Eindhoven University of Technology MASTER Analysis of suppliers' sustainability learning behavior Kayhan, M.

(1)

Eindhoven University of Technology

MASTER

Analysis of suppliers' sustainability learning behavior

Kayhan, M.

Award date:

2020

Link to publication

Disclaimer

This document contains a student thesis (bachelor's or master's), as authored by a student at Eindhoven University of Technology. Student theses are made available in the TU/e repository upon obtaining the required degree. The grade received is not published on the document as presented in the repository. The required complexity or quality of research of student theses may vary by program, and the required minimum study period may vary in duration.

General rights

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

(2)

E INDHOVEN U NIVERSITY OF T ECHNOLOGY

Department of Industrial Engineering & Innovation Sciences Operations, Planning, Accounting and Control (OPAC) group

Analysis of Suppliers’ Sustanability Learning Behavior

M. Kayhan

In partial fulfillment of the requirements for the degree of Master of Science in Operations Management and Logistics

Supervisors:

Dr. T. Tan (TU/e) Dr. S. Dabadghao (TU/e)

Dr. Z. Atan (TU/e) M. Baren (Philips) D. A. McNeill (Philips)

August 19, 2020

(3)

Abstract

As the supplier sustainability is getting more attention from focal firms, governments and researchers, many studies investigate sustainable supplier management and supplier sustainability development. However, there is still a significant gap in the literature related to supplier sustainability development, especially, in quantitative studies. Consequently, this study aims to fill the gap by (i) understanding the suppliers’ sustainability learning behavior which means evolution of supplier’s sustainability levels throughout the time, and their relationship to supplier characteristics,(ii) modeling suppliers’ learning process by using learning curve theory and construct predictive models for parameters of the learning curves and (iii) increasing the effectiveness of suppliers sustainability improvement process by considering suppliers’ learning capabilities and several buyer involvement strategies. A complex machine learning model (XGBoost) is used to understand the relationship between the learning behavior of suppliers and supplier characteristics by using SHAP model. Results show that, starting score of the suppliers, activities that are performed by suppliers, facilities located at suppliers’ sites, worker distribution variables and certificates have a relationship with the learning behavior of suppliers. While modeling the learning curves of suppliers using a learning curve model (time constant learning curve model), historical data is used, and parameters are estimated by using non-linear least square method. Afterwards, another machine learning model (random forest regressor) is applied to predict parameters of the learning curves by considering different sustainability learning behaviors. In order to structure the sustainability improvement process, several optimization models are constructed. The results show that the learning capabilities of suppliers can be included in the models to increase effectiveness of action plans. In addition to that, buyer involvement strategies such as providing training and financial assistance to suppliers, enhance the sustainability gains of suppliers. It is found that sustainability levels compared to initial sustainability levels of suppliers can be increased up to 30% by optimizing the buyer involvement strategies. By considering models, suppliers sustainability improvement process can be organized more efficiently and managed more proactively by making use of buyer involvement strategies and including learning capabilities and behaviors of suppliers.

(4)

Executive Summary

Introduction Due to increase in division and global distribution of businesses, it is crucial to take suppliers into account as an important driver of company’s competitive power (Krause et al., 1998; Mol, 2003). This is especially the case when there is a limited opportunity of substitution. Supplier sustainability assessment and development is thus becoming more important for multi national organizations(MNOs) to work with suppliers which comply with the sustainability demands of the buyer companies. It must be always considered that outcome of sustainability efforts of MNOs depend mainly on the suppliers and their commitment (Orji and Wei, 2015). In the context of supplier development, supply chain learning can be referred.

However, there are very few quantitative works that examine learning in supplier development context, whilst modelling of learning provides insights on managerial actions (Jaber, 2011). Implementing supplier development actions is important for enhancing the suppliers’

performances. A supplier development action can be defined as an effort of a focal firm collaborate with its suppliers to enhance the performance and/or competence of the supplier and to meet needs of the focal firm. Depending on the learning capabilities of suppliers, supplier development actions can be optimized not to lose investments due to timing and risks.

Problem Statement The study aims to analyze the sustainability learning behaviors of suppliers by considering their characteristics and to incorporate those aspects into decision making processes while forming improvement action plans for suppliers. The sustainability learning behavior is referred as the evolution of suppliers’ sustainability levels throughout the time. Un- derstanding the sustainability learning behavior is important for buyer firms as it can reveal the degree of how much a supplier is willing to collaborate towards the sustainability practices.

Therefore, investigating the potential sustainability learning behavior in suppliers is important for providing tailor-made support for enhancing the sustainability levels of the suppliers in an efficient manner. By considering sustainability learning behavior of suppliers, action plans for suppliers can be provided in a way that the action plans are within suppliers’ capacities. In addition to that, two different buyer involvement strategy which are conducting on-site visit and directly investing in suppliers can be examined as additional decisions. As the buyer involvement strategies result in various costs to buyer firms, it is important to decide which buyer involvement strategy provides the highest sustainability gain for the suppliers. Therefore, (i) relationship between learning behavior of suppliers and supplier characteristics examined, (ii) learning process of suppliers modeled by using learning curve models and parameters of the learning curve are predicted using the supplier characteristics and (iii) increasing the effectiveness of supplier sustainability improvement process by including the information related to learning behavior of suppliers while providing action plans to suppliers and choosing buyer involvement strategies for suppliers. Consequently, theoretical models are applied to Philips’

data.

(5)

Identifying the learning behavior of suppliers using suppliers’ characteristics To examine the learning behavior of suppliers, the significant learning patterns are found by using hierar- chical clustering algorithm. After understanding the significant learning patterns among the suppliers, the relationship between the learning behaviors and the supplier characteristics is investigated. Supplier characteristics refer to starting score of the suppliers, general information in self-assessment questionnaires(SAQ) such as facilities, activities performed by suppliers, whether a company is MNO or not, assessor information. Dependent variable on the other hand are learning behavior of suppliers for each topic. In order to construct the classification model, a sophisticated machine learning model (XGBoost) is selected. For interpreting the effects of features on classifications, Shapley Additive Explanation model is used (Lundberg et al., 2018).

For each topic, a model is constructed, and the effects of features on learning behavior are determined. Results show that the suppliers that have lower initial scores are expected to be a fast learner as they have more improvement gap, and they can learn fast. There is also a relationship between learning behavior and activities performed at suppliers. Suppliers who perform logistics and distribution activity are more likely to be fast learner suppliers. Logis- tics and distribution suppliers are expected to care more about sustainability issues as there are already sustainability developments in the logistics industry such as ’green logistics’ and

’inverse logistics’. That is why they are expected to improve more and faster than other suppliers in SSP program. Suppliers who are performing sub-assembly activity are expected to be fast learner suppliers in health and safety topic. As the activity can be considered as a less risky activity, it might be easier for them to improve the issues that may arise. Suppliers who perform metal stamping activity are more likely to be indifferent in health and safety topic. As the activity considered as risky, suppliers may take more time to solve health and safety issues.

Suppliers who perform clean room activity are considered as slow learners in health and safety and human capital topics. That is because clean room activity suppliers are regarded as high mature suppliers, and they have a high initial score. That is why they cannot increase their sustainability scores fast. Suppliers who have a chemical warehouse on their site are more likely to be indifferent suppliers on business ethics topic. It can result from that managing a chemical warehouse may not be easy in terms of business ethics. Therefore, it may take more time to improve issues that arise from having a chemical warehouse. Having a hospital in the suppliers’

site increases the likelihood of being a slow learner supplier as it can be seen as a sign of high maturity. Consequently, those suppliers have less room for improvement and their evolution of sustainability score can be less than other suppliers because of that reason. Certificates also have a relationship with the learning behavior of suppliers. Having ISO 14001 increases the chance of being a fast learner supplier in environment topic as the certificate is closely related to environmental issues. Worker distribution variables also affect learning behaviors on various topics. Suppliers with higher management and staff employees have a higher level of the corporate structure. That is why those suppliers are expected to be fast learner suppliers. Be- ing a part of a multi-national organization(MNO) is also another critical factor for being a fast

(6)

learner supplier in human capital topic. Being a part of MNO brings suppliers more training and education to their workforce. Consequently, those suppliers are expected to learn faster in human capital topic.

Prediction of Suppliers’ Learning Curves After investigating the learning behavior of suppliers, their learning process is modeled by using a learning curve model. For that purpose, the most appropriate learning curve model (Time Constant Learning Curve Model) is selected for Philips as the parameters of the model satisfy the requirements of the company. By using non-linear least square method, parameters are estimated from the historical data of suppliers.

Different models are constructed with machine learning models (XGBoost regressor, random forest regressor) while considering different learning behaviors among suppliers. After comparing their performances, the selected model is implemented for predicting the learning curve of suppliers. A tool that can classify learning behaviors and learning curves of suppliers for each topic is created and applied at Philips.

Structuring the sustainability improvement process by incorporating suppliers’ learning behaviors Optimization models are constructed to increase the effectiveness of the sustainability improvement process. As Philips do not conduct any analysis for suppliers learning capabilities, the estimation of suppliers’ capabilities is not structurally defined. As the suppliers’ learning capabilities are estimated from research question 1(RQ1) and research question 2(RQ2), they are included in the model as capacities of suppliers. There are also two different buyer involvement strategies(conducting on-site visits and investments to suppliers) discussed in the models. Their effects are incorporated as an increase in the learning capacities of suppliers’. Two different models are compared. The model with learning behavior and both buyer involvement strategy optimizes how many days to conduct on-site visits for suppliers as providing guidance and how much monetary unit to allocate for each supplier as a support for their sustainability efforts by considering the suppliers’ learning capabilities. Results show that 19.7% increase in sustainability scores compared to initial sustainability scores of the suppliers can be achieved by only optimizing both buyer involvement strategy and incorporating the suppliers’ learning capabilities. In addition to that, by increasing resources allocated to suppliers by Philips, the sustainability levels can be increased up to 30% compared to initial sustainability levels of the suppliers.

Conclusion The research demonstrates several methods to increase the effectiveness of the supplier sustainability process by considering suppliers’ learning behavior and buyer involvement strategies. In conclusion, while providing action plans for suppliers, learning behavior and learning capabilities of suppliers obtained from the results of RQ1 and RQ2 can be applied.

By doing so, suppliers do not experience action plans that are not within their resource limits.

Secondly, optimization can be implemented for buyer involvement strategies. As buyers can provide support to their suppliers with training and investments, the degree of support to each supplier is vital to understand the optimal solution in the system. The research showed that

(7)

both buyer involvement strategies could be an effective tool to increase the effectiveness of the supplier sustainability improvement process as optimizing the both buyer involvement strategy provides the highest sustainability improvement percentages compared to initial scores during the analysis. In addition to that, increasing the collaboration degree with suppliers by increasing the resources allocated for each supplier, sustainability improvement percentages can be increased until the system limit which refers the maximum learning capacities of the suppliers. The findings can be further investigated by implementing the theoretical models to other companies who perform supplier sustainability assessment to its suppliers. By doing so, the effects can be tested. Furthermore, as other companies may have different information related to supplier characteristics, the relationship between additional supplier characteristics and the sustainability learning behaviors can be revealed. In addition to that, other companies can enhance their supplier sustainability improvement process if the decision models mentioned in the study are carefully implemented by optimizing the buyer involvement strategies and considering sustainability learning behaviors while providing the action plans.

(8)

Preface

As this thesis means the end of my master’s study at Eindhoven University of Technology, I would like to thank all the people who have supported me during this adventure.

Above all, I would like to express my gratitude to my first supervisor, Dr. Tarkan Tan. It was a pleasure to have such a great supervisor like him. He was always there to help in our biweekly meetings. He was always clear and direct about his feedbacks, which helped me to keep the project on the track. His comments always pushed me to think outside the box. Thank you for everything. I would like to thank Dr. Shaunak Dabadghao for being my second supervisor. At the critical moments, he contributed to the project, and prevented a groupthink problem. He always asked the right question to remind me what the project is about and what the ultimate goal is. I also would like to thank Dr. Zumbul Atan for being my third supervisor and providing your feedback at the very last moment. I also truly appreciate my company supervisors, Marco Baren and Dylan McNeill. Thanks Marco for your lessons about how academic life and business life are different from each other, how to make an effective presentation and how to manage people effectively. I will, indeed, remember these precious lessons that touched my life. Thanks Dylan for being such a nice supervisor and always supporting me. It was great to work with you. You contributed to project with your strategic thinking and helpful feedbacks in our meetings. I also would like to thank Hakan Akyuz for his contribution to the thesis.

Your feedbacks at the critical stages always helped me to carry on.

I also would like to thank my friends. Thanks, Fercan, Seba, Adelayda and Yasey for sharing wonderful moments with me during my study. I am so glad to have you in the last 2 years. I will miss our gatherings. Thanks, Mahir, Toycar and Mahmut for cheering me up no matter the distance especially during the most stressful days of the thesis. Also, I appreciate providing me your opinions on my work in return of some calamari. Thank you all for spending some hours to proof read my thesis. I am also grateful to my aunts and my cousin who helped me tremendously to adapt a new country and a new culture. Finally, I would like to thank my parents, Ibrahim and Nezihe Kayhan, and my brother, Onder Kayhan, for their continuous support and always believing in me.

At an inn with only two doors, we all keep walking day and night. This journey was just a stressful and valuable part of the road. I am so lucky to have you all in my life and I again would like to thank all of you for being a part of this interesting and great experience!

Mert

(9)

CONTENTS

List of Figures

1 Frame of reference of Philips’ Sustainability Agreement . . . 4

2 Sustainability Dashboard used at Philips . . . 4

3 Supplier sustainability classification at Philips . . . 5

4 Conceptual framework constructed for RQ1 . . . 14

5 Illustration of Fast Learner Suppliers . . . 15

6 Illustration of Lagged Fast Learner Suppliers . . . 16

7 Illustration of Slow Learner Suppliers . . . 16

8 Illustration of Indifferent Suppliers . . . 17

9 Illustration of Forgetful Suppliers . . . 17

10 Illustration of Sinusoidal Suppliers . . . 17

11 An example dendrogram for illustration purposes . . . 18

12 An example of applied conceptual framework to data on hand . . . 20

13 Dendrogram as an output of hieararchical clustering for overall score . . . 24

14 Conceptual Framework for RQ2 . . . 36

15 The applied framework to construct prediction models . . . 38

16 Illustration of learning curves of lagged fast learner suppliers . . . 40

17 Relationship between allocated capacity and obtained improvement . . . 50

18 Relationship between duration of on-site visit and capacity increase . . . 51

19 Effects of sequence number on maximum learning capacity increase . . . 52

20 The relationship between investments and learning capacity increase . . . 56

21 Improvement percentages for each topic and for each supplier in the base model 61 22 Improvement percentages for each topic and for each supplier in the model with sustainability learning behaviors . . . 62

23 Deviation of improvement percentages when the learning behavior is included . 62 24 Improvement percentages amount for each topic and for each supplier in the model with optimizing buyer involvement strategies . . . 63

25 Sensitivity analysis of buyer’s budget on the total improvement percentage . . . 63

26 Sensitivity analysis of on-site visit capacity on total improvement amount . . . . 64

27 Sensitivity analysis of on-site visit capacity on the total improvement percentage 64 28 Results of different set of cost parameter for predecessor action . . . 65

29 Dendrogram of sequence 4 suppliers in environment topic . . . 82

30 Dendrogram of sequence 4 suppliers in health & safety topic . . . 82

31 Dendrogram of sequence 4 suppliers in business ethics topic . . . 83

32 Dendrogram of sequence 4 suppliers in human capital topic . . . 83

33 SHAP summary plot of data-set from NHANES I (Miller, 1973). . . 91

34 SHAP summary plot of Lagged Fast Learner Suppliers in Validation (Overall) Score . . . 92

35 SHAP summary plot of Fast Learner Suppliers in Validation (Overall) Score . . . 92

36 SHAP summary plot of Slow Learner Suppliers in Validation (Overall) Score . . 93

(13)

LIST OF FIGURES

37 SHAP summary plot of Indifferent Suppliers in Validation (Overall) Score . . . . 93

38 SHAP summary plot of Lagged Fast Learner Suppliers in environment topic . . 94

39 SHAP summary plot of Fast Learner Suppliers in environment topic . . . 94

40 SHAP summary plot of Slow Learner Suppliers in environment topic . . . 95

41 SHAP summary plot of Indifferent Suppliers in environment topic . . . 95

42 SHAP summary plot of lagged fast learner suppliers in health & safety topic . . 96

43 SHAP summary plot of fast learner suppliers in health & safety topic . . . 96

44 SHAP summary plot of slow learner suppliers in health & safety topic . . . 97

45 SHAP summary plot of indifferent suppliers in health & safety topic . . . 97

46 SHAP summary plot of lagged fast learner suppliers in business ethics topic . . 98

47 SHAP summary plot of fast learner suppliers in business ethics topic . . . 98

48 SHAP summary plot of slow learner suppliers in business ethics topic . . . 99

49 SHAP summary plot of indifferent suppliers in business ethics topic . . . 99

50 SHAP summary plot of lagged fast learner suppliers in human capital topic . . . 100

51 SHAP summary plot of fast learner suppliers in human capital topic . . . 100

52 SHAP summary plot of slow learner suppliers in human capital topic . . . 101

53 SHAP summary plot of indifferent suppliers in human capital topic . . . 101

54 Illustration of time constant learning curve model . . . 105

55 Layout of learning behavior classification tool . . . 110

56 Output of learning behavior classification tool . . . 110

57 Learning Capacities of Suppliers without considering learning behavior of suppliers . . . 111 58 Learning Capacities of Suppliers with considering learning behavior of suppliers 111

(14)

LIST OF TABLES

List of Tables

1 Example of a data structure for suppliers . . . 18

2 Example of calculating clustering features from example data . . . 18

3 Possible independent variables according to the literature . . . 19

4 An example of applied conceptual framework with different learning behaviors and supplier characteristics for each topic . . . 21

5 Data size for each subset and explanation of subsets . . . 21

6 Supplier characteristics that are used for classification model . . . 22

7 Amount of suppliers that have specific learning behavior in each topic . . . 24

8 Amount of average sustaianability evolution of suppliers between each sequence . . . 24

9 Amount of suppliers that have specific learning behavior in each topic after including subset 2 . . . 25

10 Amount of average sustainability evolution of suppliers between each sequence after including subset 2 . . . 25

11 Performance of the classification model for each topic . . . 27

12 Comparison of performances between binary and multi-class models . . . 28

13 Activities that affect learning behavior of suppliers in each topic . . . 30

14 Facilities that affect learning behavior of suppliers in each topic . . . 30

15 General Information that affect learning behavior of suppliers in each topic . . . 32

16 Effects of independent variables on overall score . . . 33

17 Interaction effect between independent variables for validation score . . . 34

18 Analysis of data used for RQ2 . . . 39

19 Dependent variables for each learning behavior . . . 40

20 Performance of models for predicting improvement potential parameter . . . 41

21 Performance of models for predicting learning rate parameter . . . 42

22 Performance of models for predicting initial score parameter . . . 42

23 Assumptions of the models . . . 49

24 An illustration of converting categorical variable into binary variables -1 . . . 85

25 An illustration of converting categorical variable into binary variables -2 . . . 85

26 An example of Contingency table to calculate phi coefficient . . . 86

27 Important variables for the models and their explanations . . . 89

28 Effects of features on each class for overall score . . . 102

29 Interaction effects of features on each class for overall score . . . 102

30 Effects of features on each class for environment topic . . . 103

31 Interaction effects of features on each class for environment topic . . . 103

32 Effects of features on each class for health & safety topic . . . 103

33 Interaction effects of features on each class for health & safety topic . . . 103

34 Effects of features on each class for business ethics topic . . . 104

35 Interaction effects of features on each class for business ethics topic . . . 104

36 Effects of features on each class for human capital topic . . . 104

(15)

LIST OF TABLES

37 Interaction effects of features on each class for human capital topic . . . 104

38 Parameter set for hyper parameter tuning of XGBoost classification model . . . 108

39 Parameter set for hyper parameter tuning of XGBoost regressor model . . . 108

40 Parameter set for hyper parameter tuning of random forest regressor model . . . 109

(16)

1 Introduction

Sustainability as a topic is getting popular as multinational companies understand its importance for businesses that they are operating in. In the recent years, many undesired incidents happened in the supply chain of those multinational organizations (MNO) especially in the upstream parts of their supply chains. First example of those incidents occurred in Mattel which obliged the company to withdraw huge number of toys since traces of lead were detected in the paints (McIlroy, 2007). Another example can be given for Zara as the customers of the company found out some notes sewn into their clothes. The notes that were sewn by employees stated that they were not getting paid (Young, 2017). In 2013, more than 1,000 people died in Bangladesh due to collapse of Rana Plaza in Bangladesh, and those people were employees of suppliers for large clothing companies (Kerppola et al., 2014; Lee and Rammohan, 2017).

These examples can be increased where non-compliant attitudes of upstream of a supply chain to codes of conducts of the sustainability programs. In addition to that, 74% of carbon foot- print of an industry results from upstream echelons’ emissions (Matthews et al., 2008). There- fore, multinational companies have started to focus on increasing sustainability performance of their suppliers in terms of compliance to their sustainability programs, codes of conduct and agreements. Due to increase in division and global distribution of businesses, it is crucial to take suppliers into account as an important driver of company’s competitive power (Krause et al., 1998; Mol, 2003). This is especially the case when there is a limited opportunity of substitution. Supplier sustainability assessment and development is thus becoming more important for MNOs to work with suppliers which comply with the sustainability demands of the buyer companies. It must be always considered that outcome of sustainability efforts of MNOs depend mainly on the suppliers and their commitment (Orji and Wei, 2015). In the context of supplier development, supply chain learning can be referred. However, there are very few quantitative works that examine learning in supplier development context, whilst modelling of learning provides insights on managerial actions (Jaber, 2011). Depending on the learning capabilities of suppliers, supplier development actions can be optimized not to lose investments due to timing and risks.Implementing supplier development actions is important for enhancing the suppliers’ performances. A supplier development action can be defined as an effort of a focal firm collaborate with its suppliers to enhance the performance and/or competence of the supplier and to meet requirements of the focal firm. By implementing supplier sustainability development action, buyer firms can benefit from various aspects such as creating a positive company image, increasing product quality and increasing profit and decreasing cost as a result of sustainability development actions. Therefore, sustainability improvement actions should not be seen as a dead money investment. On the contrary, if the actions are implemented properly, it can bring several profit increase opportunities with an increase in the sustainability level.

The purpose of this research is to analyze the sustainability learning behaviors of suppliers by considering their characteristics and incorporate those aspects into decision making processes

(17)

1.1 Company Introduction while forming improvement action plans for suppliers. The sustainability learning behavior can be addressed to evolution of suppliers’ sustainability levels throughout the time. As it is expected, it can be different for each supplier. For example, a supplier may learn faster than other whereas a supplier may not learn but instead it may forget its sustainability experience.

Therefore, understanding and incorporating the sustainability learning behavior of suppliers is important to increase the effectiveness of decision making process as it can provide useful insights on how much a supplier is willing to collaborate for sustainability. However, investigating learning behavior in the context of supply chains is not well understood (Bessant et al., 2003; Childerhouse and Towill, 2003; Towill et al., 2000) and (Jaber, 2011) states that learning effects in supply chain can be a great contributor to direct managerial actions, and supply chain learning can be managed by strategically managed buyer-supplier relationships (Carr and Pearson, 1999). The drivers of supplier sustainability level were investigated by some papers (Locke et al., 2007; Toffel et al., 2015; Short et al., 2016), but revealing factors that affect sustainability learning behavior of suppliers is not well investigated, which is one of the aim of this research. In addition to that, developing a decision-making model to increase sustainability improvement of suppliers by incorporating different sustainability learning behavior of suppliers will be another contribution after investigating sustainability learning behavior of suppliers. In the decision making process, several buyer involvement strategies such as providing training to suppliers in order to increase know-how and experience of suppliers, supporting financial efforts of suppliers by directly investing into them is considered. Buyer companies can greatly enhance supply chain sustainability by using collaboration strategies by taking a more thoughtful approach with including different sustainability learning behaviors while selecting strategies for collaboration, their choice of suppliers for collaboration and the way they organize their collaboration efforts.

1.1 Company Introduction

Philips is one of the pioneer companies that started auditing its suppliers’ sustainability performance. In 2004, Royal Philips (hereafter: Philips) decided to assess sustainability performance of their suppliers. Those assessments were done by third party companies with supplier sustainability audits. Philips refuses the idea that audits reveal the true sustainability levels of the suppliers as those audits are conducted via ‘violation-based’ methods. Philips decided to shift towards a more structured method where trustworthiness and continuous improvement are more important. In 2016, Philips started “Beyond Auditing” approach to measure sustainability performance of the suppliers by using self-assessment questionnaires (SAQ) under “Supplier Sustainability Performance” program (Philips, 2018). In the program, suppliers are promoted not to hide the information about current circumstance of their sustainability performance instead of evaluating with “Pass-Fail “criteria. In return, Philips provides support for improvement of its suppliers, regardless of the situation. Thus, the aim is increasing transparency, commitment and meeting agreed targets.

Philips is trying to make “Beyond Auditing” approach more proactive instead of reactive. The

(18)

1.2 Data Description current process in SSP(Supplier Sustainability Improvement Plan) program has 6 steps. First, Self-Assessment Questionnaire (SAQ) is sent to the suppliers to answer the questions and provide evidences such as documents or pictures for their responses if necessary. Then, responses to questionnaire are assessed by Philips, and sustainability scores of each supplier are revealed.

According to revealed scores, each supplier receives an improvement plan. It can be the case that a supplier has an on-site visit to assess the issues more deeply. The aim of Philips is shifting the program into proactive side by predicting sustainability scores of suppliers beforehand. By doing so, Philips can optimize its actions for their suppliers such as on-site visits and providing action plans. Detailed explanation for the steps of SSP program can be found in § 1.2.

So far, relationship between SAQ and the final sustainability scores were analyzed and the set of most important questions for final scores was determined (Smouter, 2018). Secondly, consistency between answers of SAQ and provided evidence pieces was investigated, and a supplier sustainability classification framework was developed(Scholte, 2019). Thirdly, root causes of non-compliance were analyzed (Wismeyer, 2019), and prediction models for sustainability scores were studied (Zapata, 2019). Lastly, improvement potentials of suppliers were revealed by using supplier characteristics (Urlu, 2019).

1.2 Data Description

After the introduction of “Beyond Auditing” approach in 2016, suppliers fill self-assessment questionnaires (SAQs), provide evidence information such as pictures or related documents and are further visited for a site assessment (Philips, 2018). SAQ that are sent to suppliers has approximately 700 questions. It is designed in a way that some questions can pop up depending on the answers for previous questions. SAQ has seven sections. First two sections are for introduction and obtaining general information. Other five sections are for sustainability performance and correspond to five topics of sustainability determined by Philips’ requirements, international standards and Responsible Business Alliance (RBA) code (Philips, 2018). Each section has 9 subtopics to determine sustainability maturity level of the suppliers. The frame of reference can be seen in Figure 1.

SAQ has 36 unique fields meaning one field for each topic and subtopic (maturity level). Each question in SAQ belongs to one unique field and has a specific weight for calculating the weighted scores which can be seen in Figure 2. This approach is also similar with validat- ing SAQ using evidence pieces provided by suppliers such as pictures and related documents.

In the case of validation, whether the document is provided or not is examined first, and if it is provided, they score 1 and 0 vice versa. However, if the evidence is not applicable for that kind of supplier, the score is determined as N/A. In addition to that, if a non-compliant behavior can be seen from an evidence piece, the score is determined as PZT (Potential Zero Tolerance).

After that, depending on the quality of an evidence piece, it can get a score of 0, 0.5 or 1 for piece of evidence meaning invalid, partially valid or valid evidence provided respectively. By using those scores, validation score is calculated in a separate dashboard by using the same weights shown in Figure 2.

(19)

1.2 Data Description

Figure 1: Frame of reference of Philips’ Sustainability Agreement

Figure 2: Sustainability Dashboard used at Philips

To calculate final scores, the scores obtained from SAQ dashboard and Validation dashboard are used. Weights of SAQ dashboard and Validation dashboard were 30% and 70% respectively until 2018. However, after 2019, SAQ dashboard has 0% and validation has 100% meaning that SAQ dashboard contributes only in collecting information.

After realizing the final validated score, suppliers are classified in different categories. Those are Best in Class (BiC), Do It Yourself (DIY), SSP (Supplier Sustainability Improvement Plan) and No Zero Tolerance (NZT). Different characteristics for different categories can be seen in Figure 3. The most common categories of Philips’ suppliers are DIY and SSP. If a supplier has PZT during validation phase, it is directly categorized as SSP. DIY suppliers are the ones considered to have enough maturity level for managing their improvement areas and are not visited on-site. On the other hand, SSP suppliers have lower sustainability maturity and are visited on-site to check evidence pieces and determine improvement actions. After getting the final scores and taking expert opinions, focus areas of improvement are decided for each supplier. SSP suppliers implement the improvement actions under supervision of Philips whereas DIY suppliers implements by themselves as they are considered mature enough. In the study, the focus and the scope is only investigating SSP suppliers as they are the suppliers who have the most attention from Philips in the context of supplier sustainability development.

(20)

1.3 Problem Description

Figure 3: Supplier sustainability classification at Philips

1.3 Problem Description

As there were already multiple projects conducted recent years thanks to SSP, the research will focus on making use of the information collected from suppliers to have a better understanding in their sustainability learning behavior and to utilize obtained understanding in decision making process for optimizing the buyer’s effort. Ngwenyama et al. (2007) state that timing of the development effort is important for maximizing organizations’ productivity gains and using learning pattern can be the theoretical base for determining the timing. After understanding the sustainability learning behavior of suppliers, the buyer firm can have insights about how much a supplier is willing to allocate effort for increasing its sustainability level. Therefore, the objective of this research is to gain understanding of learning behavior of Philips’ suppliers on their sustainability scores and to develop a mechanism that supports sustainability assessment process and decision making on improvement strategies. The decision making includes providing action plans to suppliers within their capabilities and using buyer involvement strategies to increase sustainability improvements of the suppliers. Therefore, the trade-off is between having the highest sustainability improvement while using the budget allocated to suppliers by the buyer as effective as possible. As the buyer involvement strategies such as providing training and investing directly in suppliers need budget, it is important to understand when and at what degree the buyer should use the buyer involvement strategies as collaboration with suppliers.

1.4 Research Question

Research questions (RQs) are formed according to the problem description explained in § 1.3.

Although this thesis project is practice based, RQs are formed in a way that methodology can be followed by all companies that implement supplier sustainability assessment program.

• RQ1: What approach can be used to identify learning behavior of suppliers based on their characteristics?

After obtaining the suppliers’ characteristics from SAQ and SAQV and evaluating the sustainability score, the next step is to classify the suppliers’ learning behavior according to informa-

(21)

1.4 Research Question tion obtained. Currently there is no knowledge on using learning behavior to estimate evolution of performance. In addition to that, there can be relation between supplier characteristics and learning behavior. Farghal and Everett (1997) state that the prediction of performance is most convenient when it is necessary for estimating, scheduling or other project management necessities. At Philips, it is important as optimizing the on-site visit schedules is a vital part of the improvement process. Looking ‘behind the learning’ from the perspective of supplier development initiatives is important to understand learning performance and relational gain of both parties (Gosling et al., 2018). Therefore, the goal is to create a classification methodology on how to determine which supplier characteristics are important while investigating the learning effect on sustainability. By doing so, the features that are the main drivers of learning behavior will be revealed. The deliverable result from this research question is a structured classification method on how to categorize suppliers’ learning behavior by looking at their characteristics in order to increase understanding of sustainability score evolution through the sequences.

• RQ2: What method can be used to predict sustainability learning behavior of suppliers incorporating the factors that affect their learning behavior?

The research question contributes how to predict sustainability learning behavior by using the information gained in Research Question 1. Lall`e et al. (2015) state that it is possible to predict learning curves as users acquiring new skills, tailor made support could be provided to enhance the performance and engagement of users by using those predictions. Therefore, the goal is to create a methodology to predict sustainability learning behavior for suppliers by using the output obtained in RQ1 so that it can be used in further decision-making processes.

The deliverable result from this research question is a predictive model on how suppliers’

sustainability learning curves can be predicted by only looking at specific features.

• RQ3: How should buyer’s effort be optimized by determining on what buyer firm should spend its effort while considering suppliers’ learning behaviors and supplier characteristics so that the improvement process be more effective and structured?

The research question contributes to structure decision making process for improvement action plans by incorporating learning behaviors of the suppliers and determined features that have an influence. What actions can be taken depends on the learning behaviors of suppliers.

Mizgier et al. (2017) claim that as buyers have limited resources to invest in supplier development and the returns from those investments are not certain meaning that they are risky, capital allocation must be done with informed decisions by considering risk. In Philips case, informed decisions can be made by including suppliers’ learning behavior to structure the effectiveness of supplier development. While considering management of supply chain in terms of social, environmental and economic aspects for creating strong corporate image, complexity is one of the issues as companies should pay attention to all aspects at the same time. (Resat and Unsal, 2019). Therefore, multi objective optimization problems might be considered to integrate sustainability in Philips’ case, and the goal is to create a methodology to design a

(22)

1.5 Outline decision-making process in supplier development by considering the supplier characteristics that affect learning behavior significantly. Buyer firms can provide ideal support for each supplier to enhance sustainability development by incorporating additional information revealed by supplier characteristics instead of providing same type of support for each supplier without considering that they have different organizational learning capacity. By doing so, buyer firms can determine how and on what they should spend their effort beforehand. Effect of including the new information on decision making process in supplier development can be compared using Philips data as well. The deliverable result is a mathematical model on how buyer’s effort can be optimized to make improvement process more efficient and customized for suppliers by making use of information gained from RQ1 and RQ2.

Above mentioned RQs are related to each other as the aim of RQ1 which is identifying the learning behavior of suppliers and its relationship between supplier characteristics can be used as an input for RQ2 which is about predicting the learning curves of suppliers. It is because the learning behavior of suppliers also closely related to learning curves of suppliers. That is why information gained in RQ1 can be used in RQ2. Likewise, learning curves of suppliers can be incorporated in RQ3 which is about providing decision support model for the company in supplier development. As the learning curves of the suppliers reveal important information related to suppliers capacities’, output of RQ2 can be used in RQ3 as a parameter for suppliers.

1.5 Outline

The report starts with a literature review in § 2. In § 3, the first research question is discussed by explaining a theoretical model in § 3.1, and an applied model at Philips demonstrated in § 3.2.

The same procedure is also followed for research question 2 and 3. In § 4, the second research question is discussed with a theoretical model in § 4.1 and an applied model at Philips in § 4.2.

For the last research question, explanations can be found in § 5 with a theoretical model in

§ 5.1 and an applied model in § 5.2. Discussion of the results can be found in § 6. Overview of the implementation provided in § 7. Conclusions of the study can be found in § 8. Lastly, limitations of the study and future research directions are discussed in § 9.

2 Literature Review

With the increased attention of sustainability in academic literature, organizations pay more attention to sustainability while managing their supply chains by considering the role of suppliers in upgrading to a sustainable supply chain. Carter and Rogers (2008) define sustainable supply chain management as “the strategic, transparent integration and achievement of an organization’s social, environmental, and economic goals in the systemic coordination of key organizational business processes for improving the long-term economic performance of the individual company and its supply chains”. There are challenges while implementing sustainability into supply chains as it requires a holistic approach to benefits in terms of social and environmental and the costs such as long-term investment commitments. It also results in pres-

(23)

sure coming from consumers due to increased level of transparency in supply chains (Carter and Rogers, 2008). Lee and Rammohan (2017) suggest different reasons why to improve sustainability in supply chains. Firstly, As consumers prefer green products more in the recent years, it boosts the demand for such products. Another reason is that harmful activities in supply chains in terms of sustainability get more attention by media, and organizations are pun- ished more severely in such cases. Furthermore, Non-governmental organizations start step in more often to enhance the process of increasing sustainability efforts by the organizations.

Lastly, governmental actions such as regulations and prevention laws enforce organizations to focus more on sustainability in their supply chains. Walker et al. (2008) also categorize sustainability motivators for organization as internal and external motivators. External motivators which enhance the organizations processes proactively shift towards improving sustainability in their supply chains consist of market drivers such as customer demands and competition, government regulations and legislations and social factors in terms of societal organizations and media (Chkanikova and Mont, 2015). Economic optimization and business risk management are considered as internal motivators. Management support plays a vital role for en- suring the successful implementation of sustainability programs to supply chain management activities (Pagell and Wu, 2009).

Matthyssens and Faes (2013) states that transparency is a critical factor for embedding sustainability into supply chain especially for the upper part of supply chains. As transparency is a challenging task due to the need of collaboration of numerous stakeholders within supply chain, monitoring and evaluation approaches can be considered as effective methods to track suppliers’ efforts while enhancing transparency. Tracking suppliers’ efforts provide a basis for buyers’ firm to understand willingness of suppliers on complying with buyers’ codes of conducts in sustainability procedures (Van Weele and Van Tubergen, 2017). As economic and environmental dimensions are dominant in the researches on sustainable supply chain management, researchers might pay more attention to the social and human aspects of sustainability to form a systematic sustainability research in supply chain management (Panigrahi et al., 2019).

In 2017, Van Weele and Van Tubergen (2017) reported that sustainability criteria and practices enforced by buying organizations from their suppliers would promote supplier creativity and sustainability. This will also have advantages for buying organizations. Accordingly, buying organizations are introducing supplier development initiatives to create a specific approach to supply chain sustainability. MNOs commonly start by asking their suppliers to sign codes of conduct for sustainability. Since the early 1990s, this practice has risen from the USA, and the number of different codes of conduct has since increased drastically (Andersen and Skjoett- Larsen, 2009). Commonly, codes of conduct include subjects such as critical attention areas, compliance with policies and standards of the industry, and compliance with laws of government (Kashmanian, 2015). Regardless of the kind of code of conduct signed by suppliers, MNOs have to monitor their suppliers to follow the codes of conduct as experience has shown that signing codes of conduct do not mean that the suppliers will follow them (Van Weele and

(24)

Vivanco, 2014). Buyers will collaborate with suppliers and put in place supplier growth initiatives to improve compliance (Van Weele and Van Tubergen, 2017). Sustainable practices of suppliers are typically tested by conducting audits, either by third parties or the buying organization. A standard audit contains four parts for collecting the evidence. Those four parts are management interviews, staff interviews, facility tours, and reviewing records, according to Kashmanian (2015). There is, however, an intense dispute as to whether those who perform audits are trustworthy, reliable and competent enough to conduct successful audits (e.g., Locke et al. (2007); Toffel et al. (2015); Short et al. (2016)). If MNOs conduct internal audits, clash of interest arises because of the possible violations that could not be better known to the public. To some extent, this issue persists when third parties do audits, but another issue raised is the consistency and willingness of the auditing organization to carry out accurate and reliable checks in compliance with the buying organization’s code of conduct (Locke et al., 2007).

Audits have also shown to be potentially harmful because they result in less transparency on the part of suppliers (Distelhorst et al., 2017) and suppliers seeking to pass audits rather than address the fundamental issues (Jiang, 2009). Indeed, many different codes of conduct and protocols also exist for the audit of suppliers. Suppliers, therefore, operate in an uncertain environment, where it is often difficult to know which parts to emphasize (Locke et al., 2007).

Supervising suppliers is also essential for providing them guidance and support. The development of suppliers is the method of working one-to-one with other suppliers to enhance efficiency for buying an organization’s benefit, and It is strongly tied to the management of the supplier relationship. Measures of supplier development may be adapted for a short period (e.g., to improve supplier delivery times) or focus for an extended period (e.g., to improve the supply base of the buyer strategically) and may have a direct or indirect character (Wagner and Krause, 2009). For direct interventions, the buying organization directly spends money in a supplier, e.g., for equipment provision, training programs, for live consulting, and tem- porary staff transfer or (Sucky and Durst, 2013; Prahinski and Benton, 2004; Wagner, 2006; Bai and Sarkis, 2011). For indirect supplier development, the buying organization takes on a pas- sive role, for example, by setting performance targets or improvement targets or by offering supplier(s) incentives (Wagner, 2006; Wagner and Krause, 2009; Sucky and Durst, 2013)). Effi- cient supplier development activities depend on various internal and external elements like the corporate strategies of the organizations, technical uncertainties, supplier’s capabilities, power distribution, or the length of the relationship between client and supplier (Bai and Sarkis, 2011;

Sucky and Durst, 2013).

The development of suppliers demands that both companies execute staff, financial, and capital resources to work, develop an effective method of monitoring performance, and communi- cate sensitive information timely. Therefore, for both sides, this method is challenging. Buyer managers and workers should be persuaded that investing in a supplier’s organization is a worthy risk. Supplier managers should be persuaded that their best interest is in following their client’s guidance and support. However, success is uncertain, even though the two organizations agree that the development of suppliers is significant.

(25)

Recently, increased number of organizations faced with detrimental practices in terms of sustainability issues in upper part of their supply chain. It is important to prevent those issues as it is relevant to 12% average decrease in market capitalization (Lefevre et al., 2010). Sup- plier sustainability risks are defined as ‘pure risks’ where obtaining a gain is not possible as a result of the event (Yates and Stone, 1992; Narasimhan and Talluri, 2009). When it comes to managing supplier sustainability risks, there are three approaches that can be found in the literature: risk avoidance, monitoring-based and collaboration-based risk acceptance strategies (Blome and Schoenherr, 2011; Ritchie and Brindley, 2007).Risk avoidance strategy is the cir- cumvention of a risky situation in order to eliminate risk (Juttner et al., 2003). In essence, it is targeted at removing the risk source by reducing the probability of the risk event to zero.

Thus, the supplier sustainability risk can be avoided when supply managers switch from their current risky supplier to another supplier with a better sustainability record. For instance, the largest global trader of palm oil, Wilmar International Ltd., resolved in June 2013 to stop doing business with suppliers from Indonesia, who were involved in using illegal fires to clear land for farming (Yun et al., 2013). The focus of monitoring-oriented risk mitigation technique is to evaluate the suppliers’ actual performance or processes against certain performance criteria or particular characteristics to check whether or not they meet the required standards (Jiang, 2009). This approach usually requires buyers to collect and examine the information of suppliers, select suitable criteria, utilize audits and surveys to evaluate the sustainability-related characteristics of inbound goods and their suppliers, and request a report from the suppliers in regard to various aspects of their ecological and social performance (Bowen et al., 2001;

Seuring and Muller, 2008). These are often implemented through contracts containing written ecological/social requirements (Ciliberti et al., 2008), which need third-party endorsements (Morali and Searcy, 2013), or/and subjecting the suppliers to the buyer’s code of conduct (An- dersen and Skjoett-Larsen, 2009). For example, Toyota, GM, and Ford requested ISO 14000 certification from their suppliers, while Xerox, IBM, and Bristol-Myers Squibb urged their Chi- nese suppliers to adopt ISO-14001-complaint environmental management systems (Zhu et al., 2007). Collaboration-oriented risk mitigation technique is based on improving the social and ecological performance of suppliers via partnership, which implies interacting with them directly and implementing mutually designed social and ecological solutions (Jiang, 2009). It deals with various activities such as encouraging suppliers to share their experience and information through sponsored social or ecological summits, offsetting costs due to their adherence (e.g., combined investments in equipment that are environment-friendly) and providing them with training programs (Vereecke and Muylle, 2006). IKEA, for example, attempts to improve the understanding and knowledge of its suppliers, especially those in developing countries, on sustainability-related issues in a bid to change their mindset. It also gives out loans to its suppliers to support their capital-intensive investments relating to sustainability, like the construction of wastewater treatment facilities at their production sites (Spence and Bourlakis, 2009).

When it comes to prediction of sustainability levels of suppliers, regression methods are quite

(26)

popular for companies to monitor supplier sustainability (Short et al., 2016; Toffel et al., 2015;

Locke et al., 2007). There are also studies that are performed with classification models (Yunxia, 2017; Caro et al., 2016). Here, a classification model tries to obtain some inference by using an input data. The ultimate aim is predicting the labels or categories for a new data. A regression model also tries to have conclusions by using the input data. However, the aim of the regression model is predicting the value of the new data by making point forecast. Short et al. (2016) used a negative binomial regression model, comparing their results to a Poisson-regression model. The assessment is used to analyze the amount of breaches reported during each audit and measure the extent to which companies follow codes of practice for determining the number of infringements in each audit. Hence, the negative binomial regression model has chosen by them. Toffel et al. (2015) also used the same model (negative binomial regression model) in order to look for evidence for the importance of variables in accordance with suppliers’ la- bor standards. Another study, Rabbi et al. (2020), is conversely used Bayesian belief network in green supply chain to foresee the several parameters. They explained how this method is used to show causality and conditional probabilities between a number of factors. Moreover, if dependent variables are probabilistic, this method is practicable and convenient.

Vahdani et al. (2012) introduced a linear neuro-fuzzy model for supplier evaluations within the cosmetics industry. They began by choosing the most suitable criteria for evaluating suppliers. Subsequently they collected data-set on a numeric scale for criteria and performance evaluation. Once they had compiled historical attributes and performance information, the researchers divided the resulting data sets into two parts to implement the proposed model and test its predictive potential. In order to visualize the model’s estimation capacity, results were compared to those results obtained from a Radial Basis Function (RBF) neural network, a Multi-Layer Perceptron (MLP) neural network and a Least Square-Support Vector Machine (LSSVM).

Besides regression models, classification models are used for supplier selection and sustainability literature. Caro et al. (2016) has used a regression model, using multi-level random effects, to test the importance of factors supposedly influencing unauthorized subcontracting by outsiders within the textile industry. The correlation between indexes and measurement complexity becomes greater as supplier evaluation has tended to be more diversified, com- prehensive, and refined. To ensure that the assessment can be kept simpler for credibility and reasonability of evaluation is another major concern. Yunxia (2017) states in their study that supplier evaluation is the process of dividing the supplier’s ability according to a rating system. It can therefore be considered a classification problem which requires both huge dimensions and a wide range of capabilities. Given the strong advantage of Random Forests in classification and choice of traits, Yunxia (2017) has used a random forest as a place to assess suppliers in respect to the classification problems.

The development of suppliers is critical because of the complexity of current supply chains and business globalization. As buyers only have limited resources, they need to make calculated decisions about developing their suppliers. Also, the investment profit in the development

(27)

of suppliers is unpredictable, so buyers should consider this risk when selecting their development initiatives suppliers. In their paper, Mizgier et al. (2017) suggested a multi-objective model for allocating capital for low-risk supplier development. They implemented it to an international car manufacturer’s example and help the decision-making process with Bloomberg database. They used suppliers’ stock returns and capital costs to determine the efficiency of the suppliers. Their model explores the trade-offs that are present between supplier development risk and expenses. By evaluating the case of Toyota and its suppliers, Mizgier et al.

(2017) demonstrated that the buyer’s risk aversion plays a vital part in providing supplier development capital allocation. The higher the buyer’s risk-averse, meaning that greater focus is given on preventing huge losses, the more heterogeneous the supplier allocation will result in a more diversified supply base. On the other hand, as the buyer becomes more concentrated on capital costs, the least expensive suppliers are selected, and risk aversion no longer plays a part. In short, provided the buyer’s choice, it can consider a range of suitable solutions of the issue and determine where its supplier development efforts should be centered as per its risk-taking tendency. As buyers have a limited budget for efforts to develop suppliers, which need to be distributed between many suppliers optimally. Consequently, the budget invested in a supplier depends on the budget spent on other suppliers. Additionally, the profits vary between suppliers based on their capacities and implementation expertise. In their papers, Talluri et al. (2010) suggested a risk management model for supplier development programs.

The buyer’s objective in their model is to assign amounts of investment to accomplish aimed return at the least risk. The risk is influenced by the variation in the total amount invested and the profits of the supplier. Talluri et al. (2010) studied the production of collaborative suppliers and identified conditions under which cooperation and non-cooperation are profitable for buyers.

Position of the study in the literature Supplier sustainability continues to be an increasingly popular domain of research. However, not all facets of the profitability of manufacturers are thoroughly investigated. Current supplier sustainability literature focuses primarily on methods and approaches for implementing sustainability within the organizations and their supply chain. Analysis of the data from MNO’s supplier sustainability practices will prove the importance of variables or build predictive models. Thus, the widely used method is regression models and random forests. A technique that may be used is the classification models for the categorization of suppliers. Among the literature reviewed, clustering models primarily used for the selection and classification of suppliers. Clustering models can be used to character- ize supplier’s sustainability learning behavior that can be regarded as a gap in the literature.

MNOs are moving from tracking suppliers alone to strategies that go beyond tracking and therefore concentrate more on enhancements and developments. Suppliers are still not well- investigated in terms their sustainability learning behaviors. Implementing the learning curve model on the growth of suppliers can be seen in the literature (Gosling et al., 2018). However, pure analysis of sustainability learning behavior by using the learning curve methodology cannot be found. A direction for future research is therefore to create methods to analyze supplier

(28)

and buyer data in order to build a classification model that predicts sustainability learning behavior of suppliers. In addition to that, supplier development literature in the context of sustainability is not broad unlike supplier selection literature. Therefore, to create decision support models to optimize the effort of suppliers and buyers in order to increase the supply overall supply chain sustainability level while considering the trade-offs such as provided action plans, buyer collaboration strategies, costs and resulting sustainability gains. Especially, all the examined models in this paper use non-linear optimization models which can be re- duced to linear models as well (Mizgier et al., 2017; Talluri et al., 2010). It should be noted that it is difficult to grasp the full view of the supplier sustainability literature as they are diverse and there are different kinds of viewpoints. However, there are still literature gaps that are waiting to be analyzed and examined.

3 Identifying the Learning Behavior of Suppliers Using Their Char- acteristics

One of the research questions is about finding the relationship between supplier characteristics and learning behavior of suppliers. That’s why in this section the main focus will be on RQ1:

“What approach can be used to identify learning behaviors of suppliers based on their characteristics?“.

3.1 Theoretical Model

This section is for presenting the theoretical model which can be applied to every company which conducts supplier sustainability assessment. The conceptual framework, methodology and model will be discussed for the theoretical model.

3.1.1 Conceptual Framework

Determining the conceptual framework is necessary to construct the model itself for analyz- ing data and determining the parameters for the model. Once suppliers get into the program, their sustainability levels are revealed. By examining the change in sustainability levels of suppliers, the learning behavior of suppliers can be determined. Assumptions of the conceptual framework are: (1) Sustainability levels of suppliers are determined periodically. (2) Sustain- ability levels are defined in the form of a ‘score’ which depends on several criteria. In most of the industries, these assumptions can be considered as highly valid. Companies such as Canon, BMW, Philips and Apple which are members of Responsible Business Alliance(RBA)¹ conducts supplier sustainability assessments and constitute sustainability scores as an output of sustainability assessments. Supplier audits and/or supplier self-assessment questionnaires can be used for assessments. After having previous assumptions related to sustainability assessment procedures for industries, data related to sustainability scores of suppliers can be assumed to be stored by the buyer firm for each periodical assessment. In order to understand the learning behavior, one must examine the change in sustainability score over time. Change

1http://www.responsiblebusiness.org/