Rooting out the root causes of order fulfilment errors: a multiple case study

ABSTRACT

While recent technological advancements have enabled improved performance in warehouse operations, companies still struggle with eliminating errors such as incorrect/damaged items or the wrong quantity sent to customers. Such errors result in inefficient resource use, costly returns handling, and customer dissatisfaction. Despite the importance of errors, current knowledge of the underlying causes is limited. This paper addresses the gap by investigating error causes through a relatively new technology – intelligent video analysis (IVA), an additional tool for analysing warehouse operations. A multiple case study involving companies that have implemented IVA in their outbound warehouse operations (i.e. order picking, packing, and sorting) was conducted. This study is the first to investigate implementations of IVA, offering novel empirical insights into error root causes. It shows how many errors reported by customers are actually not made by humans in outbound warehouse operations but are attributable to faulty customer claims, inbound warehouse operation errors, and malfunctioning technology. This study offers interesting insights into the interaction of various technical, organisational, and human factors, thereby contributing to the literature on sociotechnical systems and the human-centric Industry 5.0. Finally, the study outlines a way forward for managers to address and further reduce errors occurring in warehouse operations.

KEYWORDS:

• Warehouse
• video analysis
• human factor
• error
• Industry 5.0

1. Introduction

Warehousing is experiencing a technological transformation, and companies are investing massively in a wide range of both information technologies and robotic systems for automated material handling (Azadeh, De Koster, and Roy 2019; Kembro and Norrman 2022). Recent technological advancements have improved warehouse performance in aspects such as storage utilisation, throughput per hour, and cost per pick. However, companies still struggle with eliminating order fulfilment errors perceived to occur in outbound warehouse operations (order picking, packing, and sorting). Order fulfilment errors – that is, customers receiving damaged, incorrect, or missing items (Fontin and Lin 2020) – result in the inefficient use of resources and costly returns handling, including extra costs for labour, material, transport, and administration. These errors may also lead to customer dissatisfaction, potential disruptions for business-to-business customers, and ultimately, the loss of customers (Grosse, Glock, and Neumann 2017; Setayesh et al. 2022).

Despite its importance, current knowledge on underlying causes of errors perceived to occur in outbound warehouse operations is, surprisingly, limited. It could be assumed that implementing a range of information technologies, such as the warehouse management system (WMS; Kembro and Norrman 2022), the Internet of Things (Lee et al. 2018), blockchain (Tao, Wang, and Zhu 2022), pick by voice (PBV; De Vries, De Koster, and Stam 2016), and RFID (Cao, Li, and Wang 2019), would resolve the issue. However, it seems that as technology in warehouses has evolved and companies increasingly rely on check digits, barcodes, RFID, and weight checks, warehouses have effectively become blindfolded. Therefore, the deviations that do happen are hard to understand and address. In fact, while information technologies are effective at collecting data, there is little discussion about whether the integrity of the data is good in the first place or how it impacts the quality of logistics. As noted by researchers such as Setayesh et al. (2022), order picking data could be misrecorded in the first place due to technology malfunction, a poor process, or human error. Such misrecordings could result in ineffective detective work, incorrect feedback to operators about errors they did not make, and the implementation of solutions that do not solve the underlying cause.

Addressing this research gap, the purpose of this paper is to investigate the underlying causes of errors believed to occur during order fulfilment in outbound warehouse operations. We have achieved this by conducting a multiple case study (initiated through a master thesis project: Helm and Malikova 2022) involving six companies that have implemented a relatively novel technology – intelligent video analysis (IVA; Kembro, Danielsson, and Smajli 2017). IVA technology provides a new set of eyes for warehouse management to analyse and understand warehouse operations. This offers a unique opportunity to increase the understanding of what, how, where, and why order fulfilment errors occur.

Contributing to both theory and practice, this study provides empirical insights into a relatively novel technology for analysing warehouse operations and error causes. The study provides evidence on how many of the errors reported by customers are, in fact, not made in the order picking, packing, and shipping processes. The study unveils a range of far-from-obvious errors and gives insights into the interaction of various technological, organisational, and human causes, thereby making an important contribution to the growing literature on warehouses as sociotechnical systems (Panagou, Neumann, and Fruggiero 2023; Winkelhaus, Grosse, and Morana 2021) and Industry 5.0 (De Lombaert et al. 2023). The study also highlights the importance of human factors (Grosse 2023; Grosse, Glock, and Neumann 2017).

The paper is structured as follows: First, we present a review of relevant studies. Second, we outline the method for our multiple case study. In this section, we first introduce IVA and thereafter describe our case selection, data collection, and analysis. In section 4, we present the empirical results and analysis. We then connect our study with the literature review to discuss and elaborate on implications. Finally, we present conclusions and suggest avenues for future research.

2. Literature review

2.1 Order fulfilment in outbound warehouse operations

For this study, it is essential to clarify what is meant by order fulfilment in outbound warehouse operations (see Figure 1). First, order fulfilment generally refers to the completion of customer orders, starting from the moment a customer order is received and ending with delivery of the right product in the right quantity to the right location at the right time (Kembro, Eriksson, and Norrman 2022). This process involves multiple steps inside the warehouse, as well as the physical distribution to the customer’s preferred delivery point (Chen et al. 2022). Second, warehouse operations are typically divided into two broad categories: inbound and outbound operations. While the inbound operations are the receiving, put away, and storing of goods, the outbound operations – which directly concern customer order fulfilment – are order picking, packing, and sorting (per customer, destination, and transporter; Kembro, Eriksson, and Norrman 2022) until the point where the order is shipped out of the warehouse (Bartholdi and Hackman 2016).

Figure 1. Overview of order fulfilment and warehouse operations.

A series of images displaying the flow of goods into a warehouse, after which a person picks an order into a box and sends it off to the customer via a truck.

2.2 Order fulfilment errors: Their consequences and causes

In the order fulfilment process, a customer may experience any of the following order fulfilment errors: receiving (i) incorrect items, (ii) an incorrect number of items (none, too few, too many), or (iii) damaged items (Fontin and Lin 2020; Frazelle 2016; Salhieh and Alswaer 2022). The consequence may be a customer return, which increases the total cost of work in the warehouse (Setayesh et al. 2022). It could also affect the service quality for customers, resulting in dissatisfaction and, potentially, the loss of customers (Zhang et al. 2022). For business-to-business customers, an order fulfilment error could cause a halt in production, resulting in large financial losses (Grosse et al. 2015).

Order fulfilment errors may occur in outbound warehouse operations, especially during order picking. Incorrect handling or stacking of goods on the picking route could result in damaged items (Bartholdi and Hackman 2016; Setayesh et al. 2022). A missing item can be due to it being forgotten to be picked or because it was not on the pick list. A wrong pick can also mean too few or too many of the correct item picked by the operator (i.e. the picker did not verify item integrity; Setayesh et al. 2022). The WMS helps to reduce some of these errors; if the operator has scanned the expected pick location, the SKU, and the destination pallet or box, the risk of an error is theoretically minimal (Winkelhaus, Grosse, and Morana 2021).

Errors in warehouse operations can be ‘detectable’ or ‘propagating’, with the former being easy to identify and correct as soon as it is discovered, while the latter is difficult to recognise and can have lasting consequences (Battini et al. 2015). An example of a propagating error is incorrect inspection and registration of goods in the warehouse receiving operation, which can result in consecutive errors in the outbound operations (Ludwig and Goomas 2007). Kembro and Norrman (2019, 526) add, ‘Errors in registration of product information may create issues throughout the chain’. Another example is if products are put away in the wrong quantity or the wrong location (Frazelle 2016; Setayesh et al. 2022). The actual error may appear as an ‘incorrect order pick’, but in fact, this may have been due to incorrect put away (Ludwig and Goomas 2007).

Our review indicates that the literature predominantly covers the perceived ‘what’ and ‘where’ aspects of errors – with emphasis on the order picking operation. However, the ‘why’ – that is, the underlying causes – has received limited attention. To help identify underlying causes, the literature points to the value of using metrics such as picking, put-away, shipping, and inventory accuracy, as well as error-free orders shipped (De Koster and Balk 2008; Frazelle 2016; Staudt et al. 2015). However, how to measure (or detect) these errors in practice is not described, and considering the propagating errors, the metrics do not necessarily reflect the actual cause of the errors. An error may have occurred somewhere downstream or upstream of the operation where the error was detected and logged. Hence, an operation-specific metric may not accurately point out the root cause of an error (Ludwig and Goomas 2007).

One of the few studies that have investigated underlying causes is Škerlič and Muha (2017). Based on a literature review, they concluded that operators are often blamed for errors, but ‘human error is a symptom of deeper problems within the system because the source of errors is structural in nature, rather than human’ (Škerlič and Muha 2017, 86). Therefore, it is relevant to also consider technological and organisational aspects. In a recent study, Setayesh et al. (2023) developed a tool that attempts to identify errors and underlying causes in warehouse operations through a series of detailed questions posed to the warehouse operations staff. The tool highlights layout, pick information and technology, storage assignment, palletising and batching, routing, and organisational behaviour management as areas of improvement.

There are also specific examples of error causes; one issue mentioned by Bartholdi and Hackman (2016) is that storing similar-looking products next to each other may cause incorrect picks. Another issue is that multitasking in the form of batch picking may lead to mixing orders (Frazelle 2016). Setayesh et al. (2022) mention random storage strategies resulting in cognitive confusion. Additional causes include unclear labelling of locations and goods and inexperience with technology. Technology can be the cause of errors, as it introduces additional complexity (Fontin and Lin 2020). Ludwig and Goomas (2007) exemplify one such case where two signs, marking pick locations, had been mixed up. Consequently, the operators using PBV ‘blindly’ picked from the instructed location without a description of the product available.

2.3 Human factors in warehouse operations

While automation technology is increasingly used in warehouses (Kembro and Norrman 2023), manual picking still accounts for the largest part of warehouse operation costs worldwide (Zhang et al. 2022). Thus, warehouses increasingly represent the so-called sociotechnical systems, where employees, technical systems, and the organisational environment interact and affect each other (Dul et al. 2012; Winkelhaus, Grosse, and Morana 2021).

Human factors, representing the interactions in sociotechnical systems, focus on two related outcomes, performance and well-being, and are thus important for understanding error causes (Dul et al. 2012; Grosse, Glock, and Neumann 2017). Nonetheless, human factors have only recently received increased, though still limited, attention in the warehousing literature and in research on Industry 4.0 (Grosse et al. 2015; Kolus, Wells, and Neumann 2018; Neumann et al. 2021; Panagou, Neumann, and Fruggiero 2023; Sgarbossa et al. 2020). This especially applies considering the shift toward Industry 5.0, which places human well-being at the centre of warehouse operations (Grosse 2023; De Lombaert et al. 2023; Panagou, Neumann, and Fruggiero 2023). A fundamental assumption is that humans have capabilities, limitations, and needs, which all have a direct impact on operational quality and specifically errors (e.g. picking the wrong item; Grosse, Glock, and Jaber 2013; Kolus, Wells, and Neumann 2018). Neumann et al. (2021, 7) elaborate,

If [there is a] mismatch between worker capabilities and the demands placed on them by the system, then dysfunctional results, including errors and injuries, can be expected. This leads to a chain of negative consequences for both the worker and ultimately for the system as a whole.

Grosse et al. (2015), Grosse, Glock, and Neumann (2017), and Vijayakumar et al. (2022) list five human factors that impact operational performance: perceptual (e.g. information processing, picklist reading, confusion), mental/cognitive (e.g. experience, learning by doing), physical (e.g. travel, carrying items), psychosocial (e.g. motivation, stress, boredom), and work environment (e.g. light level, noise, cold, heat). The literature lists various examples for each of these human factors. Grosse, Glock, and Neumann (2017), for example, discuss that operators processing a large amount of information in a short time may induce errors. Meanwhile, new supportive technologies, such as PBV, goods-to-picker stations, and real-time feedback systems, may help to reduce errors by alleviating the mental workload (De Lombaert et al. 2023; De Vries, De Koster, and Stam 2016; Zhang et al. 2022). Human learning is also faster during automated picking processes due to frequent and real-time feedback with multiple perception–cognition–motor–action cycles (Loske 2022). Grosse, Glock, and Jaber (2013) conclude that experience and familiarity reduce the percentage of pick errors.

Regarding psychosocial factors, Kolus, Wells, and Neumann (2018) and Vanheusden et al. (2022) discuss workload, which includes physical and mental stressors, and fatigue, which results from the increased stress. Grosse et al. (2015) mention the so-called slips, where actions are in disaccord with the actor’s intention. This could happen when an operator accidentally picks the wrong item or drops an item due to distraction. De Lombaert et al. (2023) identify that conversing with colleagues has an impact on concentration and leads to errors, though conversely also results in increased motivation. Furthermore, Grosse, Glock, and Neumann (2017) suggest that managing motivation, stress, and boredom are key for reducing errors. One example is gamification, which has been shown to help increase productivity and notify operators of errors (Winkelhaus, Grosse, and Morana 2021).

2.4 Summary of the literature review

Order fulfilment errors persist despite technological advancements. One challenge is that some errors are difficult to detect, and the underlying causes have received limited attention. Recent research has emphasised the importance of human factors for reducing errors. However, there is still insufficient knowledge of the potentials and obstacles of using supportive and substitutive technologies in manual order picking (Grosse 2023). This is where our study and IVA technology can make an important contribution. As IVA provides a new set of eyes for warehouse management to analyse and understand what goes on in a warehouse, our study aims to propel research on human factors and the root causes of order fulfilment errors into the next phase with unprecedented insights that were previously unavailable.

3. Methodology

This paper explores a phenomenon – the underlying causes of order fulfilment errors believed to occur in outbound warehouse operations – through a case study involving six companies that have implemented IVA. In this section, we first introduce and explain IVA technology. Thereafter, we present the case selection and description of the case companies, including how they use IVA. Finally, we detail the data collection and analysis processes.

3.1 Introduction to IVA technology

IVA is a tool that enables visually recording, tracking, and analysing order and goods flow where errors are perceived to occur. The first IVA tests were conducted in 2015/2016, and Kembro, Danielsson, and Smajli (2017) discussed the potential value of IVA for improving warehouse operations. We have found no other discussions of IVA implementations in warehouse operations when searching for key literature terms. One reason is that, to the best of our knowledge, the first implementation of IVA occurred in 2018.

In brief, as illustrated in Figure 2, the IVA applications investigated in this study include a set of network cameras mounted at various positions in the warehouse. The cameras are used to record key operations and locations, such as picking locations, forklifts, packing tables, and consolidation areas. The video sequences are searchable via IVA software by using transaction data (e.g. time stamps, locations, SKU, order number, and equipment) captured through a combination of automatic identification (e.g. barcode and RFID; Kembro, Danielsson, and Smajli 2017) and automatic communication technologies (e.g. PBV, pick-by-light systems, and vision picking; Glock et al. 2021), all connected to the WMS keeping track of what, how much, when, and from/to where items are moved (Kembro and Norrman 2022).

Figure 2. Illustration of how transactional data from information systems is matched with video of physical goods flow to enable the video analyst to search for and analyse recorded sequences.

A series of images that show how an order picker is being recorded via multiple video cameras. Another person is sitting by the computer analysing the recorded video material.

It is important to note that the studied implementations of IVA involve human analysts for the identification and classification of errors and analysis of root causes. This means that as a deviation (e.g. a customer complaint regarding missing item) is reported, human analysts use the IVA software to search for and analyse the relevant recorded video sequences. The next generation of IVA with artificial intelligence had limited adoption at the time of the study and is therefore not included in this paper but will be further discussed in section 5.

3.2 Case selection and description

Case study research is ideal for investigating and better understanding emerging or contemporary phenomena (Eisenhardt and Graebner 2007). Case studies offer rich empirical insights that allow for analytical rather than statistical generalisation (Yin 2014). Analytical generalisation is based on comparing a previously developed theory with empirical insights from typically 4–10 cases. To guide the reader, the criteria for case selection are explicit, and the case contexts are described with many details (Gibbert, Ruigrok, and Wicki 2008; Johnson 1997).

Targeting information-rich cases, we reached out to a total of 14 companies. Access to these companies was facilitated through a supplier of IVA technology. We narrowed the list to six case companies based on purposeful sampling (Patton 2015), considering (i) IVA use experience, (ii) data availability and richness, and (iii) cases representing similar and contrasting contexts. The selected companies (see Table 1) had at least four months of experience with IVA at the time of the data collection. To the best of our knowledge, these companies are among the first adopters of IVA. All case companies have used IVA to address different types of errors in their warehouse operations. Five cases are based in Sweden, and one is based in Belgium.

Table 1. Overview of case company characteristics.

Characteristic	Company A	Company B	Company C	Company D	Company E	Company F
WH location	Sweden	Sweden	Sweden	Sweden	Sweden	Belgium
WH type	Distribution	Distribution	Contracted distribution	Contracted distribution	Distribution	Distribution
Customers	Other WHs and direct customers	Direct customers	Stores and direct customers	Stores	Stores and direct customers	Other WHs and direct customers
Goods ownership	Self	Self	Clients	Client	Self	Self
Type of goods	Spare part	Pesticides & feed for agriculture	Electronics and accessories	Hardware and tools	Car parts and accessories	Spare part
Size of goods	Small to very large	Small to large	Small	Small to large	Small to very large	Small to very large
Value of goods	High	Low to High	High	Low to High	Low to High	High
Operations with IVA	Picking	Picking	Packing, Consolidation	Picking	Picking, Consolidation	Picking, Consolidation
Experience with IVA	6 months	1 year	2 year 6 months	2 years	1 year 3 months	7 months
IVA configuration	Stationary	Mobile	Stationary	Mobile	Stationary	Stationary, Mobile
Technologies used	Barcode, AS/RS, goods-to-picker station, WMS	Barcode, WMS	Barcode, goods-to-picker station, WMS, TMS	Barcode, forklift screen, WMS	Barcode, goods-to-picker station, WMS	Barcode, RFID, pick-by-voice, WMS

We observed two main IVA configurations: stationary and mobile. Stationary cameras were used by four case companies (A, C, E, and F), providing a focused view of individual outbound warehouse operations (order picking, packing, and sorting). Mobile cameras, used by three companies (B, D, F), were mounted on forklifts and allowed recording across multiple operations. The case companies used IVA as a complement to other existing technologies, such as barcode technology, PBV, forklift computers, AS/RS, and goods-to-picker stations. Two companies implemented IVA because the quality improvements from the goods-to-picker stations (E) and the PVB (F) had plateaued relative to the expected improvements from these technologies, while the rest wanted to improve outbound warehouse operations generally.

3.3 Data collection

The case study method allows the use of multiple data collection methods (Yin 2014); for this study, we used a combination of qualitative semi-structured interviews, observations, and archival data (Table 2). Together, these methods allowed us to triangulate and strengthen the validity of the study findings (Patton 2015). Furthermore, semi-structured interviews give flexibility to the process by not only collecting data on a predetermined set of topics but also providing room for exploration (Kallio et al. 2016). The data was collected through a master thesis project (Helm and Malikova 2022). To avoid common flaws and bias, the senior researcher (supervisor of the thesis) was involved in all planning and preparation steps, including development of the structured data collection protocol with interview questions (see Appendix A).

Table 2. Summary of data collected from the case companies.

Company	Observation	Archival Data	Interview	Interview detail
				Date	Duration	Role (years of experience)
A	No	No	Yes	3/3/22	61 min	WH operations manager (20), Outbound team manager (8)
B	No	Yes	Yes	3/9/22	55 min	Transport manager (22)
				3/28/22	25 min	Customer service team (N/A)
C	Yes	Yes	Yes	3/14/22	56 min	Security and business development manager (15)
D	Yes	Yes	Yes	3/15/22	42 min	Operating manager (16)
				3/15/22	63 min	Production manager (10)
				4/19/22	30 min	Customer service representative (N/A)
E	Yes	Yes	Yes	3/16/22	87 min	Production manager (12)
F	Yes	No	Yes	3/16/22	52 min	Production System Specialist (17)
				3/18/22	29 min	Production System Specialist (17)

First, the interviews allowed rich data collection with a diverse set of expressions from the respondents (Packer 2011). We conducted interviews with key staff (Table 2); that is, we interviewed staff who had access to the IVA system and possessed the most insights into errors and underlying causes. Thus, we interviewed a different number of staff and different roles for each of the case companies. The semi-structured interview topics focused on the reason for implementation, IVA configuration/set up, IVA uses, and observed effects. The interviews were conducted either on-site or, due to COVID restrictions, via an online platform (Teams). All interviews were recorded and transcribed. Second, we conducted observations that provided in-depth descriptions of activities in a real-life setting (Yin 2014). In four of the cases, we were allowed to observe instances where employees interacted with IVA: video observations of performed errors at three companies and images of errors at one company. These observations offered advantages in terms of objectivity compared to interviews, as the phenomenon can be seen directly without the influence of an interviewee’s feelings or attitude (Robson 2002). The observations were also discussed with the case representatives, allowing for follow-up questions to understand the root causes of and ways to resolve the errors. Third, we received archival WMS data from four out of the six companies. The archival records included performance data related to errors in warehouse operations. This data supported both observations and interviews to substantiate claims of identified and remedied errors by the case companies.

3.4 Data analysis

Our analysis, which was conducted jointly by the researchers, followed the three steps proposed by Miles, Huberman, and Saldaña (2014): data condensation, data display, and conclusion drawing and verification. Data condensation began with creating within-case summaries of IVA use and its effects. An important part was coding – that is, assigning labels to sections of the interview transcripts to categorise the content in a way that is relevant to the research question (Robson 2002). To improve the validity, the coding was conducted independently between the researchers and thereafter compared to establish consensus. Examples of codes included error types, causes, actions, and effects. Next, we created data displays in the form of matrices, charts, and networks to visualise ‘what the data is saying’, which in turn reduced the risk of hasty conclusions (Miles, Huberman, and Saldaña 2014). Causal maps were created for each case company with links between the error types, insights, actions, and effects (Figure 3).

Figure 3. Example of causal map linking error types, causes, preventative actions, and observed effects of using intelligent video analysis.

A total of 12 text boxes arranged in four levels, three boxes per level. Between the different levels, the boxes are connected to each other via multiple many-to-one and one-to-many relationship lines. The connections show how error types are connected to error causes, which require certain actions to be resolved.

The case summaries and causal maps were sent to the interviewees for validation. Next, we analysed differences and similarities between the cases (Patton 2015) concerning the uses, actions, and effects of IVA. This analysis allowed us to create a final causal map and specifically understand how IVA is used to identify errors that the warehouse did not make and how it enables follow-up actions to address actual errors (Miles, Huberman, and Saldaña 2014). Furthermore, the cross-case analysis investigated how the various physical configurations contribute to the effectiveness of IVA in identifying and acting on errors.

We validated our findings with the interviewees. We also supported our findings with context and rich case descriptions and used triangulation to verify within-case findings (Robson 2002). To ensure external validity, we provide an overview of the case company characteristics. Lastly, we ensured practically relevant focus and results (Miles, Huberman, and Saldaña 2014) by working with both a facilitating partner and companies using this technology.

4. Results and analysis

In this section, we present the various errors observed through IVA, followed by an analysis of the underlying causes and the different strategies to address errors.

4.1 Distinguishing errors made inside versus outside of the warehouse

Capturing an error (e.g. an incorrect item picked) where it occurs is difficult using conventional methods, as it would require every order to be frequently counted, weighed, or checked for quality. Instead, as displayed in Figure 4, the case companies previously relied primarily on two data sources for measuring errors: customer claims and inventory discrepancies. Relying on customer claims is problematic because these are reported after the transportation and the customer receipt process, both taking place outside the warehouse’s control. Although inventory discrepancies capture errors in the warehouse, frequent stock taking is time-consuming and could include errors made by the supplier – for example, incorrect labelling. Meanwhile, both methods of data capture rely on someone both capturing and reporting the error. If an error is not captured and reported, it is not included in the metrics. Thus, some errors are likely uncaptured and/or unreported.

Figure 4. Observed relationship between error sources, how they are measured, and resulting error metrics.

On the top, a big circle is included reading ‘error metrics’. Below, multiple text boxes are connected via solid or dotted lines. The text boxes represent captured error data and the sources of errors. The solid lines show that the error was caused inside the warehouse; the dotted lines indicate that the error was caused outside of the warehouse.

Although unreported errors cannot be easily identified as they are made, IVA can at least help with investigating the causes of those errors reported through customer claims or inventory discrepancies. Even more importantly, IVA can be used to validate the error – in our study, IVA recordings showed that multiple order fulfilment errors previously attributed to the warehouse were made outside of the warehouse, shown as dotted lines in Figure 4.

4.2 A means for addressing errors made outside of the warehouse

Without IVA, the case companies previously had limited means to confirm or invalidate customer claims. They were therefore forced to accept a customer claim as an error made in the warehouse, despite suspicions that the error occurred after the order left the warehouse. With IVA, the case companies now have recorded evidence by tracing the order fulfilment process and ensuring that the order left the warehouse as intended (correct items, correct quantities, undamaged). This enables them to refute claims based on errors made outside of the warehouse. Thanks to IVA, the case companies could identify a range of actions that could be taken externally with their customers, transport providers, and suppliers. An overview of these benefits and actions is provided in Table 3.

Table 3. Overview of benefits with intelligent video analysis for dealing with fulfilment errors made outside of the warehouse.

Benefits with IVA	A	B	C	D	E	F
Able to conclude errors were made outside the warehouse by viewing video sequences of correctly picked, packed, and shipped orders	✓	✓	✓	✓	✓	✓
Able to deny customer claims that originated from errors made outside the warehouse		✓		✓	✓
Able to act toward transporter for errors made outside the warehouse (missing or damaged item)			✓	✓		✓
Received reimbursement from transporter			✓	✓
Avoided costly RFID investment to resolve errors in consolidation, when in fact the issue was rooted elsewhere						✓
Able to identify errors made outside the warehouse at supplier site, leading to planned joint improvement project with suppliers			✓	✓

Exemplifying IVA benefits, company B concluded that out of the 70 claims they investigated at the time of the study, 32.5% were errors made outside of the warehouse. The error sources were instead found to be misunderstandings or lacking routines in the customers’ goods reception operation, or errors occurring during transport. The interviewee commented,

I would say that in the cases where we sent the complete order, the customer typically misses it in their receiving operation. The pallets we send contain multiple SKUs, and sometimes it can be hard to see the items located at the bottom or in a box with multiple SKUs.

Similarly, company E concluded that about 40% of their customer claims were not errors made inside the warehouse. The videos showed how the orders were accurately picked and packed before leaving the warehouse. The claims could thus be denied and were also excluded from company E’s warehouse performance metrics. The impact of IVA on company E’s quality KPI is shown in Figure 5. Implementing IVA early 2021, accepted claims significantly decreased and remained low throughout the year. As a result, the company could relocate three full-time employees previously occupied with handling claims and errors.

Figure 5. Accepted claims in relation to total order output at company E.

A graph covering the time period 2020 to 2022 with a plotted line indicating the percentage of accepted claims of total orders. The line is gradually falling until halfway where it levels out.

Company D also confirmed that order fulfilment errors had reduced since their implementation of IVA. Investigation of customer claims and pointing to where the customer may be wrong has helped shift the responsibility of identifying the cause of a claim from the warehouse to the customer. The interviewee elaborated,

Nine out of 10 times, the customer either finds the item, or accepts that they might have lost it, when we show video of the claimed item being picked and packed. One out of 10 times, the customer insists that they never received the item, regardless of what the video shows.

Three companies (C, D, and F) used IVA recordings to take action against transporters for causing errors. Companies C and D were reimbursed for lost or damaged goods by the transport provider. Company A plans to take similar action but has not done so yet due to concerns with the customer relationship.

Company C was able to identify recurring damages for a specific parcel type leading to a joint improvement project with their supplier. Company D discussed future work with a supplier to change the barcode on the packaging of a certain SKU, with the aim of reducing confusion during picking. Finally, company F avoided a costly RFID investment thanks to IVA. They had a perceived issue in their consolidation area (i.e. sorting of orders per customer, destination, and transporter before shipping), which was planned to be addressed through RFID implementation. However, as they analysed IVA recordings, few of the reported errors were actually observed in the consolidation area, and the root causes would not have been addressed properly with RFID.

4.3 Identifying errors and underlying causes inside the warehouse

In parallel with identifying errors made outside of the warehouse, IVA proved to be instrumental for identifying errors that actually occurred in outbound warehouse operations and for linking these to the underlying causes. All case companies, except company B, have benefited from this IVA feature; due to union rules, company B is forbidden to record videos for investigating potential employee mistakes. Figure 6 shows the observed errors and underlying causes based on cross-case analysis of video footage and supporting interviews.

Figure 6. Links between errors made in outbound warehouse operations and the underlying causes.

Multiple text boxes arranged in three levels, three boxes in the first and third level, and seven boxes in the second level. Between the different levels, the boxes are connected to each other via multiple many-to-one and one-to-many relationship lines. The connections show how error types are connected to perceived error causes based on analysis of video recordings. The error causes are grouped based on the origin of error.

First, the technical problems were easily detectable with IVA. Company E, for example, identified two malfunctions in their goods-to-picker stations. At seemingly random times, the information screen that provides the operator with instructions did not update for a few seconds. Therefore, the operators acted on old information and systematically picked the wrong quantity and SKUs for the next order. On further investigation, it was shown that the malfunction was caused by a memory leak in the screen itself. At other times, the mechanical button used to confirm a pick line acted erroneously. By investigating the claims, it was demonstrated how the pick lines, at times, were confirmed without the operator pressing the button. Preliminary analyses indicated that this malfunction was due to a worn button that reacted to vibrations from the products. Common for both errors are that, prior to IVA, these errors were assumed to be made by the operator likely due to confusion or distraction.

Next, we identified five causes that relate to human factors: distraction, confusion, inexperience, stress, and carelessness. Distraction is illustrated by two examples. At company A, picking errors, such as putting items into wrong pallets, were made by operators talking to colleagues or playing with their phones. At company E, operators went on break in the middle of an order pick, and then the next operator picked the same item again. Company E also observed employees distracted by phone use. Another issue is stress. At company C, IVA showed how operators tried to make up for a backlog of pick orders by multitasking, resulting in the picking of the wrong items and quantities. Similar mistakes were made due to inexperienced staff who did not follow established processes. Likewise, confusion played a role, as operators at companies D and E picked whole boxes instead of pieces or frequently forgot or did not understand that multiple parts were required for one SKU. Company D also observed careless behaviour while stacking items, resulting in damaged items.

Finally, organisational issues explained several of the errors in outbound warehouse operations. As observed by companies D and E, issues included unclear processes, poorly labelled items, and a lack of information available to the operators. Other issues were a lack of training for new employees and inadequate picking processes – where fragile items were picked at the beginning of the pick route, resulting in damaged items. Also, stressed operators, due to backlogs, could be linked to organisational aspects such as understaffing and stressful performance measuring and monitoring.

Our analysis showed that an error could be due to multiple causes; therefore, we observed that technological and human factors and organisation error causes are related. For example, an incorrect pick could be caused by a combination of flawed process and an inexperienced operator. Company F implemented PBV in their forklift picking – where each forklift could fit four small pallets, each for a different customer. While PBV initially increased the quality of picking, using voice to confirm which pallet to pick led to frequent confusion and mistakes. The problem observed had multiple roots. Despite new technology, operator confusion was due to a lack of training or incorrect labelling (organisational issues). This issue will be further addressed in section 5.

4.4 Addressing the error causes in outbound warehouse operations

For technical causes, the solution for company E was straightforward, in the sense that once the problem was identified, it was typically forwarded to the automation technology supplier or the internal IT department. Similarly, organisational causes were perceived and handled in a straightforward way; most case companies (C, D, E, and F) focused on revising their processes and policies, such as clarified and visualised information available for operators, updated policies for new employees, and changed pick routes so that fragile items are picked last. Some of the issues, such as packing and labelling, indicated that the root cause originated upstream in the supply chain, leading to the initiation of joint development projects with suppliers.

While the approaches to addressing technical and organisational aspects were similar across all cases, human factor–related causes were dealt with in different ways. All case companies engaged in some sort of feedback to the operators about errors; however, the scope of the feedback varied. Our analysis indicated five different levels (Table 4).

Table 4. Summary of different levels of using intelligent video analysis in feedback to employees based on our empirical findings.

	Case company
Level of action	B	F	C	E	A and D
1. Feedback to the group of operators	✓	✓	✓	✓	✓
2. Discussion on improvements with the group		✓	✓	✓	✓
3. Feedback to individual operators			✓	✓	✓
4. Video presentation to individual operators				✓	✓
5. Input for performance evaluation					✓

At level 1, the company informs the group of employees of the total number of errors – this was done at all case companies. At level 2, feedback was also provided to the group as a whole, involving the employees in a discussion of ideas how common errors could be resolved. Companies with level 2 action (all except company B) showed that this step is critical for effectively identifying corrective actions. At level 3, feedback was additionally given to individual operators (companies A, C, D, and E). Level 4 includes showing video sequences of the committed error to the employees. Companies A, D, and E confirmed that showing the video allows for effective feedback to employees, as seeing themselves making a mistake facilitates an understanding of what they have done wrong. Finally, at level 5, IVA was also used as a basis for individual performance evaluation (companies A and D). This could, for example, be a written warning or bonus deduction measured based on number of errors committed.

Importantly, the level (1–5) of feedback to employees is not just a managerial choice per company. Union negotiations and country laws were both mentioned by case companies as measures that limit how they could use IVA to a varying extent. For example, company B’s access to IVA was limited to the customer service department. Furthermore, the cameras were configured to show as little as possible of the operators working, which makes root cause analyses difficult. As a result, they could practically only use IVA to validate claims and report accurate error statistics to the warehouse. Similarly, company F reported that, legally, they were only allowed to use recorded video material for process improvements but not for individual feedback. However, union negotiations at company F allowed for IVA to be accessed by quality engineers, in addition to customer service, which allowed improvements to be discussed on a group level.

In addition to these deliberate feedback actions from companies, four companies (A, B, D, and F) observed that operators were more focused just by being aware of the network cameras. As a result, the operators made fewer mistakes. Company F provided an interesting observation regarding their IVA, which was mounted on one of their forklifts: with no targeted feedback or action other than mounting the IVA on one forklift – and clearly the operators understood which forklift was being monitored – the claims related to orders processed using that forklift were 80% lower than when using an average forklift in the warehouse. Furthermore, the operators rotated between forklifts, so it was concluded that this reduction was not connected to a specific operator.

5. Discussion and implications

Our results and analysis bring several important implications for theory and practice. Below, we connect our insights from the multiple case study with the literature review to discuss the most important findings.

5.1 A new perspective on order fulfilment errors

First, identifying errors is important for improving performance in terms of reduced costs and returns handling. This is especially true when the warehouse is not responsible for the error and has yet to pay for it in some way. The costs of completing an investigation, picking a new item, and shipping it out, in addition to the product costs, is quite expensive (Setayesh et al. 2022; Zhang et al. 2022). IVA also contributes to collaboration on improvement initiatives with other actors (such as transporters and suppliers), which could help reduce overall errors. This measure is especially useful for addressing hard-to-identify propagating errors (Battini et al. 2015), resulting in a positive impact across the supply chain (Kembro and Norrman 2019).

Second, companies need to rethink how to measure order fulfilment errors in outbound warehouse operations. The literature does not provide clear guidance (see, e.g. Frazelle 2016; Staudt et al. 2015), and as seen from our case study, companies primarily use customer claims as a source, which is far from accurate. Thanks to IVA, we have demonstrated that a significant number of claims are due to errors made outside of the warehouse. Warehouse metrics that include errors made outside of the warehouse may mask the actual problems and, to address the underlying causes, point the company in the wrong direction. This could have long-lasting negative impact on human factors and warehouse performance. For example, falsely accusing warehouse operators of errors may lead to stress, distraction, and demotivation, negatively affecting performance (Grosse, Glock, and Neumann 2017) while the underlying error cause remains unaddressed.

5.2 Human factors: Addressing the cause of many errors

Human factors receive increased attention in the warehousing literature (Grosse et al. 2015; Neumann et al. 2021) and, as shown in our study, are key for addressing order fulfilment errors made in warehouse operations. Among the case companies, human errors were by far the most common causes of errors. We identified multiple human factors, of which perceptual (confusion), cognitive (experience), and psychosocial (stress, distraction, and carelessness) seemed to dominate (cf. Grosse, Glock, and Neumann 2017). In fact, we did not observe any direct physical or work environment (light, noise, temperature) causes. This is interesting and should be explored further. A possible explanation is that companies – while neglecting mental and psychosocial aspects – to a large extent consider the physical work environment and ergonomics in warehouse and automation design (Kembro and Norrman 2023).

The literature presents limited discussion on corrective actions to address human factors. To the best of our knowledge, our study is the first to discuss actions for reducing order fulfilment errors made by human factors. We found that the main way the case companies have addressed errors made by human factors is to provide feedback on different levels. The most effective actions are the ability of the managers to discuss specific errors with the staff to jointly identify solutions (even without singling out anyone). Such feedback provides a learning opportunity (Loske 2022), which has a direct positive impact on mental/cognitive aspects and results in improved performance. Staff involvement also boosts psychosocial aspects such as increased motivation, which could help explain the positive impact on errors (Grosse, Glock, and Neumann 2017).

At the same time, our findings suggest that errors made by underlying human factors may be quite difficult to reduce with corrective actions. For example, as discussed during our interviews, it is hard to prevent an operator from being distracted by a personal matter and miscounting a pick. In our study, no warehouse technology was able to eliminate this type of error. Automated picking stations and PBV were implemented and resulted in initial improvements (cf. De Lombaert et al. 2023; De Vries, De Koster, and Stam 2016; Zhang et al. 2022); however, the new technology benefits at our case companies plateaued, indicating that these technologies have limitations in how much they can reduce errors. From a sociotechnical perspective, it may be that the technologies are not designed with human factors in mind (Grosse et al. 2015; Kolus, Wells, and Neumann 2018; Neumann et al. 2021; Panagou, Neumann, and Fruggiero 2023; Sgarbossa et al. 2020) or may not be suited for addressing all types of human errors.

5.3 The underlying technological and organisational causes

It is well-documented that technology helps improve operations, but as discussed by Fontin and Lin (2020) and supported in our study, technology also introduces new complexities and thus new potential error sources. There is a risk of operators becoming too reliant on the technology, almost blindly following and overtrusting the machine (Loske 2022; Ludwig and Goomas 2007).

IVA provides a useful tool to quickly identify technical issues. This was identified as valuable for increasing the operators’ motivation and confidence in the technology, as well as for resolving the issue at hand with the help of the IT department and/or the technology supplier. Furthermore, in line with Loske (2022), IVA in combination with other picking technologies could facilitate frequent feedback loops that accelerate human learning. This provides valuable training and experience – which, according to De Vries, De Koster, and Stam (2016), are key for increasing productivity while decreasing errors.

The importance of learning connects to underlying organisational issues. Our study shows that a lack of training leads to operator inexperience and reduced cognitive performance, which in turn could result in errors and have a negative impact on work performance. Other examples of organisational issues include unclear processes, a lack of information, and poor labelling, which all impact operator perceptual factors. These examples point to the fact that human factors are closely linked with organisational factors to get the full picture of underlying error causes (Škerlič and Muha 2017). Being able to identify an error cause from observations, instead of anecdotes, may enable a company to address an issue faster and more accurately. The ability to identify underlying causes, as opposed to blaming the error on the operator, can help increase motivation and both directly and indirectly – through positive impact on human factors – enhance their operational performance (Grosse, Glock, and Neumann 2017).

5.4 Errors in sociotechnical systems: A combination of multiple underlying causes

While IVA helps to identify errors made within the warehouse, drawing accurate conclusions on the underlying causes and remedies is not straightforward. As many warehouses represent sociotechnical systems, the underlying causes may be a combination of technological, organisational, and human factors (Dul et al. 2012; Winkelhaus, Grosse, and Morana 2021). Not uncommonly, as our study corroborates, operators are blamed for errors, while the errors may be symptoms of deeper structural problems (Kolus, Wells, and Neumann 2018; Škerlič and Muha 2017).

Our analysis indicates that errors can be due to multiple causes, and the observed causes could, in line with the human factor literature (Grosse, Glock, and Jaber 2013; Neumann et al. 2021), have an additional layer of root causes. To give an illustrative example, picking the wrong item may appear to be caused by a confused operator. This confusion could come from a technical problem, unclear labelling, and/or a distracted operator. In turn, this distraction could originate from multiple psychosocial factors, such as a lack of motivation, boredom, or stress (Grosse, Glock, and Neumann 2017). Boredom may be caused by monotonous tasks in an automated environment. Meanwhile, stress could have various underlying causes – for example, poor management, an inconducive work environment, or an operator trying to make up for a backlog. For a company to take the right action, this example shows the necessity of broadly investigating the underlying root causes. IVA has been shown to provide important clues pointing an investigation in the appropriate direction. Without IVA, it would be quite difficult to confirm and address underlying error causes.

5.5 Legal and ethical aspects of using IVA in manual warehouse operations

Our findings indicate several positive aspects of using IVA in warehouse operations. If used appropriately, it reduces the risk of falsely putting the blame on warehouse operators. From a human factor perspective (Grosse et al. 2015; Grosse, Glock, and Neumann 2017), operators may therefore perceive IVA positively, as it reduces the stress of being falsely accused of fulfilment errors. With that said, the positive take on IVA must be discussed vis-à-vis the fact that the staff is being ‘watched’ on a regular basis.

An important aspect of using IVA with corrective actions is country laws and regulations, and unions. While Sweden and Belgium have strong privacy laws due to the General Data Protection Regulation, our study indicates that unions are a more significant factor in how IVA can be used to address errors within the warehouse. However, this finding reflects six companies that have implemented IVA, so future research can identify whether this can be generalised.

Likewise, our findings indicate that the mere presence of cameras (or the perception of being monitored) leads to operators making fewer errors. This is relevant to discuss from a human factor perspective, as the perception of being monitored (by management) could have implications for a range of psychosocial aspects (e.g. motivation and stress) and the well-being of operators (Grosse et al. 2015; Grosse, Glock, and Neumann 2017; Vijayakumar et al. 2022). This is key for Industry 5.0, which places human well-being at the centre of warehouse operations (De Lombaert et al. 2023; Grosse 2023; Panagou, Neumann, and Fruggiero 2023). Our research did not specifically study the impact of IVA on operators; therefore, this should be studied in more detail. For example, the finding that operators were more focused on IVA cameras could be due to negative stress and other unknown factors, as the cameras may be considered intrusive. Hence, the value of IVA technology could be questioned if it negatively impacts the quality of the operator’s work environment.

5.6 Can order fulfilment errors ever be eliminated?

Summing up all insights from our study, it is interesting to discuss if and how fulfilment errors can be (fully) eliminated. A conceptual illustration of this idea is presented in Figure 7.

Figure 7. Conceptual illustration of how intelligent video analysis can impact a warehouse performance metric for fulfilment errors.

A graph showing a solid line that is gradually declining, after which it becomes dotted and levels out. The line represents number of errors in three stages. First, without using video recording. Second, when video analysis is used. The final stage, which is dotted, indicates a potential future using artificial intelligence, questioning if errors can be completely eliminated.

It is clear from our case companies that IVA helps to identify and filter out errors, using warehouse metrics, made outside of the warehouse to a significant extent. IVA is also useful for quickly addressing errors rooted in technical and organisational issues. Moving forward, improved technologies and capabilities for preventing errors made in the warehouse and analysing their causes should help to further reduce errors over time. However, corrective actions take time to investigate, develop, and implement. Another challenge is that new product introductions, process changes, and new employees continuously create new errors. Similarly, for technological underlying causes, hardware wear and software updates can also introduce new errors that must be analysed before they are dealt with. As such, we argue that there is a limit to the extent that order fulfilment errors can be eliminated with the current implementation of IVA (that is, retroactively analyse video sequences to suggest improvements). It is also important to highlight that many errors are never reported or detected. Based on our empirical insights, this particularly concerns ‘overpicks’, including picking boxes instead of pieces. Because these errors are difficult to identify, using IVA for retroactive analysis may not be able to resolve them. Altogether, and supported by the case companies, these insights indicate that IVA using artificial intelligence (real-time IVA) for proactive analysis is the next step. Real-time IVA includes real-time monitoring and corrective actions, where the operator receives immediate feedback from the IVA to recognise an error as it happens. This would also have implications for the possibilities and motives of analysing underlying causes by reducing the need to investigate. Errors made by distraction and recklessness, for example, could potentially be reduced if they are prevented in real time compared to the time and resources needed to analyse and target a corrective action for the root cause, which may not even be that effective.

6. Conclusions and future research

The purpose of this study was to investigate the underlying causes of fulfilment errors believed to occur in warehouse operations. Surprisingly, this topic has received limited attention despite its impact on customer satisfaction, the cost of returns handling, and the efficient use of resources (Grosse, Glock, and Neumann 2017; Setayesh et al. 2022).

We conducted a multiple case study of companies that have implemented IVA, a relatively novel technology that provides a new set of eyes for warehouse management to analyse and understand warehouse operations (Kembro, Danielsson, and Smajli 2017). The technology revealed that many errors thought to be caused in the warehouse are often made outside the warehouse on the supplier, transport, or customer side. However, errors caused in the warehouse are far from easy to prevent and resolve – especially with increasing automation, which constantly introduces new types of errors.

6.1 Contribution to theory and practice

Our study has multiple contributions. First, it adds to the literature stream on warehouse metrics (e.g. De Koster and Balk 2008; Frazelle 2016; Staudt et al. 2015) by concluding that customer claims overstate errors made in the warehouse. This metric thus puts unmotivated blame on warehouse operations, hides many of the actual underlying causes, and may point improvement initiatives in the wrong direction. We also provide an important contribution to the literature on new technology in warehousing, including smart warehouses, Industry 5.0 (De Lombaert et al. 2023), and sociotechnical systems (Kembro and Norrman 2022; Neumann et al. 2021; Panagou, Neumann, and Fruggiero 2023; Winkelhaus, Grosse, and Morana 2021). We demonstrated that IVA can support a warehouse’s transition to automation by providing insights into other warehouse technologies (RFID, PBV, goods-to-picker stations) – for example, indicating why fulfilment errors made in warehouse operations persist after implementation. We also corroborate previous studies (e.g. Fontin and Lin 2020) showing that novel technology introduces new complexities and thus new potential error sources, as well as overtrust in the new technological phenomenon (Loske 2022).

This study also makes an important contribution explaining how and why fulfilment errors are caused in warehouse operations. Our findings confirm that an error and its cause can be due to multiple technical, organisational, and human factors (Škerlič and Muha 2017) and that operators may be erroneously blamed for errors that reflect larger structural problems. This provides insights into the complexity of analysing sociotechnical systems and identifying root causes for accurate improvement initiatives (cf. Loske 2022), though IVA helps address this complexity by enabling managers to quickly correct technical issues and flawed processes.

Finally, our study contributes to understanding and addressing human factors – an important topic that has, until recently, received limited attention in the literature (Grosse et al. 2015; Neumann et al. 2021). Our case analysis showed that perceptual (confusion), cognitive (experience), and psychosocial (distraction, stress, and carelessness) factors are predominant causes, and the main remedy to reduce errors is to provide feedback on different levels and involve staff in identifying possible solutions, though errors due to human factors are still difficult to resolve. IVA could help with incorporating human factors into technology by not only offering insights into how humans interact with new technologies (cf. Sgarbossa et al. 2020) but also providing frequent feedback loops that can accelerate human learning. The presence of cameras also seemed to result in error reduction – implying that the perception of being monitored (by management) increases the operator’s focus, though there may be undiscovered psychosocial impacts (e.g. motivation and stress).

6.2 Limitations and future research

Our study provides multiple avenues for future research. Further research should focus on the consequences of IVA on the operators and their environment – especially a psychosocial perspective on operators’ ‘increased awareness’, as this study was limited to managerial input. A multidisciplinary approach should be taken and should consider legislation and policymaking regarding video monitoring. Furthermore, a real-time application of IVA, currently being piloted at one company at the time of this research, can alert the operator if they are making a mistake, potentially preventing the mistake from occurring. This shows promise in resolving human-related errors that seem difficult to address and prevent with current corrective actions and technologies. Further research into real-time IVA would not only advance the understanding of the effects of this new technology but also show how it can complement the IVA studied in this paper and other technologies.

The novelty of IVA as a technology in warehouse operations comes with some limitations. To the researchers’ best knowledge, only a few companies currently use the technology, which limited the number of participants in the study. Furthermore, IVA has been in use for a limited time, and therefore, additional consequences (e.g. on financial and performance metrics) of the technology use may have been missed due to its novelty and a lack of data. It is also relevant to further investigate if, how, and why IVA substitutes, complements, and/or supports other technologies for certain purposes. As more companies implement IVA, it would be valuable to study more cases from various contexts, including different countries, different types of warehouses, and different types of product and order characteristics.

Acknowledgements

We gratefully acknowledge the Editors and the anonymous reviewers for their guidance and constructive comments, and Dr. Ebba Eriksson for support as a friendly reviewer.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability

The data that support the findings of this study are available from the corresponding author [[email protected]] upon reasonable request.

Additional information

Notes on contributors

Max Helm

Max Helm holds a Master’s degree in Mechanical Engineering specialised in Logistics and Production Economics at Lund University. Previously, he worked as a logistician with a focus on warehouse management systems for a Norwegian company in the dairy industry. He currently works with business development at a logistics technology company, focusing on novel applications of artificial intelligence and video technology in warehouse operations.

Alexandra Malikova

Alexandra Malikova holds a Master of Science in Transportation from Massachusetts Institute of Technology, and a Master’s degree in Logistics and Supply Chain Management from Lund University. In her professional career, Alexandra worked at two of the leading transportation providers in New York City and recently conducted a project with the UN World Food Programme in Nairobi, Kenya. She currently works as a consultant specializing in supply chain management, logistics, and start-up operations.

Joakim Kembro

Joakim Kembro is an Associate Professor, awarded the distinction of Excellent Teaching Practitioner at Lund University, Sweden. Prior to joining academia, he worked as a logistician in various positions with the United Nations. His research includes omnichannel retailing and logistics, warehouse operations, information sharing and technology, and humanitarian logistics. He publishes in leading OM and SCM journals such as Journal of Operations Management, Production and Operations Management, and Journal of Supply Chain Management. He has received multiple Emerald Awards and is currently a member of the JHLSCM Editorial Advisory Board.

Appendix A