Mapping the market for remanufacturing: An application of “Big Data” analytics

https://doi.org/10.1016/j.ijpe.2020.107807Get rights and content

Abstract

Remanufacturing is one of the most examined topics in the closed-loop supply chain (CLSC) literature. However, we still have limited knowledge on the characteristics of the market for remanufactured products. This study addresses this gap by using a big data analytics framework. We employ off-the-shelf, pre-trained vectors created with the Global Vectors for Word Representation (GloVe) word embedding method from a data set crawled from the Internet. The Louvain method subsequently provides us with clusters based on remanufacturing and related terms, without requiring human interactions. Our findings provide the following main insights. First, remanufacturing and related terms are associated with specific industries and products, among which printing equipment, automobiles and car parts, treadmills, consumer electronics, and household appliances. Among the terms capturing remanufacturing activity, remanufactured, reconditioned, and rebuilt are strongly associated with business-to-business and slow clockspeed products, while refurbished is mostly associated with business-to-consumer and fast clockspeed products. Second, original equipment manufacturers (OEMs) are much more salient than independent remanufacturers, and Japanese OEMs are especially well represented as players in the market for remanufacturing. Third, environmental concerns only appear weakly in the discourse surrounding product recovery, while consumers do seem to place emphasis on quality and price. In a final part of the study, we contrast the CLSC academic literature with the clusters obtained through our big data analysis, thereby identifying industries, products, and brands that are understudied. We also outline the practical implications of our work for managers involved in setting up a remanufacturing strategy, as well as regulators.

Introduction

Remanufacturing recovers value from used products by replacing components or reprocessing used parts to bring the product to a like-new condition (Atasu et al., 2008). Within the literature on closed-loop supply chains (CLSC), a large number of studies examine the operational, tactical, and strategic decisions associated with remanufacturing (Galbreth and Blackburn, 2010, Goodall et al., 2014, Jeihoonian et al., 2017, Samarghandi, 2017, Abbey and Guide Jr., 2018, Pazoki and Samarghandi, 2020). While these studies take a corporate (supply) perspective, a growing stream of papers examines remanufacturing from a consumer (demand) perspective, by analysing the drivers of the willingness to pay (WTP) or the size of the market for remanufactured products (Guide Jr. and Li, 2010, Subramanian and Subramanyam, 2012, Quariguasi Frota Neto et al., 2016, Jakowczyk et al., 2017).

Despite the substantial academic coverage of remanufacturing in the CLSC literature, we still know relatively little about the basic characteristics of this multi-billion dollar activity. Back in 1996, Lund called the remanufacturing industry a hidden giant (Lund, 1996). Reinforcing this point, [50] argued that remanufacturing is in some respects an invisible activity, since it is not recognized as an industry by standard industry classification systems. More recently, the CEO of the Ellen MacArthur Foundation argued during his speech at the 2019 Annual World Remanufacturing Conference: “Nobody knows what reman is” (Source: https://bit.ly/34rlwo6).

In this paper, we aim to address the substantial gap in knowledge regarding the main characteristics of the market for remanufacturing products. We study three research questions: First, what are the key industries and products represented in the market for remanufactured products? Second, who are the most prominent players, i.e. firms and brands, represented in this market? And third, what are the most salient attributes associated with remanufactured products?

One option to examine the above questions would be to collate information from reports, newspaper articles, and academic papers, among others, and summarize these. However, this approach would be extremely time-consuming and potentially subjective, since the researcher would have to make an a priori selection of the material to review. We adopt an alternative framework based on big data analytics, allowing us to obtain a computationally inexpensive overview of the market for remanufactured products without relying on human interactions in the data collection.

More particularly, we use a novel approach that combines Natural Language Processing (NLP) and community discovery techniques. NLP, a powerful tool that attempts to make sense of human communication, has been successfully deployed in fields as diverse as bio-diversity science (Thessen et al., 2012), journalism (Cohen et al., 2011), law enforcement (Voigta et al., 2017), medicine (Hripcsak et al., 1995), and psychotherapy (Althoff et al., 2016). Community discovery, broadly speaking a mechanism for identifying individuals who exhibit similar preferences or speak the same language, in turn, has been applied in criminology, telecommunications, and sociology, among other disciplines (Ferrara et al., 2012, Ferrara et al., 2014).

To our knowledge, we are the first to adopt this approach for mapping the characteristics of the market for remanufactured products. In a first step of our analysis, we obtain off-the-shelf, pre-trained vectors from spaCy, an open-source software library for advanced natural language processing (Source: https://bit.ly/2vorTeZ). The vectors are created with Global Vectors for Word Representation (GloVe), one of the most well-known word embedding methods (Pennington et al., 2014), and are based on a data set obtained from Common Crawl, a nonprofit organization dedicated to providing a copy of the Internet (Source: https://bit.ly/2xJegYJ). In a second step, we apply the state-of-the-art Louvain method for community detection (Blondel et al., 2008), thus obtaining a series of clusters related to remanufacturing, without requiring further human input.

An analysis of these clusters provides us with the following insights regarding our three main research questions. First, with regards to the most popular industries and products linked with remanufacturing and related terms, we find evidence of strong associations with printing equipment, automobiles and car parts, treadmills, consumer electronics, and household appliances. Interestingly, among the terms capturing remanufacturing activity, remanufactured, reconditioned, and rebuilt are strongly associated with business-to-business (B2B), slow clockspeed products, while refurbished is closely linked with business-to-consumer (B2C), fast clockspeed products. The latter pattern suggests that refurbishing should not be considered a synonym for remanufacturing. Second, with regards to the key players in the market for remanufacturing, we find that original equipment manufacturers (OEMs) are much more salient than independent remanufacturers. Japanese OEMs are especially well represented, while Chinese OEMs have no salience. In a robustness test, we find that those OEMs identified by our cluster analysis as having high semantic similarity with remanufacturing tend to be effectively engaged in product remanufacturing. Third, in terms of the attributes most commonly associated with remanufacturing, the clusters suggest that environmental considerations are second to quality and price in the eyes of consumers.

Although our study is exploratory in nature and therefore does not start from a theoretical framework, we refer to relevant theoretical concepts of isomorphism (DiMaggio and Powell, 2000), clockspeed (Fine, 2000), and resource-constrained innovation (Weiss et al., 2011) when interpreting our findings. We also position our results relative to those of related other empirical studies on remanufacturing.

Our paper has two main implications for academic research. First, our findings can serve as a point of departure for the design of future empirical academic studies on remanufacturing. More particularly, our cluster analysis may help researchers identify industries and products that are currently most closely associated with CLSC and thus warrant further research, as well as some of the industries and products that have been overlooked by previous research. This could result in more cross-sectional variation and therefore increased power in empirical tests, by enhancing the diversity of industries and products under consideration. Our findings could also inform researchers on the potential main players in each of the remanufacturing markets, which could be relevant to those interested in intra-firm case studies. Moreover, our cluster analysis may help researchers identify a list of suitable keywords associated with remanufacturing, which could serve as a useful sample collection tool for empirical studies on characteristics of remanufactured products (see, e.g., [75]). Second, our findings on the differences between remanufacturing and refurbishing, as well as on the attributes commonly associated with remanufactured products, contribute to ongoing academic debates about these so far unresolved issues (Thierry et al., 1995, Abbey et al., 2015a, Abbey et al., 2015b).

Practitioners and regulators, in turn, also stand to benefit from a deeper knowledge into the characteristics of the market for remanufacturing. To give an example, OEMs not yet engaged in product remanufacturing may benefit from being able to gauge the popularity of remanufactured product types, a proxy for cannibalization potential, which is an issue of increasing importance (Guide Jr. and Li, 2010). Firms faced with the prospect of developing a remanufacturing marketing strategy might also find it useful to know that green credentials seem to be less important than price and quality cues for these products. Finally, regulators might find our work valuable for remanufacturing policy setting purposes, as it provides them with a birds-eye overview of the main characteristics of the market. More generally, our work showcases the ability of big data analytics to generate crucial insights into CLSC market dynamics in a fast, efficient, and objective way.

The remainder of this paper is structured as follows. In the next section, we briefly outline the main strands of literature relevant to our work. Section 3 outlines the big data analytics methodology used in our study. Section 4 describes the clusters generated by the framework. Section 5 documents the main insights related to our three research questions that can be derived from these clusters. Section 6 discusses the implications of our work for academics and practitioners. Section 7 discusses limitations and avenues for future research.

Section snippets

Literature review

Our work sits at the intersection of three relevant streams of literature. Topic-wise, it is positioned within the literature on the market for remanufactured products. More broadly, it is also related to studies tapping the Internet for data, as well as studies using the NLP technique for data analysis. We now briefly describe the most relevant papers in each of these three streams and outline the incremental contributions of our study.

Methodology

This section outlines the methodology used for our analysis. In a first subsection, we discuss the concepts of semantic similarity and word embedding in general terms. In a second subsection, we briefly describe the GloVe word embedding method and resulting word vectors used in our paper. In a third subsection, we discuss how we use the Louvain method for community detection to obtain clusters related to remanufacturing based on the word vectors obtained through GloVe.

Summary of clusters for remanufactured and related terms

In this section, we document the clusters obtained for the kernel term remanufactured, which is the key focus of our analysis, as well as for the three related kernel terms refurbished, reconditioned, and rebuilt.

In the graphs initialized with remanufactured, and depicted in Fig. 1(a), eight clusters were formed. Each of the clusters generates a meaningful interpretation, save for the small cluster entitled Unclear interpretation. From the kernel term refurbished, presented in Fig. 1(b), half

Mapping the market for remanufacturing: main insights from the cluster analysis

In this section, we discuss the main conclusions that can be drawn from the clusters presented in the previous section. We have split the section in subsections capturing our three research questions.

Implications of our findings for academics and practitioners

Our research could be useful for academic researchers for two main reasons. First, our findings can serve as a foundation for future empirical work on remanufacturing. Second, our findings can contribute to a number of ongoing academic discussions in the area of CLSC. We now motivate these two implications in some more detail.

Our findings can serve as a starting point for future empirical studies by highlighting areas of remanufacturing activity that are currently under-researched. To make it

Conclusions, limitations, and further research

This paper describes the use of pre-trained word vectors trained on an unlabelled general corpus obtained from the Internet (rather than a collection of texts derived on the topic of remanufacturing), and an application of a non-supervised algorithm to extract important information about the market for remanufactured products.

The procedure outlined here has brought into light, the industries and products that are perceived to be most closely associated with remanufacturing activity. It has also

Acknowledgement

We are grateful to Marc Reimann and seminar participants at Cass Business School and University of Graz for their useful comments.

References (89)

  • KerrW. et al.

    Eco-efficiency gains from remanufacturing: A case study of photocopier remanufacturing at Fuji Xerox Australia

    J. Cleaner Prod.

    (2001)
  • LiN. et al.

    Using text mining and sentiment analysis for online forums hotspot detection and forecast

    Decis. Support Syst.

    (2010)
  • MaiF. et al.

    Deep learning models for bankruptcy prediction using textual disclosures

    European J. Oper. Res.

    (2019)
  • MuggeR. et al.

    How to sell refurbished smartphones? An investigation of different customer groups and appropriate incentives

    J. Cleaner Prod.

    (2017)
  • PazokiM. et al.

    Take-back regulation: Remanufacturing or eco-design?

    Int. J. Prod. Econ.

    (2020)
  • ReimannM. et al.

    Managing a closed-loop supply chain with process innovation for remanufacturing

    European J. Oper. Res.

    (2019)
  • SamarghandiH.

    Studying the impact of merged and divided storage policies on the profitability of a remanufacturing system with deteriorating revenues

    Int. J. Prod. Econ.

    (2017)
  • SinghA. et al.

    Social media data analytics to improve supply chain management in food industries

    Transp. Res. E

    (2018)
  • van WeeldenE. et al.

    Paving the way towards circular consumption: Exploring consumer acceptance of refurbished mobile phones in the dutch market

    J. Cleaner Prod.

    (2016)
  • WangY. et al.

    Consumer product knowledge and intention to purchase remanufactured products

    Int. J. Prod. Econ.

    (2016)
  • XuX. et al.

    The influence of e-services on customer online purchasing behavior toward remanufactured products

    Int. J. Prod. Econ.

    (2017)
  • YearworthM. et al.

    Spontaneous emergence of community OR: Self-initiating, self-organising problem structuring mediated by social media

    European J. Oper. Res.

    (2018)
  • ZhangW. et al.

    Weakness finder: Find product weakness from Chinese reviews by using aspects based sentiment analysis

    Expert Syst. Appl.

    (2012)
  • AbbeyJ.D. et al.

    A typology of remanufacturing in closed-loop supply chains

    Int. J. Prod. Res.

    (2018)
  • AbbeyJ.D. et al.

    The role of perceived quality risk in pricing remanufactured products

    Prod. Oper. Manage.

    (2017)
  • AbbeyJ.D. et al.

    Consumer markets for remanufactured and refurbished products

    Calif. Manage. Rev.

    (2015)
  • AbrahamsA.S. et al.

    An integrated text analytic framework for product defect discovery

    Prod. Oper. Manage.

    (2015)
  • AgrawalV.V. et al.

    Remanufacturing, third-party competition, and consumers’ perceived value of new products

    Manage. Sci.

    (2015)
  • AlthoffT. et al.

    Large-scale analysis of counseling conversations: An application of natural language processing to mental health

    Trans. Assoc. Comput. Linguist.

    (2016)
  • AtasuA. et al.

    So what if remanufacturing cannibalizes my new product sales?

    Calif. Manage. Rev.

    (2010)
  • AtasuA. et al.

    Remanufacturing as a marketing strategy

    Manage. Sci.

    (2008)
  • BlondelV.D. et al.

    Fast unfolding of communities in large networks

    J. Stat. Mech. Theory Exp.

    (2008)
  • BodnarukA. et al.

    Using 10-k text to gauge financial constraints

    J. Financ. Quant. Anal.

    (2015)
  • BojanowskiP. et al.

    Enriching word vectors with subword information

    (2016)
  • CabralL. et al.

    The dynamics of seller reputation: Evidence from eBay

    J. Ind. Econ.

    (2010)
  • ClayK. et al.

    Prices and price dispersion on the web: Evidence from the online book industry

    J. Ind. Econ.

    (2001)
  • CohenS. et al.

    Computational journalism: A call to arms to database researchers

  • De MeoP. et al.

    Generalized louvain method for community detection in large networks

  • DeephouseD.L.

    Does isomorphism legitimate?

    Acad. Manage. J.

    (1996)
  • DiMaggioP.J. et al.

    The iron cage revisited: Institutional isomorphism and collective rationality in organizational fields

    Econ. Meets Sociol. Strateg. Manage.

    (2000)
  • EatonD.H.

    Valuing information: Evidence from guitar auctions on eBay

    J. Appl. Econ. Policy

    (2005)
  • FengY. et al.

    The state of the art in semantic relatedness: A framework for comparison

    Knowl. Eng. Rev.

    (2017)
  • FerraraE. et al.

    Visualizing criminal networks reconstructed from mobile phone records

    (2014)
  • FerraraE. et al.

    The role of strong and weak ties in facebook: A community structure perspective

    (2012)
  • Cited by (20)

    • Blockchain adoption and channel selection strategies in a competitive remanufacturing supply chain

      2023, Computers and Industrial Engineering
      Citation Excerpt :

      Annual sale of the global remanufacturing industry exceeded 150 billion U.S. dollars and the United States alone occupied 100 billion U.S. dollars (Liu et al., 2017). In practice, the remanufacturing market such as printing equipment, automobiles and car parts, consumer electronics and household appliances has been widely and successfully applied by third-party remanufacturers (TPRs) (Neto and Dutordoir, 2020; Yang et al., 2021). Foxconn is authorized by Apple to recycle and remanufacture End-of-life (EOL) iPhones (Zou et al., 2016).

    • A sustainable step forward: Understanding factors affecting customers’ behaviour to purchase remanufactured products

      2023, Journal of Retailing and Consumer Services
      Citation Excerpt :

      Customers purchase intentions and purchase behaviours are strongly linked, according to research using the theory of planned behaviour (TPB) (Ajzen, 2011). Although prior studies have confirmed that the behavioural intentions are a significant variable for actual behaviours, the actual behaviours often varies because customers do not always follow through on the intentions, particularly in unique contexts when presented with options or given restrictions (Neto and Dutordoir, 2020). It is possible that remanufactured products may provide one of these contexts in which intentions are inconsistent with actual behaviours (Ho et al., 2018).

    View all citing articles on Scopus
    View full text