data mining approach in exploring the expectations of different electronic manufacturers on their suppliers in and around the Pearl River Delta region.
- Literature Review
The business environment of today is so competitive that many companies are forced to explore the sources of competitive advantages from internal activities to external suppliers. A partnering buyer supplier relationship is widely recognized as an important source of competitive advantage. The Michigan State University's Global Procurement and Supply Chain Electronic Benchmarking Network (GEBN, 1995) surveyed a group of supply chain management executives from 77 companies and defined strategic supplier alliances (SSA) as the long-term, cooperative relationships designed to leverage the strategic and operational capabilities of individual participating companies to achieve significant ongoing benefits to each party.
Based on this definition, it can be inferred that a partnering supplier relationship consists of two groups of attributes. The first group refers to the relational aspects of the relationship, i.e., ‘long-term and cooperative’ or other related traits. The second group is concerned with management practices relating the leverage of capabilities of the supplier and the buying firm. In this paper, a total of nine attributes were selected as the major factors constituting a partnering supplier relationship. Four of them are relational aspects of the partnership, and the other five are purchasing practices that are related to integrating the capabilities of participating companies. Definitions of these nine attributes are presented in Table 1.
SSA Practices
/Definition
Relational Practices
Cooperative Attitude / A cooperative norm in which the participating companies work together to achieve mutual and individual goals (Cannon and Perreault 1999).Long-term Commitment / The belief that the relationship will continue to the future and benefit both participating companies in the long run (Ganesan 1994).
Communication / The extent to which critical, often proprietary, information is communicated with suppliers (Mohr and Spekman 1994).
Trust / A concern in the partner’s honesty and benevolence (Doney and Cannon 1997).
Purchasing Practices
Supplier Development / Activities undertaken by a buying company to improve either supplier performance, supplier capabilities, or both, so as to meet the buying company’s short- and long-term supply needs (Krause et al. 2000).Joint Action / The buying company working jointly with the supplier on certain activities that are important to both parties (Joshi and Stump 1999).
Information Technology
(IT) Application / The use of IT for enhancing the internal efficiency and effectiveness of the purchasing function and for improving inter-company communication and transaction processing (Sriram et al. 1997).
Supplier Evaluation / A formal programme to assess the performance of potential or existing suppliers.
Purchasing Knowledge
and Skills / The knowledge and skills of a company’s purchasing personnel in the supply market and managing suppliers by modern purchasing practices.
Table 1 The Definitions of Attributes about Partnering Supplier Relationships.
These attributes have been widely studied by researchers. For instance, the work of Siguaw et al. (1998) showed that a cooperative relationship is associated with trust, commitment and financial performance. Findings of Heide and John (1990) showed the existence of linkages between resources commitment and joint action or continuity. Anderson and Weitz (1992) found that open sharing of information leads to increased commitment to a relationship. The study of Carr and Smeltzer (1999) indicated that communication is related to the strategic level of a purchasing function). Further, direct linkage between trust and company performance was found by Siguaw et al. (1998) and Zaheer et al. (1998). The results of a recent survey(Krause et al. 2000)also indicated that supplier development activities including supplier incentives, supplier assessment, etc., are associated with performance improvement. Moreover, with data collected from the US (Carr and Smeltzer 1997), and Taiwan (Carr et al. 2000), a correlation between purchasing knowledge and skills and strategic purchasing was revealed.
Although the results of these existing studies are capable of offering new insights to the literature, the research design of these studies relied on explicit assumptions concerning the associations between practices or attributes of supplier management. However, such assumptions were often made on the basis of intuitive judgments of the researchers. The lack of quantitative evidence in the formulation of research design hinders the credibility of the results. This paper provides a demonstration to operations management researchers on the use of a novel data analysis method, data mining, to identify the supplier attributes that were preferred by customers. More specifically, the linkages between attributes were explored without making the association, mostly un-directional and one-to-one, assumptions of conventional statistical analysis.
- Methodology
3.1The Context of the Study
The data of this study was collected from the electronics industry of Hong Kong. The electronics industry is the second largest manufacturing industry in terms of gross output and domestic exports in Hong Kong. In 2003, there were around 900 companies in the Hong Kong electronics industry that accounted for 42% of Hong Kong’s total exports. The gross output of the industry was HK$30,376 million (USD 3,894 million) in 2001 (Hong Kong Trade Development Council 2004). The industry is export-oriented with the United States as the major market. It predominantly focused on labor-intensive, cost-based strategies, and typically manufactures electronic watches and clocks, telecommunications equipment, medical electronics, automotive electronics, and computers and peripherals (Hong Kong Industry Department 2000) with major manufacturing bases in Pearl River Delta.
3.2Instrument Development
The detailed development of the instrument of this study was documented in the work of Mak and Lee (2005). Briefly speaking, based on a review of the literature relating to partnering supplier relationships, a preliminary version of the questionnaire was developed. The draft questionnaire was pilot-tested with eight purchasing management practitioners and three experienced researchers. Based on the comments obtained from the pilot-test, several minor changes were made to the questionnaire in order to improve readability.
After completing the pilot test, a large-scale survey was conducted to collect data from the Hong Kong electronics industry. Based on the directory published by Hong Kong Electronic Industries Association Limited (1999), 900 companies were randomly selected for telephone contact. This initial contact served to confirm that the target companies were available and to identify the key informants who were responsible for looking after the purchasing function and/or the manufacturing function. The calls to the 900 companies found that 758 of them had at least one manufacturing plant within the region. At the ended of the survey, a total of 175 useable questionnaires were collected, resulting in a response rate of 23.1% (i.e. 175/758). With respect to the measures, all attributes of this study were measured by multi-item indicators on a 7-point Likert scale. The reliability and validity of the measures were validated by Coefficient Alpha test and factor analysis (Principal Component Analysis), respectively.
3.3Data Mining
3.3.1 The concept
Suppose in the absence of idea about the set of supplier attributes that buyers desire, one have to group the supplier and buyer attributes without a pre-specified dependent attribute. Thus, an unsupervised learning technique in data mining was adopted here. It is the association rule. Association rule finds interesting associations and/or correlation relationships among large set of data items. It shows attribute values that occur frequently together in a given dataset (Agrawal 1993 and 1994). In the beginning, association rule focused on market “basket data” that storeditems purchased on a per-transaction basis. A typical and stunning example of an association rule is that it finds out that people who buy beer also buy diapers in supermarket analyses (Berry 1997 and 2002 and MacDougall 2002).
3.3.2 The method
In 1993, Agrawal proposed the original method. Optimizing association rule provides an effective way to focus on the most interesting characteristics involving certain issues (Rastogi 1998). The application here could be described as follows. Let I = {i1, i2,…, im} and J = {j1, j2,…, jm} be a set of buyer and supplier attributes, respectively. Technically, the i’s and j’s are all called items. A filled questionnaire with XI and YJ is called a set of items, or an itemset. Each questionnaire keeps an identity, and technically it is called a database transaction. An association rule is an implication of the form AB, where AI∪J and BJ. Ais said to be the antecedent, while B is the consequent. If the supplier attributes are included in an antecedent,a supplier can match his own characteristics with the client’s so as to predict the buyer’s desired on further supplier attributes.
Let say D be a set of all relevant database transactions. The association rule AB holds in the transaction set D with support s. s(AB) = P(A∪B). The support is simply the proportion of transactions that include all items in the antecedent and consequent parts of the rule. It is expressed as a percentage of records in the database.
Another important parameter in the association rule is confidence c. c(AB) = P(A︱B). It is the proportion of transactions that include all items in the consequent as well as the antecedent to the number of transactions that include all items in the antecedent.
The confidence of a rule (AB) is an estimate of the conditional probability of itemset B given itemset A. It does not measure the strength of the implication. It can be misleading when the occurrence of itemset B is more that itemset A; where the confidence can be easily achieved. To filter out this misleading association, a lift of the association is introduced to evaluate a speculated correlation. lift is defined as P(A∪B) / ( P(A)‧P(B) ). ItemsetsA and B are said to be positively correlated when the lift is greater than 1; the occurrence of one itemset positively implies the occurrence of the other.
The result is a model of association rules providing information in the form of “if-then” statements. These rules are computed from the data and, unlike the if-then rules of logic, association rules are probabilistic in nature.
3.3.3 The process
With the core method described above, the raw data have to be preprocessed including data cleaning, aggregation and transformation (Sang2002). Our transformation here is to fit the 7-point scale to 3-point so that we would have just the high-medium-low scale. Two of the advantages of the association rule algorithm worth mentioning here are that it typically only identifies patterns that occur in the database, and it is useful in processing missing data, which probabilistically contributes to the support. (Berry 2000 and Nayak 2001). And the algorithm is especially good for sparse data, such as items in supermarket. Association rule is generalized to dependence rules, which identify statistical dependence in both the presence and absence of items in itemsets (Silverstein 1998). Thus, it is fitting for the association rule algorithm to process questionnaire data.
After data preprocessing, pattern searching started with the method described in the last section. Patterns found with preset criterion, that is the confidence and support, were checked against their corresponding lift so as to assure the correlations.
4.Results and Interpretations
With minimum requirement of support and confidence of 0.2 and 0.75, respectively, Table 2 lists the rules found in the analysis, when supplier attributes known to the buyer are considered.
Antecedent / Hits / Consequence / Hits / Conf / LiftCOMMUN=high TRUST=high / 71 / COOPRELA=high / 61 / 0.86 / 1.67
LTCOM=high TRUST=high / 74 / COOPRELA=high / 62 / 0.84 / 1.63
COOPRELA=high LTCOM=high / 75 / TRUST=high / 62 / 0.83 / 1.66
COOPRELA=high COMMUN=high / 76 / TRUST=high / 61 / 0.8 / 1.61
TRUST=high / 87 / COOPRELA=high / 67 / 0.77 / 1.5
COMMUN=high LTCOM=high / 88 / COOPRELA=high / 67 / 0.76 / 1.48
Table 2 Data Mining Results
The results indicate the presence of a number of possible propositions in the data.
Proposition 1:When the buyer has a high level of communication and it knows the supplier has a high level of trust, a supplier with a high level of cooperative attitude will be preferred.
Proposition 2:When the buyer has a high level of long-term commitment and it knows the supplier has a high level of trust, a supplier with a high level cooperative attitude will be preferred.
Proposition 3:When the buyer has a high level of long-term commitment and it knows the supplier has a high level of cooperative attitude, a supplier with a high level of trust will be preferred.
Proposition 4:When the buyer has a high level of communication and it knows the supplier has a high level of cooperative attitude, a supplier with a high level of trust will be preferred.
Proposition 5:When the client knows a supplier has a high level of trust, the supplier with a high level of cooperative attitude will be preferred.
Proposition 6:When the buyer has a high level of communication and long-term commitment, a supplier with a high level of cooperative attitude will be preferred.
Theses results clearly indicate that cooperative attitude and trust are the critical attributes of a partnering supplier relationship among all attributes analyzed, since these two attributes are generally preferred by buyers. They also clearly demonstrate that data mining is able to offer quantitative evidence to show the possible correlation among attributes or constructs without the need to make associative assumptions. With such a quantitative evidence, more accurate conceptual models or hypotheses can be developed for testing by rigorous methods such as multivariate or structural equation modeling in which assumptions on the cause-effect relationships between constructs are made inevitably.
- Conclusions
The data mining approach of analyzing a questionnaire survey using an association rule algorithm was demonstrated. Although the patterns revealed were not surprising ones, the results illustrated the power of this unsupervised method of patterns identification. The analyzed results also indicated the possible correlation among attributes without setting up pre-assumptions on the underlying mechanism on supply chain management. By relaxing support and confidence on the analyzed attributes, it was found that higher-order interactions could be explored and interesting combinations of buyer-attributes and supplier-attributes were associated.
References
Agrawal R., and Srikant R. (1994) “Fast Algorithms for Mining Association Rules in Large Databases”,Proceedings of the 20th International Conference on Very Large DataBases, September 1994, pp.478-499.
Agrawal R., Imielinski T., and Swami A. (1993) “Mining Association Rules between Sets of Items in Very Large Databases”,Proceedings of the ACM SIGMOD Conference on Management of Data, Washington D. C., May 1993, pp.207-216.
Anderson, E. and Weitz, B.A. (1992), “The Use of Pledges to Build and Sustain Commitment in Distribution Channels”, Journal of Marketing Research, Vol. 29, pp. 18-34.
Bagchi P.K. (2004) “Supply Chain Integration in Nordic Firms”, Proceedings of the 2nd World Conference on POM, April 30-May 3, Cancun, Mexico.
Berry, M. J. A. and Linoff, G. (1997) Data Mining Techniques for Marketing, Sales and Customer Support, New York: John Wiley & Sons, Inc.
Berry, M. J. A. and Linoff, G. (2000) Mastering Data Mining: The Art and Science of Customer Relationship Management, New York: John Wiley & Sons, Inc.
Cannon, J.P. and Perreault, W.D., Jr. (1999), “Buyer-Seller Relationships in Business Markets”, Journal of Marketing Research, Vol. 36, pp. 439-460.
Carr, A.S. and Smeltzer, L.R. (1997), “An Empirically Based Operational Definition of Strategic Purchasing”, European Journal of Purchasing & Supply Management, Vol. 3, Issue 4, pp. 199-207.
Carr, A.S. and Smeltzer, L.R. (1999), “The Relationship among Purchasing Benchmarking, Strategic Purchasing, Firm Performance, and Firm Size”, The Journal of Supply Chain Management, Fall, pp. 51-60.