Exploring the processing of product returns from a complex adaptive system perspective

Jennifer A. Espinosa (Department of Marketing and Business Information Systems, Rowan University, Glassboro, New Jersey, USA)
Donna Davis (Department of Marketing, University of South Florida, Tampa, Florida, USA)
James Stock (Department of Marketing, University of South Florida, Tampa, Florida, USA)
Lisa Monahan (Department of Marketing, Meredith College, Raleigh, North Carolina, USA)

The International Journal of Logistics Management

ISSN: 0957-4093

Publication date: 12 August 2019



The purpose of this paper is to explore the processing of product returns at five case companies using a complex adaptive systems (CAS) logic to identify agent interactions, organization, schema, learning and the emergence of adaptations in the reverse supply chain.


Using a multiple-case study design, this research applies abductive reasoning to examine data from in-depth, semi-structured interviews and direct researcher observations collected during site visits at case companies.


Costly or high-risk returns may require agents to specialize the depth of their mental schema. Processing agents need freedom to interact, self-organize and learn from other agents to generate emergent ideas and adapt.

Practical implications

Limiting the depth of individual agent schema allows managers to better allocate labor to processing product returns during peak volume. To boost adaptability, managers need to craft a dynamic environment that encourages agents with diverse schema to interact, anticipate, and self-organize to brainstorm new ideas. Managers need to resist the urge to “control” the dynamic environment that ensues.


This paper builds on existing research that studies the key decision points in the analysis of product returns by exploring how processing-agent behaviors can create adaptability in the reverse supply chain. Additionally, this research follows in the tradition of Choi et al. (2001) and Surana et al. (2005) and proposes the application of CAS to a specific part of the supply chain – the processing of product returns.



Espinosa, J.A., Davis, D., Stock, J. and Monahan, L. (2019), "Exploring the processing of product returns from a complex adaptive system perspective", The International Journal of Logistics Management, Vol. 30 No. 3, pp. 699-722. https://doi.org/10.1108/IJLM-08-2018-0216



Emerald Publishing Limited

Copyright © 2019, Emerald Publishing Limited

1. Introduction

Customers today expect more than ever before when shopping (e.g. instantaneous access to products, seamless shopping experiences), and forward supply chains have begun to respond to these demands by developing fast moving retail supply chains and omnichannels (Jones, 2018; Petro, 2018). Increasingly, customers are assimilating desires for accessibility and convenience into their expectations for returning products. So far, few supply chains have made modifications to their product return strategies to offset the costs associated with these evolving expectations (Dennis, 2018; Lindsey, 2016). However, the crucial role that product returns play in forging productive and sustainable consumer relationships is undeniable (Hjort et al., 2013). While every consumer/firm transaction provides an opportunity to create consumer value (Kumar et al., 2010), returns are unique in that they begin with a negative experience that precipitates the return. When a firm acts to turn a negative experience into a positive one, consumers are often motivated to purchase again from the same firm and spread the word to others about the turnaround experience (Gesenhues, 2017).

Product returns require companies to perform a balancing act between setting policies that are simple enough to encourage repeat shopping, yet strict enough to prevent return abuse, while also maintaining an effective process to dispose of returned products (Goldman, 2016; Jack et al., 2010). Traditional product return strategies have focused on optimizing return policies to minimize returns as much as possible and/or identifying characteristics of “serial” returners (Daunt and Harris, 2012; Davis et al., 1998; Janakiraman and Ordonez, 2012). However, this optimization approach significantly downplays the importance of human behavior in product returns and largely ignores insights generated from the processing of returns. Returning a product involves the interaction of at least two people (e.g. the customer and the employee accepting the return) but typically more, all of whom have different motivations, power, and information-processing capabilities. Companies have traditionally assumed customers behave rationally when returning products (i.e. return a product because it is defective). However, shopping trends indicate customers are increasingly prioritizing their needs at the expense of the seller’s (e.g. buying multiple sizes of the same product with the intention of returning the sizes that do not fit; Welson-Rossman, 2018). Furthermore, the employees who evaluate and process returned products are not immune to decision-making errors. Thus, in comparison to return behaviors by customers, behaviors of employees processing product returns have received little attention in the literature, despite their substantial impact on a firm’s profitability.

Embracing and leveraging employee behaviors during the returns process represents one relatively untapped, alternative approach to the optimization of customer behaviors, with the potential to reveal product quality issues and suggest solutions to those issues in real time (Storer et al., 2014). This research applies complex adaptive systems (CAS) theory (Holland, 1995) to explore the different parts of the system that interact during the processing of product returns. More specifically, we will closely examine how employee behaviors can benefit or harm the system. Unlike an optimization strategy that assumes customer and employee behaviors do no harm, CAS theory studies both the benefits (e.g. creativity) and drawbacks (e.g. conflict) of human behavior on the system (Nilsson and Gammelgaard, 2012). One important research question that arises is:


How can employee behaviors during the processing of product returns increase the reverse supply chain’s adaptability?

The objective of this paper is to explore the product returns process from a CAS perspective. Specifically, this research examines how interactions, learning, and adaptations occur during the processing of product returns. To achieve this purpose, we examine the product returns process at five different case companies to develop an in-depth understanding of the roles employee behaviors play. In the next section, we review the relevant literature, and offer a brief description of our abductive case study methodology. Next, detailed descriptions of the product return processes observed at each of the five case companies are provided, followed by a discussion of CAS-grounded patterns that emerged across the cases. The paper closes with a discussion of the implications of the current research and directions for future research.

2. Literature review

The conceptual research framework (see Figure 1) for this research is based on the intersection of the product returns process and CAS theory, and is grounded in research insights from the reverse logistics, supply chain management, natural and social sciences literatures.

2.1 Processing product returns

Timely processing is a strategic weapon companies can utilize to reduce the financial impact of returned products and build customer loyalty despite the failure of a product (Kocabasoglu et al., 2007). Monitoring product return rates has the potential to identify product issues that can be corrected immediately to prevent additional returns of the same product, however, few companies dedicate the necessary resources to gather and act on this type of information (Goldman, 2016; Ho et al., 2012; Lindsey, 2016).

Research by Stock et al. (2006) and Rogers et al. (2002) describes the typical product returns process as a series of steps products pass through sequentially. Stock et al. (2006) outline five key steps during which a returned product arrives at the processing center, is organized (by type of return, date received, etc.), inspected, and reconciled against the returns authorization before being dispositioned to recoup value. Rogers et al. (2002) take a broader view of the overall returns process, starting with the customer’s initial request to return a product and ending with monitoring return rates to identify problematic products. In practice, the order of the steps varies. For example, original equipment manufacturers of home domestic products (e.g. vacuum cleaners) that use third-party logistics (3PL) providers to facilitate product returns analyze product returns earlier in the process because they must determine if the retailer or 3PL will disposition the product (Bernon et al., 2013). Following an established returns process allows companies to reduce excess inventory, avoid storage costs and minimize the risk of product obsolescence (Autry et al., 2001; Blanchard, 2010; Stock and Mulki, 2009).

Analyzing product returns is the most important step in the returns process because employee decision making during this step will directly affect the disposition option chosen, and subsequently the value recouped from a returned product (Stock et al., 2006). Despite the significant financial implications of analyzing product returns, surprisingly few researchers have explored how employees make these decisions. Hazen et al. (2012) study employee disposition decisions and outline seven components of a disposition decision, such as scanning the external environment to determine if a market exists for an imperfect product. Tan and Kumar (2006) model key decision points (e.g. make or buy repair parts) that shape the profitability of returns. Beyond these few studies that identify important decision points in dispositioning returned products, very little is known about how employee behaviors impact the efficiency of the product returns process.

In addition to studying the analysis step in the product returns process, prior research also examines system design options to manage the flow of returned items (e.g. closed loop supply chain; Blackburn et al., 2004; Turrisi et al., 2013), including optimizing the put away of returns with picking orders for outbound shipments (Schrotenboer et al., 2017), and different disposition options (Hazen et al., 2012; Stock and Mulki, 2009). Ultimately, issuing partial refunds to customers to offset the costs of coordinating the processing of product returns may be the ideal return process outcome from the company’s point of view (Altug and Aydinliyim, 2016; Su, 2009).

2.2 Optimizing customer return behaviors

A second strategic weapon companies can rely on to protect their bottom line is their return policy. During the creation of return policies, companies often include specific restrictions (e.g. a 30-day return window) that stipulate the conditions under which a product can be returned (Piron and Young, 2001). The restrictions included in a return policy tend to increase the level of effort a customer must exert to return a product (Janakiraman and Ordonez, 2012), and the return policy itself acts as a pre-purchase signal of the quality of the retailer (Bonifield et al., 2010). Many companies view lenient return policies as a competitive way to signal their goodwill to customers and increase spending post- return, particularly in online transactions (Bower and Maxham, 2012; Janakiraman et al., 2016a; Oghazi et al., 2018). In fact, online firms such as Amazon have made it so easy to return, that consumer expectations have evolved to expect a no-questions-asked return policy (Safdar, 2018).

A sizable stream of research in the product returns literature studies how companies can reduce returns by optimizing the leniency of their policies (e.g. Petersen and Kumar, 2009; Powers and Jack, 2013). However, newer research is increasingly finding that too much leniency may damage the retailer–customer relationship. For example, Hjort and Lantz (2016) warn retailers to use caution when crafting lenient return policies, finding free returns tend to attract less profitable customers. Janakiraman et al. (2016b) instead suggest companies be selectively lenient based upon the reason for the return or the established relationship quality of the customer. However, optimized return policies often overlook the employee interactions involved in a product return, and ultimately limit the value that can be created during the returns process.

2.3 Complex adaptive systems theory

A CAS is “an interconnected network of multiple entities that exhibit adaptive action in response to changes in both the environment and the system of entities itself” (Pathak et al., 2007, p. 550). CAS theory studies how agents (defined as entities in a system that can be changed; Holland, 2014) relate and interact within a system to create super-additive outcomes (Nilsson and Gammelgaard, 2012). Agents in CAS can be humans, processes or important items that possess unique abilities to influence other agents in interactions and subsequently, the system as a whole (Holland, 1992; Nilsson and Gammelgaard, 2012).

One unique ability human agents possess in CAS are agent schema, or internal mental networks of information learned through prior experience, that agents rely on when processing information and making decisions (Choi et al., 2001; Pathak et al., 2009; Surana et al., 2005). Agents organize their schema by developing nodes (defined as a central mental location that stores information about similar topics) and linkages (defined as mental associations that connect related nodes; Hawkins and Mothersbaugh, 2010). With experience, agents develop heuristic information-processing mechanisms for routine decisions, so that in future interactions, the activation of one node will automatically activate other nodes heuristically associated with it (Hawkins and Mothersbaugh, 2010; Vohs et al., 2008).

Agents in CAS frequently learn from their own experiences and through interactions with other agents. When agents encounter a novel situation, they may perform first-order learning where they attempt to craft a solution by comparing familiar aspects of the situation to information stored in their mental schema (Dooley, 1997). Based upon the comparisons and their intuition, agents experiment with solutions to solve the problem (Allen, 2000). CAS also encourages second-order agent learning (defined as an agent’s ability to learn how to learn in new ways; Schon, 1975) through interactions with other agents. Effective strategies generated by one agent through first-order learning can be shared directly with other agents in interactions, which become adaptive interactions when those other agents learn how to respond to novel problems through the experience of the first agent (Holland, 2014).

As agents make decisions after interactions with other agents, feedback loops form (Stacey et al., 2000; Wysick et al., 2008). Even seemingly small decisions (such as “is the information I have obtained from this interaction useful?”) can quickly be magnified and influence other agents in the system (Holland, 2014). For example, if an agent decides the information learned in a recent interaction is not useful, s/he may refrain from sharing it with other agents and/or avoid interacting with specific agents in the future (Nilsson, 2006; Stacey et al., 2000). This chaotic, non-linear behavior, where a small decision by an agent has a disproportionately larger effect on CAS, coupled with the frequent occurrence of unusual events (called fat-tailed behavior) keep the CAS in a state of perpetual novelty and require agents to constantly adapt (Choi et al., 2001; Holland, 2014; Surana et al., 2005).

CAS rarely reach a state of equilibrium, which forces agents to make decisions and adapt based on imperfect information about the state of other agents and the consequences of their decisions (Holland, 2014; Nilsson and Gammelgaard, 2012; Simon, 1991). Agents compensate for this dynamic state through self-organization, whereby they rely upon information provided by fellow agents rather than a formalized system leader to anticipate necessary changes (Choi et al., 2001; Potts, 1984). Anticipatory agent actions add additional non-linearity to CAS, but yield emergent adaptations, where the whole adaptation is more than the sum of the ideas generated by individual agents (Holland, 2014). The diverse schema, shared experiences and skills of agents within an uncertain, dynamic CAS environment are the essential ingredients for the emergence of adaptations to occur (Allen, 2000; Choi et al., 2001).

In the supply chain management literature, Choi et al. (2001) were among the first to apply CAS theory to supply networks and identify specific behaviors complex adaptive supply networks (CASN) exhibit. Choi et al. proposed that firms with similar levels of shared schema within a supply network will perform better than firms with different levels of shared schema. Surana et al. (2005) also argue for supply chains to be treated as CAS, and utilize CAS concepts in conjunction with network dynamics to improve the modeling of collaboration relationships in supply chains. Nilsson and Darley (2006) offer insights to improve the modeling of interrelationships between manufacturing and logistics operations using agent-based simulations to increase the knowledge and intuition of decision makers. Wysick et al. (2008) study supply networks through the lens of CAS theory, and explore the vulnerability of these networks to fat-tailed behavior, like the bullwhip effect, and recommend the implementation of neural networks to help manage extreme events. Li et al. (2010) examine the internal and external forces that drive the development of CASN, and offer a self-organizing evolutionary model to explain how CASN form.

In the reverse logistics literature, the application of CAS theory has revealed how to evaluate different value recovery options for original equipment manufacturers (Lehr et al., 2013) and how to best manage green supply chains (Sarkis et al., 2010). CAS theory has many useful applications in reverse logistics beyond these initial studies, particularly as a means of understanding agent behaviors during the processing of product returns.

3. Methodology

3.1 Research approach

The purpose of the current research study is to explore the product returns process through a CAS theoretical lens. Agents within the product returns system serve as the unit of analysis for the current research. We collect the knowledge and opinions of reverse logistics managers across different reverse supply chains to understand the various types of agent behaviors in the product returns system (Yin, 2018). All informants generously granted us access to their managers who oversee product returns, who by the nature of their positions, were able to provide us with the in-depth information we required. Understandably, most informants could not grant us full access to their product return processor employees due to the time-sensitivity of their business.

A case study methodology was selected because the processing of product returns is a function embedded within a company’s supply chain, and case study research is well-suited for studying a phenomenon “in depth and within its real-world context, especially when the boundaries between phenomenon and context may not be clearly evident” as the boundaries between forward and reverse logistics often are not ( Yin, 2018, p. 15). Specifically, an abductive case study methodology allowed us to identify generalizable and specific CAS properties among our five case companies (Kovacs and Spens, 2005). Following the established abductive research procedures discussed by Kovacs and Spens (2005), our research occurred in three stages. In the first stage of the research, we were contacted by a multi-billion dollar US-based technology products company and asked to evaluate their product returns processes. To do this, we recorded detailed observations of their processes while touring their distribution center and conducted in-depth interviews with employees including the logistics director and senior reverse logistics manager.

During the second stage, we reviewed the academic literature to identify theoretical insights that aligned with the product returns process at the technology products company. We found that existing research describing the steps in the product returns process did not fully match those of the technology products company, and we therefore used CAS theory as a theoretical framework to better understand the company’s processes. In addition, we also constructed a case study protocol (see Table I) that included broad discussion points related to components of the reverse supply chain of interest (e.g. software used to process returns), but carefully avoided making any direct or indirect references to CAS theory, so we did not “lead” respondents to discuss complexity (Edmondson and McManus, 2007; Yin, 2018). Instead, interview protocol questions focused on understanding the full breadth and depth of informants’ experiences with product returns. Prior to each site visit, these questions were reviewed by two academics to identify ambiguous wording or topical omissions.

In the third stage of the research, interviews were conducted with four additional, unique companies to determine how well our theoretical propositions resonated with firms selling different types of products (Kovacs and Spens, 2005; Yin, 2018). We collected and analyzed our data simultaneously, going back and forth between our empirical data and theory, to identify general and specific CAS properties in our data (Dubois and Gadde, 2002; Kovacs and Spens, 2005; Strauss and Corbin, 1998). This multiple-case replication procedure allowed us to draw cross-case insights to deepen and refine our understanding of the product returns process (Yin, 2018).

3.2 Overview of research setting

The technology products company, which served as the starting point for this multiple-case study, is an American-based multinational distributor of IT products and services with over $25bn in sales annually. This firm processes approximately 2m returned products each year, and has the opportunity to act as a third-party returns processor for several potential clients. In addition, four other national or multinational companies that sell different types of products were included in this multiple-case study to allow for the identification of generalizable and specific properties (Kovacs and Spens, 2005). The four other companies were purposively selected for inclusion in our research based upon the type of product sold and their position in the supply chain to maximize variation in our informants. Table II summarizes the case companies included in the multiple-case study and contains descriptions of their products, primary selling location, sales and return volume.

3.3 Data collection

Data collection started in 2014 and ended in 2015. We conducted six semi-structured interviews with employees from five different types of companies. All five companies were visited in person. Site visits began with a tour of the distribution facilities, during which direct observations were hand-recorded by a member of the research team in a small notebook. Following the tours, we conducted the in-depth interviews. The audio of all interviews was recorded on a tape recorder. Table III details the profile of each informant, including their pseudonyms, titles and background information.

3.4 Data analysis

All interviews (lasting 20–120 min) were digitally recorded and professionally transcribed. Immediately following the visit, all direct observations from site visits were typed out based on the research team’s handwritten notes. To conduct our data analysis, we adopted a theoretical orientation and utilized insights from CAS to organize our analysis (Yin, 2018). All interview and direct observation documents from each case were coded and analyzed in the NVivo 10 software (QSR International, 2012). The analysis was conducted after each site visit in order to deepen the research team’s understanding of the data and identify any interview protocol refinements that should be made based upon emerging themes or issues (Chakkol et al., 2014). Results of the analysis were then shared with each case organization, and each informant verified the accuracy of our records and interpretations to ensure the validity of our data (Yin, 2018). The next section describes the within-case patterns observed at each company, followed by a discussion of the cross-case patterns observed across all of the case companies.

4. Within-case findings

This section details the results for each case without discussing generalizable patterns across the cases, which will be done in Section 5. The case results proceed as follows:

  • description of agents and resources;

  • description of agent interactions and organization;

  • description of agent schema and learning; and

  • description of emergent adaptations.

4.1 Case A – processing technology product returns

Processing product returns at the technology products company involves several agents of varying influence, including the customer, customer service representative, front-line return processors, exception specialists and the senior manager of reverse logistics. Of these human agents, the front-line and exception processors make the key decisions, and therefore have the most influence, on the returns process. The primary resources include the returned products and information system.

To mitigate the risk of customer deviance, product returns are processed in one of two tracks- a “normal track” and an “exception track.” Direct researcher observations reveal that all returned products begin processing when they arrive at the distribution center and are sorted by date of arrival on pallets. Front-line processors begin to interact with the returned products by moving the pallets to their workstation and inspecting the product actually returned to determine if it matches the returns authorization. If the product and returns authorization match, the returned product enters the “normal track” and the front-line processor will make a disposition decision based upon the condition of the product.

Products that do not match the returns authorization enter the “exception track” and are sent to an exception specialist for extra processing. Exception specialists have additional training beyond front-line processors, and are particularly adept at handling problematic returns. Dave (see Table III) describes exception processing at the technology products company as more interactive, requiring back and forth communication between the exception specialist, customer service representative and the customer:

And that communication [about the product’s identity] is essentially a back-and-forth communication between us at our site – because we are hands on and we can visually see what we have – versus the decision makers at [customer service], who have to determine whether or not we issue the credit to the customer.

(Dave, Logistics Director, Technology products company)

Human agents at the technology products company interact frequently with the reverse supply chain’s information system to access information required to process each return, including customer transaction data. However, this information system operates on a unique programming language developed in-house (termed legacy system going forward), and does not communicate well with the newer enterprise resource planning (ERP) system used to manage forward logistics. The legacy system has a significant impact on processing product returns, requiring front-line processors to go through at least five to ten extra screens to process a single return and update the salable stock database.

As they process product returns, front-line and exception processors at the technology products company rely on specialized schema that house information about the different types of products sold, common issues with those products, and potential fixes for each product. Processors use their schema when evaluating a returned product’s identity and condition, and often assume a detective role to look for hidden clues (e.g. incorrect product box used to return an item) to determine the legitimacy of a return. Front-line processors handle a broad variety of returns that require different amounts and types of knowledge. Processors learn new information through resources available online – typically a vendor’s website for product descriptions or Google for pictures of the product.

Dave asks employees to share ideas they come up with from their day-to-day job responsibilities with the rest of the department. While Dave’s leadership style has generated some best practices to improve the usage of recyclable materials in outbound shipments, few ideas have been generated to improve the processing of product returns. Managers at the technology products company are acutely aware of their need to improve their legacy system, but struggle with generating ideas on how to adapt the information system:

As we are taking on new types of business that are a little bit outside of our norm, we’re finding that we are trying to have to manipulate this archaic system to a tremendous amount, and it is increasing a lot of steps – extra steps and things like that – that we’re not accustomed to doing with our normal business, that has made it very difficult to adapt to new business opportunities.

(Dave, Logistics Director, Technology products company)

4.2 Case B – processing apparel returns

Agents involved in the processing of returned products at the apparel company include the customer, customer service representative, processors, and the reverse logistics manager. The processors have the most influence. Resources include the product, information system and product offering catalogs.

Processing returns at the apparel company is a research-intensive, highly manual process which begins when returned clothing items arrive at the distribution center and enter the returns area:

So, typically in the returns process, if it’s coming from a retailer, they’ve got to reach out to customer service […] let us know what they are planning on returning […] whether it was a defective item, whether or not we made an agreement from an inventory standpoint that they can return stuff that they didn’t sell, or for other purposes. So, they reach out to customer service, customer service will generate a returns authorization and then that will allow them to start shipping the product back to us.

(John, Director of Distribution, Apparel company)

Processors check the returns authorization in the information system, but must visually identify each product returned because the apparel company screen prints and distributes for numerous retailers and does not put their own SKU information on each finished clothing item. Instead, processors work backwards from the description of the items on the returns authorization and what the items look like to determine the style of shirt and locate the correct SKU to reenter into inventory. For popular styles, this process takes only a few minutes. If the style is difficult to locate, processing one item could take up to a few hours. In a typical month, the apparel company has 40,000 to 50,000 active SKUs eligible to be returned, which includes SKUs for blank clothing items (i.e. no team or player printed on the clothing item) and customized blanks for 30+ teams and/or players across several sports leagues. Once the correct SKU is located, processors decide whether the clothing item can be re-sold, recycled or donated.

Processors at the apparel company develop highly specialized schema through experience researching the different apparel items sold in company catalogs or online. Therefore, during high volume periods, it is easy for a backlog of returns to pile up since labor cannot be easily reassigned from pickers and packers of outbound shipments:

So, I end up having to use my packing capacity, cross-training people there to help do returns. Returns associates tend to have specialized knowledge, so outbound associates cannot easily fill in or help with returns.

(John, Director of Distribution, Apparel company)

Processors must also be knowledgeable about sports seasons, player trades, team logo/color changes and time-sensitive game schedules when making a disposition decision about returned items.

John has visions of what an ideal information system should function like for processing apparel returns:

Well, we’d love to just be able to take a picture of the shirt. and have it automatically look up and tell you what style it was. If we were going to start from scratch. If Apple was designing the system. so, you could take your iTouch or your iPhone or iPad […] you’d take a picture and it would tell you exactly what it is.

(John, Director of Distribution, Apparel company)

However, John and his team face challenges associated with the technological shortcomings of the legacy system, and the internal processes required to adapt the system to meet business needs:

When you have old legacy systems, it becomes very difficult and sometimes painful to get changes approved. So, there’s two different ways we’ve created. One way is to come up with a creative way to work around the system to get what you want, or a creative way to change the system that doesn’t require extensive amounts of resources. [The second way], if we want a process change is we will go to the systems group and say, “currently the system does this, and we want it to do that.” They’ll say, “okay, based on what you want, your specifications, it will take us 700 programming hours.” We know we can’t afford that, so how do we modify that and maybe change it up a little bit, so we can get the hours down to an affordable amount, so there will be a payback on the process […] So, you have to prioritize what changes are the most important to get us the biggest bang for our buck, and how do you make sure that we’re not investing too much money into those legacy systems?

(John, Director of Distribution, Apparel company)

Many processors at the apparel company have worked there for years, so even changes that do ultimately get implemented still need to gather agent buy-in to be successful.

4.3 Case C – processing automotive part returns

Agents involved in the processing of automotive part returns include the company’s retail stores, processors, and the in-bound goods manager. Processors have the most influence on the processing of returned products. However, the retail stores have some power because they decide what returns to accept, and the processors often feel obligated to keep the stores happy. The primary resources include product returns and the information system.

Returns processing at the automotive parts company begins at the store level, and ideally each store will sort the products being returned into different categories (e.g. overstock, defective parts) on different pallets. However, in practice, the stores rarely sort the returned products and this adds additional processing steps at the distribution center:

If [the pallet] is really mixed, what happens is we need to have a person kind of separate it out. And then process everything that way. And then once everything is separated out, it goes into the avenues that it needs to go into […] There’s an authorization to return that is there, they will scan that. The system will cross-reference that authorization to return number with what type of SKU the store said they were going to send back […] If everything matches up, the system will then say, “okay, it’s received, put it in a container that they set up for put away.” And basically, our put away is zoned out. Every location in our building is zoned out. So, it will say, “you already have a container set up for zone 87” and if they receive something into inventory that is for zone 87, it will say, “put it in this container.”

(Mike, Out-bound operations manager, Automotive parts company).

The automotive parts company has been able to use a popular application-developed IBM DOS-based system to manage returns for 20+ years because they can easily find programmers versed in the universal programming language of DOS. In-house programmers are available to make necessary changes to the information system, which has resulted in a quick, efficient routine for processors to interact with:

Yeah, it is actually one of the fastest systems you could use because the overhead is so small on a DOS-based system […] You know, if you are a cost making facility like we are – we are basically a money sink, this building – that’s what it is. Speed is the key.

(Mike, Out-bound operations manager, Automotive parts company)

Processor interactions during the returns process are fairly limited by automation, and errors occur frequently because most returned parts have multiple identification numbers attached to them. Processor schema of the different types of parts is the primary defense against these types of errors. Processors learn new product information through experience, and also can join specialized groups (e.g. slotting group) to learn additional information (e.g. how products are slotted in the warehouse):

Every SKU has a couple identification numbers. For us in the building, the SKU is the master number. But, we also have a parameter ID number […] and then we have a part number […] Just the nature of having three numbers to identify one product, you can see we could have multiple numbers to scan and one number that identifies two different things […] our team members have to be smart enough, that when they are receiving back an engine- for some reason, it happens a lot with engines maybe because of how long a sales life of an engine lasts- you don’t want to give store credit for a hose. And the store will be like “oh, hey you shorted me like $3000” and if you look back, the team member scanned the SKU, but didn’t realize that there were two options. And that they were supposed to pick the engine part.

(Mike, Out-bound operations manager, Automotive parts company)

Overall, the automotive parts company has attempted to streamline the analysis of product returns as much as possible, with the goal of one day completely eliminating the need for a reclamation department. To accomplish this goal, the automotive parts company acts on innovation ideas that emerge from processors or managers:

But one of the things that we did […] is a module or a conveyor for reclamation. As you can see there is a lot of sorting that needs to be done. To me, it’s no different than selecting for a store […] I don’t see why we can’t have a conveyor that will sort out each [returned] product by the zones in which it goes in the building, and then for us to put it away, instead of it being such a manual process.

4.4 Case D – processing furniture returns

Several types of agents are involved in the processing of furniture returns, including the customer, customer service representative, line processors, shop processors, quality control processors and the distribution manager. Shop processors and quality control processors have the most influence. The primary resources include product returns and the information system.

Product returns arrive at the distribution center primarily on company delivery trucks and are stored in “yard-dogs” (“yard-dogs” are semi-trailers used for storage in cargo yards or warehouse facilities) until the line processors are ready for them. Each processing bay in the returns department has two conveyor lines in it – a large line used for big furniture pieces and mattresses, and a second slender line used for mirrors and headboards. Line processors first match the furniture pieces to the return manifests, and then send the pieces to the quality department for a safety inspection and to determine the best disposition option for the pieces:

When a piece comes back, it’s inspected by our quality department. If a piece is deemed that it’s perfect, we will just box it back up or wrap it back up and put it back on the shelves or in the racks. If the piece has some damage, it will be repaired. If it can be repaired to new, it will be put back in the racks, if it can be repaired and sold as is, then that piece will go over to our clearance area to be shipped out to one of our clearance stores and be sold. If a piece is damaged beyond repair and cannot be sold because it’s a liability if we did sell it because the chair might not be strong enough, we just total loss that piece, where we will take a hit on it in terms of inventory. If the piece is in between, then we will use liquidators and they’ll buy the product and sell it.

(Elliot, Regional Manager of Distribution, Furniture company)

The furniture company utilizes a customized, legacy system to store customer transaction data, returns manifests, and track inventory. Elliot cited the ability to take ideas from processors and customize their legacy system as key to the facility’s success in processing returns. To make modifications to the information system, the furniture company has an established chain of command that starts with an approval from the Vice President of Distribution before being sent to the in-house programmers who will actually make the requested modifications.

An agent’s established schema is of paramount importance in the quality and shop departments, as it is the primary line of defense in protecting customers and ensuring all returned furniture pieces are safe enough to be resold:

Just because the customer said that one [problem], we have to make sure that every inch of that piece is still good. So [the quality control processor] will go and inspect it, and based on that inspection and the years he has been here, and knowing what can and can’t be fixed, it will go to the shop and [the shop processor] will decide whether he can fix it 100% so it can be sold as new, or what the next step would be, on down the line to either clearance or salvage.

(Elliot, Regional Manager of Distribution, Furniture company)

Because of the amount of knowledge required to work in the quality or shop departments, the furniture company does not hire people directly into these job positions, but instead promotes employees from within other areas of the company who have developed a well-rounded schema. This promotional process ensures that quality and shop agents are adept at learning new product information, as well as learning from other agents:

The jobs that are most technical are shop and are quality control just in terms of the knowledge you need to have in terms of how to fix things properly so [the furniture piece] wears correctly. So those promotional positions-most people don’t come in in a shop position, they grow into it based on learning the rest of the business. So, in order to get that job, you have to pass a test in terms of knowing how to repair things and how to do things. But then you’re continuously taught and given tests to make sure that you haven’t lost the ability to do certain things because maybe you say you’re fixing something, but you’re not fixing something properly.

(Elliot, Regional Manager of Distribution, Furniture company)

The furniture company relies on insights generated from the day-to-day processing of returns in the shop and quality departments to identify furniture pieces that are repeatedly returned:

So, when [processors] see pieces coming back and they see a problem they might take a whole dresser apart. They might take an upholstered sofa apart just to look at how it was actually built. And then if they find a problem, that’s when pictures are taken and communicated back to the buyer so changes can be made.

(Elliot, Regional Manager of Distribution, Furniture company)

The furniture company routinely shares insights generated from agents to other distribution centers through videos, and encourages the agents who came up with the idea to create the videos.

4.5 Case E – processing laboratory equipment returns

Agents involved in the processing of laboratory equipment returns include the customer, customer service representative, sales representative, in-bound processor, and service center processor. The reason for a product return has the most influence on the returns process, so the customer service representatives actually have the most influence. Resources include the product and information system.

Because of the lower volume and potential costs of product returns to the laboratory equipment company, returned items are screened extensively on the reason for the return:

So, we look at the reason that drives [the return]. So, either the product is defective – whether it’s an out of box failure or it’s some type of warranty issue. The incorrect application was developed, so the sales guys go out there and determine what the needs of the companies are. Sometimes they don’t always get it right. So, the type of application will vary what type of instruments they need – do they need wet chemistry, dry chemistry in that field? So, if the sales guy gets that wrong, it’s an acknowledgement that the sales guy got that wrong, we will bring that product back. We’re probably going to get another product that will soothe their needs […] There’s also some customer driven reasons. So, the customer made a mistake ordering the product. They change their mind and they no longer want it. Or the damage was caused in transit, but they provided the transportation, instead of us providing the transportation.

(Nick, Operations Manager, Laboratory equipment company)

Once the reason for the return is determined, either an onsite fix will be attempted, or a return merchandise authorization is generated and the product is shipped back to the warehouse to be repaired:

So, the process then is that [the salespeople] try to do an onsite fix if they can. If they can’t, then they go back through their chain of command to get permission to get the out-of-box failure authorized. So, then customer service gets contacted, they do a no charge order, which generates a movement for me to ship out a replacement item and for [them] to return the broken one. The [sales] guy on site will repack that and ship it back to us.

(Nick, Operations Manager, Laboratory equipment company)

Once shipped back to the warehouse, in-bound processors reenter returned items in new condition into inventory or send out-of-box failure equipment to the service center to be fixed. Opened returned items or irreparable out-of-box failure equipment cannot be resold and are typically scrapped or returned to the manufacturer.

The main role processor schemas play during the analysis of returned products at the laboratory equipment company is to determine whether a returned item is in new condition or determining if an instrument can be repaired. Processors learn information related to product condition by looking it up in the company’s database, but may reach out to product managers involved in product design with specific repair questions:

The person that works in receiving has a decision to make- is it in new condition? So, new condition means it’s originally packaged, it’s sealed, and can go right back on the shelf. If it’s been opened, because of the nature of our products, because chemistry gets ran through and chemicals get ran through it, it goes right to scrap if it’s consumable. Versus the instruments will go through the service center, and they will look at them.

(Nick, Operations Manager, Laboratory equipment company)

The laboratory equipment company uses a return materials authorization procedure within a Sage ERP system to store transaction data, track the movement of returned items within their warehouse and issue credit to the customer. This information system and the product returns process was only a year old at the time of the site visit. Prior to that, processing and analyzing returns was messy and riddled with inaccurate information about what a customer was authorized to return:

We changed our customer returns process about a year ago. So, what we did was we flow-charted everything out. And by using the different departments, we kind of brainstormed together to come up with, well what’s the best answer? Because we didn’t have a great answer at the time, everything that we’re talking about now has been active for about a year. So, before there was a lot of miscommunication between customer service and what was authorized to be returned- what was the dollar amount, who was authorizing it? So, it changed because we didn’t have a good solid policy.

(Nick, Operations Manager, Laboratory equipment company)

5. Cross-case findings

This section analyzes the data from each case company jointly with the goal of synthesizing cross-case patterns (Yin, 2018), of agent interactions and organization, schema and learning and emergent adaptations. Table IV summarizes the key cross-case CAS findings by company.

5.1 Agent interactions and organization

In CAS, agents interact and make decisions based on imperfect information and their intuition, which impacts other parts of the system and creates feedback loops (Chan, 2001; Holland, 2014; Wysick et al., 2008). Our data indicates all five case companies have some type of customer service, front-line processing, and manager agents involved in processing product returns, however, customer service and manager agents play a background role during the actual inspection and analysis of returned products. The following discussion focuses on the processing agents. Cases A, D and E have more than one level of processing agents, and the more senior processing agents (e.g. exception specialists in Case A) possess greater training and experience.

Each case company took a different approach to structuring agent interactions and organization:

  • In Case A, agent interactions are limited by design to reduce risk, and returns requiring more intensive interactions are routed to an exception specialist. The ability of individual agents to self-organize is constrained by the rigid separation of processing agents into specialized levels.

  • In Case B, agent interactions follow a general process of looking up products, but each agent is afforded autonomy in determining the best way to process a return. Agents are given the freedom to self-organize, however, many prefer not to due to their tenure and wanting to stick with the status quo.

  • In Case C, agent interactions are automated as much as possible to conserve agent capabilities for problematic returns that arise. Agents are encouraged to self-organize and join company-sponsored groups around topics that interest them.

  • In Case D, agent interactions and decisions are split across several different agents, so that agents with the most experience make the more important decisions. Agents self-organize and are encouraged to share ideas with other agents and management.

  • In Case E, agent interactions depend on the reason for the return, but all must follow the newly implemented company-wide policy. Agents can self-organize to process out of box failures, but otherwise may not, due to the low volume of returns.

Looking across these cases, informants utilize different types of processing agents – some have only a single type of agent while others have two or more different types. Cases B and C have only one general type of processing agent, while Cases A and E have two types of processing agents, and Case D has three types of processing agents. The required level of agent specialization appears to be correlated with the risk and/or costs associated with the returns. Product returns in Cases A, D and E can often be of expensive items, so maintaining general and more specialized processing agents can help to minimize the financial risk of these returns. Returns in Case D require a third type of processing agent due to the additional risk of serious injury associated with reselling a defective piece of furniture.

When it comes to how agents interact, agents in Cases B, C and D follow a general processing pattern, but are allowed to self-organize and learn from other agents when analyzing a returned item. Agents in Cases A and E have less autonomy and are more restricted in how they interact and organize due to rigid track structures and low return rates, respectively.

5.2 Agent schema and learning

In CAS, agents evaluate interactions they participate in to identify relevant portions of their existing schema to apply when making routine decisions or identifying novel situation-specific solutions (Dooley, 1997; Holland, 2014). Agents can also learn new solutions from other agents (Schon, 1975). Across the five case examples, different depths of agent schema and amounts of agent learning were described:

  • In Case A, agent schema cover types of products, common issues, and potential fixes. Agents often encounter new products and rely on first-order learning via the company’s website or Google to gather new information.

  • In Case B, agent schema include both product-related and sports-related nodes of information. Agents rely on first-order learning via printed catalogs to learn new product information.

  • In Case C, agent schema consist of general knowledge of the types of car parts and their condition. Agents access new information through first-order learning with the information system or second-order learning by joining specialized company groups.

  • In Case D, agent schema cover types of furniture, materials, common quality issues, potential fixes, and future wear patterns. Agents acquire new information through first-order learning by working in different parts of the warehouse and through second-order learning from other agents throughout the company.

  • In Case E, agent schema include product condition and potential fixes. Agents learn new information primarily through first-order learning via the company’s database, but occasionally through second-order learning with other agents.

Looking across these cases, agent schemas are smaller and possession of important information is split amongst different types of agents in Cases A, D and E. Cases A, D and E all have agents with shallower schemas do the initial screening of product returns. These agents are able to process simple returns, but more complicated returns are handled by agents with more developed schema. Agents in Cases B and C must handle all types of returns, regardless of the development of their schema. This is not an issue for agents in Case C due to the sophisticated automation of the information system, but did contribute to bogging down the processing of product returns during peak volume times in Case B.

When it comes to how agents learn, agents in Cases A, B and E learn primarily through first-order learning and company resources, while agents in Cases C and D encourage agents to learn primarily through second-order learning in agent-to-agent interactions. Learning in interactions appears to occur more quickly and effectively than through resources, since managers in Cases C and D were better able to reassign labor from other parts of their warehouse to help out with processing product returns during high volume. Managers in Cases A and B typically do not reassign labor, and instead let processing agents work through the backlog as quickly as they can.

5.3 Emergent adaptations

Varied agent schema, self-organization and non-linearity allow new emergent ideas to develop in CAS that are super-additive in nature, which enable agents to adapt their behaviors to better match their surrounding environment (Choi et al., 2001; Holland, 2014; Surana et al., 2005). Comparing the five cases reveals informants possess different abilities to generate product-return-focused emergent ideas, and different success rates of agents implementing new adaptations:

  • In Case A, the company has successfully implemented agent-generated ideas in out-bound shipments, but has been unable to generate product-return-focused ideas. Without emergent ideas, agents are unable to adapt.

  • In Case B, the company relies on agents to generate emergent ideas, but often settles for incremental changes due to a lack of resources. Here, emergent ideas are generated, but the success rate of agent adaptations is low because agents often resist adapting and prefer to stick with the status quo.

  • In Case C, the company relies on agent-generated emergent ideas as the primary means for identifying cost savings. The success rate of agent adaptations is high, since sufficient resources and support are available to help agents adapt. One of the most successful adaptations was the development of a conveyor system for putting away returned products.

  • In Case D, the company has established pathways for collecting, evaluating, and implementing agent-generated emergent ideas. The success rate of agent adaptations is high, since adaptation is rewarded and shared with other distribution centers. One of the most successful adaptations is leveraging processing agents to identify and solve quality control issues associated with frequently returned furniture pieces.

  • In Case E, the company is open to agent-generated emergent ideas, but often waits to implement them until problems become painful and change must occur. While overhauling the reclamation process was a successful adaptation, agents often lack the motivation to adapt unless absolutely necessary.

Looking across the cases, Cases C and D stood out in their ability to generate emergent ideas and implement adaptations. Case A could not generate or implement product-return-focused emergent ideas at all, and Cases B and E were able to generate emergent ideas, but often fell short when it came to motivating agents to implement them.

Connecting these findings with the findings in Sections 5.1 and 5.2 helps to explain the differences observed across informants. In Cases C and D, the company grants processing agents more autonomy to self-organize, make decisions and encourages second-order learning, resulting in an internal culture where processing agents feel comfortable enough to adapt and generate new ideas. In contrast, processing agents rely on first-order learning only and are given very little autonomy to self-organize and make decisions in Case A. While the company states they are open to emergent ideas, the rigid structure embedded in the product returns department inhibits emergent ideas and adaptations from forming because they are viewed as risky. Cases B and E fall in between Cases C/D and A, and show some ability to generate emergent ideas and adapt. In Cases B and E, processing agents primarily rely on first-order learning and have some autonomy to self-organize and make decisions, although agents often are unmotivated to generate emergent ideas and adapt.

6. Discussion and conclusions

The current research explores the processing of product returns from a CAS perspective across five different companies. The cross-case insights indicate that companies who encourage agent autonomy in interactions, decision making, and learning are better able to generate emergent ideas and adapt. These findings are consistent with previous CAS research (e.g. Choi et al., 2001) that revealed the critical role human and social behaviors play in forward supply chains. However, the current research uniquely demonstrates that the tenants of CAS theory also apply to the processing of product returns. The academic and managerial implications of these findings are discussed in greater detail in the following section.

6.1 Implications

For scholars, the current research builds on the work of Hazen et al. (2012) and Tan and Kumar (2006) by revealing how the behaviors of agents at key decision points in the returns process create adaptability in the reverse supply chain. From a theoretical perspective, we extend previous work (e.g. Choi et al., 2001; Surana et al., 2005) demonstrating how the dimensions of CAS theory are relevant in understanding key aspects of the supply chain, and extend this work specifically into the realm of processing product returns. Thus, through the exploration of product returns processing at five different companies, the current research contributes to the reverse logistics, supply chain management and CAS literatures by beginning to fill in a gap in our understanding of the complex adaptive mechanics inherent in product return processes.

Supply chain and reverse logistics managers can use the results of this research to inform how they manage processing agents on a day-to-day basis. Managers need to begin by evaluating the cost, risk, and return rate associated with their returned products. If returned products are high cost, high risk and/or have a high return rate, managers need to seriously consider developing different types of processing agents or sophisticated automated systems to boost the adaptability of the reverse supply chain.

Whether managers prefer to develop new types of agents or automated systems, the goal of both approaches is to decrease the depth of mental schema each processing agent is required to possess. Decreasing the depth of processing agents’ mental schema gives managers more flexibility with reassigning labor from shipping outbound orders. Reassigned processing agents can do the initial screening of returns as long as they are familiar with the types of products sold. Managers can then dedicate the experience of full-time processing agents to the more “difficult” returns, without overstressing or fatiguing experienced processing agents. During normal volume seasons, decreasing the depth of agents’ mental schema also boosts the adaptability of the reverse supply chain because processing agents are not overburdened with maintaining too much information in their schema.

Instead, processing agents have the capacity to experiment with ideas, interact and share information with other agents. While managers need to ensure processing agents have access to appropriate company resources, agent-to-agent interactions may be the most efficient way to create new and invaluable firm-specific insight. In interactions, agents not only learn factual information about products, but also get a sense of the past experiences and heuristic information other agents possess about specific products (e.g. blue jerseys with short sleeves have NFL teams printed on them, while long sleeve jerseys have NHL teams printed on them). It is important to note, however, that processing agent interactions will likely need to be encouraged and/or incentivized initially, to motivate agents to interact and break long-standing agent habits. Firm-wide initiatives such as departmental social outings and formalized mentor/mentee programs may be beneficial in achieving this objective.

The biggest insight of the current research for managers echoes the tenets of CAS theory – in order to truly nurture the adaptability of the reverse supply chain and generate emergent ideas, managers need to promote processing–agent interactions, allow self-organization, optimize the depth of processing-agent schema and encourage second-order learning. As Cases A, B and E showed, supporting only a portion of these behaviors will limit the ability of processing agents to generate emergent ideas and/or adapt. Cases C and D demonstrated that in order to boost the adaptability of the reverse supply chain, managers need to craft a dynamic environment that encourages processing agents to self-organize based upon their diverse schema and intuition, to anticipate problems, and brainstorm solutions. Once set up, managers need to resist the urge to create rigid guidelines or policies to “control” the processing of product returns, as these desires will “choke” the organic adaptability of the system. Instead, managers will need to take a step back and look to the ideas that emerge from processing–agent interactions and self-organization.

6.2 Limitations and directions for future research

The current research is not without limitations which provide opportunities for future research. The first limitation revolves around the context of the research. Informants at five companies, selling a variety of products were interviewed to deepen our understanding of the processing of product returns. However, the findings may not be generalizable to companies selling different types of products not included in the current research. Exploring the types of agents, agent interactions, and agent schema present in other reverse supply chains can further expand upon the patterns present in the five case companies. Additionally, future research might seek to validate and extend our findings by including employees holding more varied positions within the supply chain (i.e. front-line and executive level) to help us further understand how agents form decision-making rules and approach interactions with other agents.

Second, the current research explores the importance of types of agents, agent interactions, and agent schema in the processing of product returns, but cannot evaluate the relative quantitative importance of these components in a product returns system. Future deductive, quantitative research is needed to better understand the relative importance of different types of agents, agent interactions and the depth of agent schema in the processing of product returns.

Understanding how human-agent behaviors impact the processing of product returns remains a relatively unexplored avenue for future research. Little is known about how human limitations (e.g. mental fatigue, pressure, or conflict) color the analysis of product returns. While giving processing agents more autonomy to decide how to process product returns has the potential for incidents of human error to occur, it also allows managers to tap into the innate creativity and innovation these agents possess both individually and collectively. With additional research, we hope managers will no longer view product returns as a problem to optimize away, but as a valuable source of information and adaptation that can revolutionize their current product returns paradigm.


CAS framework of processing product returns

Figure 1

CAS framework of processing product returns

Case study protocol questions

 1. What systems and processes do you use to manage returns?
 2. What are the different end-life options for returned products?
 3. How many vendors/customers do you accept returns from?
 4. What similarities in your returns process exist across most of your vendors? Unique to each vendor?
 5. How do you try to reduce returns?
 6. Who is involved in reducing returns? (Functions? Supply chain partners?)
 7. Walk me through what happens to products/components that are returned.
 8. How important is [the returns process] to your organization?
 9. How are return initiatives developed? Who helps develop them? What areas of your business do these initiatives help?
10. What kind of software do you use to manage your returns?
11. What are the key functionalities of this software?
12. What are the drivers and motivators for product return activities? Why are they important?
13. What are the outcomes of product return activities? Why are they important?

Summary of case companies

Case Product Location Customer type Sales Annual return volume
A Technology products US based Multinational $25bn 2m pieces
B Apparel US based Multinational $12bn 150,000 pieces
C Automotive parts US based Multinational $9bn 5m pieces
D Furniture US based National $2bn 500,000 pieces
E Laboratory equipment Internationally based Multinational $350m 200 pieces

Profile of informants

Industry Position in supply chain Types of returns Informant name Informant job title Informant description
Case A – Technology products company
Technology products Distributor/manufacturer Devices, parts Belle Senior Logistics Manager of Reverse Logistics With 8 years in the industry, 8 years at the company, and 1 year in her current position, Belle is responsible for overseeing reverse logistics teams and taking care of their administrative duties
Technology products Distributor/manufacturer Devices, parts Dave Logistics Director With 12 years in the industry, 12 years at the company, and 5 years in his current position, Dave is responsible for all facets of logistics operations
Cases B–E comparison case companies
Case B – Apparel Manufacturer Clothing items John Director of Distribution With 13 years’ experience in logistics, 5 years at the company, and 1.5 years in his current position, John is responsible for the financial and service aspects of all in-bound and out-bound shipments
Case C – Automotive parts Distributor/retailer Machinery, parts Mike Out-bound Operations Manager With 4 years in the industry, 4 years at the company, and 1.5 years in his current position, Mike is responsible for selection, replenishment and shipping
Case D Furniture Retailer Furniture Elliot Regional Manager of Distribution With 25 years in the industry, 9 years at the company, and 3 years in his current position, Elliot is responsible for overseeing the distribution of three facilities and outlet stores
Case E – Laboratory equipment Retailer Machinery, parts Nick Operations Manager With 9 years’ experience in logistics, 2 years at the company and 2 years in his current position, Nick is responsible for out-bound shipping, returns, and facility management

Note: Names are pseudonyms to protect the anonymity of informants

Summary of cross-case key findings

CAS dimension
Case: return category Agent interactions Agent organization Agent schema Agent learning Emergence Adaptations Adaptation examples
A: technology products Interactions limited to reduce risk Limited self-organization allowed Covers products, common issues and potential fixes First-order only Non-Existent Unable to adapt n/a
B: apparel Interactions not limited Self-organization allowed, but agents may avoid due to effort required Include product-related and sports-related knowledge First-order primarily Ideas emerge from agents Agents often resist adapting n/a
C: automotive parts Interactions encouraged Self-organization encouraged General knowledge of parts and condition First order and second order Relies on agent-generated emergent ideas Agents adapt successfully Conveyor system for putting returns away
D: furniture Interactions separated across 3 types of agents based on experience Self-organization is encouraged Covers types of furniture, materials, common issues, potential fixes, and wear patterns split amongst different agents First order and second order Relies on agent-generated emergent ideas Agents adapt successfully Relying on agents to identify frequently returned pieces and deconstructing them to identify the source of the problem(s)
E: laboratory equipment Interactions guided by new company policy Self-organization can occur with some returns, but often does not due to low return volume Includes product condition and potential fixes First-order primarily Ideas emerge from agents Agents unmotivated to adapt until absolutely necessary Overhauling the product returns process and educating employees company-wide


Allen, P.M. (2000), “Knowledge, ignorance, and learning”, Emergence, Vol. 2 No. 4, pp. 78-103.

Altug, M.S. and Aydinliyim, T. (2016), “Counteracting strategic purchase deferrals: the impact of online retailers’ return policy decisions”, Manufacturing & Service Operations Management, Vol. 18 No. 3, pp. 376-392.

Autry, C.W., Daugherty, P.J. and Richey, R.G. (2001), “The challenge of reverse logistics in catalog retailing”, International Journal of Physical Distribution & Logistics Management, Vol. 31 No. 1, pp. 26-37.

Bernon, M., Upperton, J., Bastl, M. and Cullen, J. (2013), “An exploration of supply chain integration in the retail product returns process”, International Journal of Physical Distribution & Logistics Management, Vol. 43 No. 7, pp. 586-608.

Blackburn, J.D., Guide, V.D.R. Jr, Souza, G.C. and Van Wassenhove, L.N. (2004), “Reverse Supply Chains for Commerical Returns”, California Management Review, Vol. 46 No. 2, pp. 6-22.

Blanchard, D. (2010), Supply Chain Management: Best Practices, 2nd ed., Hoboken Wiley & Sons, Hoboken, NJ.

Bonifield, C., Cole, C. and Schultz, R.L. (2010), “Product returns on the internet: a case of mixed signals?”, Journal of Business Research, Vol. 63 Nos 9-10, pp. 1058-1065.

Bower, A.B. and Maxham, J.G. (2012), “Return shipping policies of online retailers: normative assumptions and the long-term consequences of fee and free returns”, Journal of Marketing, Vol. 76 No. 5, pp. 110-124.

Chakkol, M., Johnson, M., Raja, J. and Raffoni, A. (2014), “From goods to solutions: how does the content of an offering affect network configuration?”, International Journal of Physical Distribution & Logistics Management, Vol. 44 Nos 1/2, pp. 132-154.

Chan, S. (2001), “Complex adaptive systems”, ESD. 83 Research Seminar in Engineering Systems, Vol. 31, pp. 1-9.

Choi, T.Y., Dooley, K.J. and Rungtusanatham, M. (2001), “Supply networks and complex adaptive systems: control versus emergence”, Journal of Operations Management, Vol. 19 No. 3, pp. 351-366.

Daunt, K.L. and Harris, L.C. (2012), “Motives of dysfunctional customer behavior: an empirical study”, Journal of Services Marketing, Vol. 26 No. 4, pp. 293-308.

Davis, S., Hagerty, M. and Gerstner, E. (1998), “Return policies and the optimal level of ‘Hassle’”, Journal of Economics and Business, Vol. 50 No. 5, pp. 445-460.

Dennis, S. (2018), “The ticking time bomb of E-commerce returns”, Forbes, February 14, available at: www.forbes.com/sites/stevendennis/2018/02/14/the-ticking-time-bomb-of-e-commerce-returns/#4430c15f4c7f (accessed August 28, 2018).

Dooley, K.J. (1997), “A complex adaptive systems model of organization change”, Nonlinear Dynamics, Psychology, and Life Sciences, Vol. 1 No. 1, pp. 69-97.

Dubois, A. and Gadde, L.-E. (2002), “Systematic combining: an abductive approach to case research”, Journal of Business Research, Vol. 55 No. 7, pp. 553-560.

Edmondson, A.C. and McManus, S.E. (2007), “Methodological fit in management field research”, Academy of Management Review, Vol. 32 No. 4, pp. 1246-1264.

Gesenhues, A. (2017), “Report: 95% of online shoppers say a positive return experience drives loyalty”, Marketing Land: Third Door Media, available at: https://marketingland.com/e-commerce-report-95-online-shoppers-say-positive-return-experience-drives-loyalty-216820 (accessed January 23, 2019).

Goldman, I. (2016), “Managing returns in a complex omnichannel environment: retail touch points”, available at: www.retailtouchpoints.com/features/executive-viewpoints/managing-returns-in-a-complex-omnichannel-environment2016) (accessed August 10, 2018).

Hawkins, D.I. and Mothersbaugh, D.L. (2010), Consumer Behavior: Building Marketing Strategy, 11th ed., McGraw-Hill/Irwin, New York, NY.

Hazen, B.T., Hall, D.J. and Hanna, J.B. (2012), “Reverse logistics disposition decision-making: developing a decision framework via content analysis”, International Journal of Physical Distribution & Logistics Management, Vol. 42 No. 3, pp. 244-274.

Hjort, K. and Lantz, B. (2016), “The impact of returns policies on profitability: a fashion e-commerce case”, Journal of Business Research, Vol. 69 No. 11, pp. 4980-4985.

Hjort, K., Lantz, B., Ericsson, D. and Gattorna, J. (2013), “Customer segmentation based on buying and returning behaviour”, International Journal of Physical Distribution & Logistics Management, Vol. 43 No. 10, pp. 852-865.

Ho, G.T.S., Choy, K.L., Lam, C.H.Y. and Wong, D.W.C. (2012), “Factors influencing implementation of reverse logistics: a survey among Hong Kong businesses”, Measuring Business Excellence, Vol. 16 No. 3, pp. 29-46.

Holland, J.H. (1992), “Complex adaptive systems”, Daedalus, Vol. 121 No. 1, pp. 17-30.

Holland, J.H. (1995), Hidden Order: How Adaptation Builds Complexity, Helix Books, New York, NY.

Holland, J.H. (2014), Complexity: A Very Short Introduction, 1st ed., Oxford, New York, NY.

Jack, E.P., Powers, T.L. and Skinner, L. (2010), “Reverse logistics capabilities: antecedents and cost savings”, International Journal of Physical Distribution & Logistics Management, Vol. 40 No. 3, pp. 228-246.

Janakiraman, N. and Ordonez, L. (2012), “Effect of effort and deadlines on consumer product returns”, Journal of Consumer Psychology, Vol. 22 No. 2, pp. 260-271.

Janakiraman, N., Syrdal, H.A. and Freling, R. (2016a), “The effect of return policy leniency on consumer purchase and return decisions: a meta-analytic review”, Journal of Retailing, Vol. 92 No. 2, pp. 226-235.

Janakiraman, N., Syrdal, H.A. and Freling, R.E. (2016b), “How to design a return policy”, Harvard Business Review, available at: https://hbr.org/2016/08/how-to-design-a-return-policy (accessed December 10, 2018).

Jones, K. (2018), “Why marketers can’t survive without investing in omnichannel”, Forbes, August 10, available at: www.forbes.com/sites/forbesagencycouncil/2018/08/10/why-marketers-cant-survive-without-investing-in-omnichannel/#5e8f354c49e1 (accessed August 28, 2018).

Kocabasoglu, C., Prahinski, C. and Klassen, R.D. (2007), “Linking forward and reverse supply chain investments: the role of business uncertainty”, Journal of Operations Management, Vol. 25 No. 6, pp. 1141-1160.

Kovacs, G. and Spens, K.M. (2005), “Abductive reasoning in logistics research”, International Journal of Physical Distribution & Logistics Management, Vol. 35 No. 2, pp. 132-144.

Kumar, V., Aksoy, L., Donkers, B., Venkatesan, R., Wiesel, T. and Tilmanns, S. (2010), “Undervalued or overvalued customers: capturing total customer engagement value”, Journal of Service Research, Vol. 13 No. 3, pp. 297-310.

Lehr, C.B., Thun, J.-H. and Milling, P.M. (2013), “From waste to value – a system dynamics model for strategic decision-making in closed-loop supply chains”, International Journal of Production Research, Vol. 51 No. 13, pp. 4105-4116.

Li, G., Yang, H., Sun, L., Ji, P. and Feng, L. (2010), “The evolutionary complexity of complex adaptive supply networks: a simulation and case study”, International Journal of Production Economics, Vol. 124 No. 2, pp. 310-330.

Lindsey, K. (2016), “Why retail has a growing reverse supply chain problem- and how to fix it”, available at: www.retaildive.com/news/why-retail-has-a-growing-reverse-supply-chain-problemand-how-to-fix-it/422136/ (accessed June 16, 2018).

Nilsson, F. (2006), “Logistics management in practice – towards theories of complex logistics”, International Journal of Logistics Management, Vol. 17 No. 1, pp. 38-54.

Nilsson, F. and Darley, V. (2006), “On complex adaptive systems and agent-based modelling for improving decision-making in manufacturing and logistics settings: experiences from a packaging company”, International Journal of Operations & Production Management, Vol. 26 No. 12, pp. 1351-1373.

Nilsson, F. and Gammelgaard, B. (2012), “Moving beyond the systems approach in SCM and logistics research”, International Journal of Physical Distribution & Logistics Management, Vol. 42 Nos 8/9, pp. 764-783.

Oghazi, P., Karlsson, S., Helltrom, D. and Hjort, K. (2018), “Online purchase return policy leniency and purchase decision: mediating role of consumer trust”, Journal of Retailing and Consumer Services, Vol. 41, pp. 190-200.

Pathak, S.D., Dilts, D.M. and Mahadevan, S. (2009), “Investigating population and topological evolution in a complex adaptive supply network”, Journal of Supply Chain Management, Vol. 45 No. 3, pp. 54-67.

Pathak, S.D., Day, J.M., Nair, A., Sawaya, W.J. and Kristal, M.M. (2007), “Complexity and adaptivity in supply networks: building supply network theory using a complex adaptive systems perspective”, Decision Sciences, Vol. 38 No. 4, pp. 547-580.

Petersen, J.A. and Kumar, V. (2009), “Are product returns a necessary evil? Antecedents and consequences”, Journal of Marketing, Vol. 73 No. 3, pp. 35-51.

Petro, G. (2018), “All fashion is fast fashion”, Forbes, August 12, available at: www.forbes.com/sites/gregpetro/2018/08/12/all-fashion-is-fast-fashion/#4a3decfa62f6 (accessed August 28, 2018).

Piron, F. and Young, M. (2001), “Retail borrowing: insights and implications on returning used merchandise”, International Journal of Retail & Distribution Management, Vol. 28 No. 1, pp. 27-36.

Potts, W.K. (1984), “The chorus-line hypothesis of manoeuvre coordination in avian flocks”, Nature, Vol. 309 No. 5966, pp. 344-345.

Powers, T.L. and Jack, E.P. (2013), “The influence of cognitive dissonance on retail product returns”, Psychology & Marketing, Vol. 30 No. 8, pp. 724-735.

QSR International (2012), “NVivo qualitative data analysis software (Version Version 10)”.

Rogers, D.S., Lambert, D.M., Croxton, K.L. and Garcia-Dastugue, S.J. (2002), “The returns management process”, The International Journal of Logistics Management, Vol. 13 No. 2, pp. 1-18.

Safdar, K. (2018), “How your returns are used against you at best buy, other retailers”, Wall Street Journal, March 13, available at: www.wsj.com/articles/how-your-returns-are-used-against-you-at-best-buy-other-retailers-1520933400 (accessed January 23, 2019).

Sarkis, J., Zhu, Q. and Lai, K.-h. (2010), “An organizational theoretic review of green supply chain management literature”, International Journal of Production Economics, Vol. 130 No. 1, pp. 1-15.

Schon, D.A. (1975), “Deutero-learning in organizations: learning for increased effectiveness”, Organizational Dynamics, Vol. 4 No. 1, pp. 2-16.

Schrotenboer, A.H., Wruck, S., Roodbergen, K.J., Veenstra, M. and Dijkstra, A.S. (2017), “Order picker routing with product returns and interaction delays”, International Journal of Production Research, Vol. 55 No. 21, pp. 6394-6406.

Simon, H.A. (1991), “Bounded rationality and organizational learning”, Organizational Science, Vol. 7 No. 1, pp. 125-134.

Stacey, R.D., Griffin, D. and Shaw, P. (2000), Complexity and Management: Fad or Radical Challenge to Systems Thinking?, Routledge, New York, NY.

Stock, J.R. and Mulki, J.P. (2009), “Product returns processing: an examination of practices of manufacturers, wholesalers/distributors, and retailers”, Journal of Business Logistics, Vol. 30 No. 1, pp. 33-62.

Stock, J.R., Speh, T. and Shear, H. (2006), “Managing product returns for competitive advantage”, MIT Sloan Management Review, Vol. 48 No. 1, pp. 57-62.

Storer, M., Hyland, P., Ferrer, M., Santa, R. and Griffiths, A. (2014), “Strategic supply chain management factors influencing agribusiness innovation utilization”, International Journal of Logistics Management, Vol. 25 No. 3, pp. 487-521.

Strauss, A.L. and Corbin, J. (1998), Basics of Qualitative Research: Techniques and Procedures for Developing Grounded Theory, Sage Publications, Thousand Oaks, CA.

Su, X. (2009), “Consumer returns policies and supply chain performance”, University of Pennsylvania Scholarly Commons, Vol. 11 No. 4, pp. 595-612.

Surana, A., Kumara, S., Greaves, M. and Raghavan, U.N. (2005), “Supply chain networks: a complex adaptive systems perspective”, International Journal of Production Research, Vol. 43 No. 20, pp. 4235-4265.

Tan, A.W.K. and Kumar, A. (2006), “A decision-making model for reverse logistics in the computer industry”, The International Journal of Logistics Management, Vol. 17 No. 3, pp. 331-354.

Turrisi, M., Bruccoleri, M. and Cannella, S. (2013), “Impact of reverse logistics on supply chain performance”, International Journal of Physical Distribution & Logistics Management, Vol. 43 No. 7, pp. 564-585.

Vohs, K.D., Baumeister, R.F., Schmeichel, B.J., Twenge, J.M., Nelson, N.M. and Tice, D.M. (2008), “Making choices impairs subsequent self-control: a limited-resource account of decision making, self-regulation, and active initative”, Journal of Personality and Social Psychology, Vol. 94 No. 5, pp. 883-898.

Welson-Rossman, T. (2018), “How the challenge of finding the right shoe size launched this woman’s company”, Forbes, August 21, available at: www.forbes.com/sites/traceywelsonrossman/2018/08/21/one-size-does-not-fit-all/#18cbab545273 (accessed August 28, 2018).

Wysick, C., McKelvey, B. and Hulsmann, M. (2008), “‘Smart Parts’ supply networks as complex adaptive systems: analysis and implications”, International Journal of Physical Distribution & Logistics Management, Vol. 38 No. 2, pp. 108-125.

Yin, R.K. (2018), Case Study Research and Applications, Sage, Thousand Oaks, CA.


The authors gratefully acknowledge and thank the anonymous reviewers and editor Gammelgaard for their helpful comments during the review process. The authors gratefully thank the University of South Florida’s L. Rene Gaiennie Endowment and the University of South Florida’s Center for Supply Chain Management and Sustainability for funding travel to collect data and the NVivo 10 software necessary to complete this research.

Corresponding author

Jennifer A. Espinosa is the corresponding author and can be contacted at: espinosaj@rowan.edu

About the authors

Jennifer A. Espinosa, PhD, is Assistant Professor at Rowan University in the Marketing and Business Information Systems Department of the Rohrer College of Business in Glassboro, New Jersey. Her main research interests lie in problems that occur on both sides of front-line transactions between companies and customers, including brand image formation and product returns. Her work has been published in the Journal of Business Research, the Journal of Product and Brand Management and the International Journal of Value Chain Management.

Donna Davis, PhD, is Full Professor in Marketing at the University of South Florida in the Marketing Department of the Muma College of Business in Tampa, Florida. Her main research interests include supply chain management and business-to-business branding. Her work has been published in several leading supply chain management journals, including the International Journal of Physical Distribution and Logistics Management, the Journal of Business Logistics and the Journal of Supply Chain Management.

James Stock, PhD, is Distinguished University Professor at the University of South Florida in the Marketing Department of the Muma College of Business in Tampa, Florida. His main research interests include supply chain management, logistics and reverse logistics. His work has been published in several leading journals, including the International Journal of Physical Distribution and Logistics Management, MIT Sloan Management Review, the Journal of Business Logistics and the Journal of the Academy of Marketing Science.

Lisa Monahan, PhD, is Assistant Professor in Marketing at the School of Business, Meredith College in Raleigh, North Carolina. She has 14 years of experience working in advertising and brand management at a variety of firms including J. Walter Thompson, Unilever, Hanes Brands, and Citigroup. Her main research interests include branding and advertising. Her work has been published in the Journal of Product and Brand Management, the proceedings of the American Marketing Association’s Winter Educator’s Conference, the Academy of Marketing Science Annual Conference and the Society for Marketing Advances.