Managed by a Machine: Workers' Job Crafting Abilities in the Case of Lieferando Riders in Germany

(1)

Managed by a Machine:

Workers’ Job Crafting Abilities in the Case

of Lieferando Riders in Germany

MASTER’S DEGREE PROJECT

THESIS WITHIN: Business Administration PROGRAM OF STUDY: Digital Business AUTHORS: Sandra Henkel & Gesa Köhrbrück Jönköping, May 2020

(2)

i

Master Thesis in Business Administration

Title: Managed by a Machine: Workers’ Job Crafting Abilities in the Case of Lieferando Riders in Germany

Authors: Sandra Henkel and Gesa Köhrbrück

Tutor: Matthias Waldkirch

Date: 2020-05-18

Key terms: Job Crafting, Algorithmic Management, Gig Economy, Food Delivery Riders

Acknowledgments

The authors would like to thank everyone who supported them throughout the process of writing the master thesis.

Most importantly, the authors would like to express the deepest gratitude to their supervisor

Matthias Waldkirch for his persistent help and valuable feedback on this thesis. Without his

extensive support, the goal of this thesis would not have been realized.

Furthermore, the authors would like to thank the members of the seminar group (Finn Brand,

Filip Höcker, Ouafaa Cherradi, Cansu Tetik, Stephanie Muth & Marius Rauscher) who provided

constructive and valuable feedback in each seminar. Additionally, the authors would like to thank family and friends who took the time to read the thesis and give constructive and precious criticism from an outside perspective.

Lastly, the authors would like to thank all the participants of the interviews who provided valuable insights into their work. Without them, this thesis could not have been accomplished.

(3)

ii

Abstract

Background:

Despite the utilization of algorithms as data management tools, they are increasingly used as people management tools to allocate, optimize and evaluate workers. This is especially popular among digital labor platforms of the gig economy as it is seen as one of the core innovations that enabled such platforms. Usually, these platform workers are self-employed, which results in an apparent autonomy while working under a rigid algorithm. For those workers, proactively shaping the job according to their own needs and abilities, commonly known as job crafting, may be increasingly important. As research suggests that job crafting occurs across professions and industries, how is it possible under the constraints of algorithmic management?

Purpose:

This thesis investigates the abilities of German food delivery riders of the company Lieferando to perform job crafting while being managed by an algorithm.

Method:

To meet the purpose of this study, the authors conducted a qualitative study. The data was collected through technology-mediated interviews with riders of the company Lieferando in Germany. The authors applied an online recruitment strategy through various social media websites to find suitable interviewees. Interview partners were picked with a random sampling strategy. The interviews were semi-structured, and the researchers guided the interviewees through a previously prepared topic guide with open-ended questions.

Conclusion:

The results of this study provide empirical evidence that riders of the food delivery company Lieferando engage in job crafting activities although working under the constraints of algorithmic management. The outcomes further show that all riders performed task crafting and cognitive crafting in various ways, whereas engagement in relational crafting was less developed. Riders not only have the ability to modify their work but also enrich it.

The findings of this study allow to draw several theoretical and managerial implications as well as provide possible research opportunities for future studies.

(4)

iii

4.1 Algorithmic Management ... 28 4.2 Task Crafting ... 33 4.3 Cognitive Crafting ... 45 4.4 Relational Crafting ... 50 CONCLUSION ... 56 DISCUSSION ... 57 6.1 Theoretical Implications ... 57 6.2 Managerial Implications ... 58 6.3 Limitations... 59 6.4 Future Research ... 60 REFERENCES ... VII APPENDICES ... XII

(6)

v

Figures

Figure 1: Job Crafting in the Context of Algorithmic Management ... 28

Appendices

Appendix 1: Article Search ... xii

Appendix 2: Empirical Context – Lieferando ... xiii

Appendix 3: Interview Sampling ... xv

Appendix 4: Overview of Interviews ... xvi

Appendix 5: Topic Guide with Example Interview Questions (English)... xvii

Appendix 6: Topic Guide with Example Interview Questions (German) ... xx

Appendix 7: Data Structure based on the Interviews... xxiii

Appendix 8: Screenshots Scoober App Contact Form ... xxv

(7)

1

Introduction

The purpose of this section is to provide background information on algorithmic management and to introduce the reader to platform work in the gig economy. Additionally, the problem examined in the thesis is formulated followed by the purpose and the study’s contribution to theory and practice.

______________________________________________________________________

1.1 Background

In recent years, several industries have experienced a rise in algorithmic management, where algorithms are being used to allocate, optimize and evaluate workers (Lee, Kusbit, Metsky, & Dabbish, 2015). Despite the utilization of algorithms as a data management tool in traditional industries such as engineering, logistics and delivery, algorithmic management is seen as one of the core innovations that enabled digital labor platforms like Uber, Amazon MTurk, Upwork or Takeaway.com (Lee et al., 2015). These platforms are all part of the rising gig economy, which refers to a work environment with limited projects (gigs), rather than full-time employment (Jarrahi, Sutherland, Nelson, & Sawyer, 2019; Wood, Graham, Lehdonvirta, & Hjorth, 2019).

Worldwide, around 70 million workers are supposed to be part of the gig economy and have registered with an online labor platform (Heeks, 2017). It is even estimated, that the number will grow further at an annual rate of 26 percent (Kässi & Lehdonvirta, 2018). The gig economy can loosely be described as “platform”, “sharing” or “on-demand” economy (De Stefano, 2016). Most commonly, platform workers are individual, self-employed contractors (Duggan, Sherman, Carbery, & McDonnell, 2020), which results in apparent autonomy and higher perceived flexibility (Duggan et al., 2020; Rosenblat & Stark, 2016) while at the same time working under the constraints of an algorithm. For those workers, proactively shaping the job according to their needs and abilities may be increasingly important.

Literature distinguishes four main types of digital platforms in the gig economy: food delivery, lodging marketplace, ride sharing and freelancing (Jarrahi et al., 2019; Kaine & Josserand, 2019). However, one sector has gained momentum in recent years and is undergoing a rapid change: food delivery (Hirschberg, Rajko, Schumacher, & Wrulich, 2016). By 2025, online food delivery is expected to reach a global supersize with a revenue of 200

(8)

2

billion US dollars, which corresponds to 185 billion Euros (Singh, 2019). Takeaway.com is one of the largest food delivery companies in Europe, having acquired several smaller companies in recent years (Olson, 2019). Besides Takeaway.com, UberEATS and Deliveroo are further food delivery platforms in Europe. After its merger with Just Eat in 2019, Takeaway.com is the market leader in online food delivery in Europe (Statista, 2019b). In Germany, Takeaway.com has a monopoly and operates under the brand name Lieferando (Hergert, 2019).

With the help of different theories and concepts such as job quality theory (Wood et al., 2019), boundary resources (Jarrahi et al., 2019) and self-determination theory (Jabagi, Croteau, Audebrand, & Marsan, 2019), previous studies have already provided some insights into algorithmic management in the gig economy, as well as the impact of algorithms on work. Nevertheless, little research has been conducted regarding job crafting theory (Wrzesniewski & Dutton, 2001) which would, however, be desirable (Kim, Im, Qu, & Namkoong, 2018; Lazazzara, Tims, & de Gennaro, 2020). Previous qualitative studies have already used this theory (Berg, Grant, & Johnson, 2010; Berg, Wrzesniewski, & Dutton, 2010).

1.2 Problem Discussion

Algorithms are no longer exclusively used as data management tools but also to manage employees (Lee et al., 2015). Previous research has shown, which role algorithmic management takes when workers are controlled by it (Jarrahi et al., 2019). This new form of control, where individuals are managed by a machine, is especially present in the gig economy (Kellogg, Valentine, & Christin, 2020; Sun, 2019).

Researchers have already investigated this phenomenon and uncovered several new challenges and opportunities for the workers (Kellogg et al., 2020; Lee, 2018; Lee et al., 2015). In particular, studies have examined the lack of transparency that algorithmic management implies as one of the main challenges for workers of the gig economy (Jarrahi et al., 2019; Möhlmann & Zalmanson, 2017). The complexity of the algorithm makes it very difficult for them to understand the processes of control and management (Jarrahi et al., 2019). Nevertheless, many individuals make an effort to develop an understanding of the technology (Jarrahi & Sutherland, 2019). Therefore, a decent amount of the recent research about the gig economy and algorithmic management has focused on how gig workers try to

(9)

3

make sense of algorithms regarding the lack of transparency and existing information asymmetry (Jarrahi et al., 2019; Lee et al., 2015).

Furthermore, some researchers have investigated the opportunity for workers to take advantage of autonomy and flexibility, which is facilitated by the algorithm (Jarrahi et al., 2019; Kellogg et al., 2020; Shapiro, 2018; Wood et al., 2019). According to Wrzesniewski and Dutton (2001), autonomy in the job encourages workers to customize their job and push the boundaries of tasks and relations. Bader and Kaiser (2019) even claim that as a consequence of the given autonomy and existing information asymmetry, workers start to build workarounds. Lee et al. (2015) examined the impact of algorithmic management on workers and also discovered that riders use workarounds for their tasks to keep control over their assignments. In addition, research has shown that workers transform work practices and norms to influence the interaction with the management tool (Jarrahi et al., 2019).

Literature has already analyzed the working conditions for gig workers, arguing that there is a tradeoff between autonomy, flexibility, task variability and social isolation, low pay, exhaustion and irregular working hours (Jarrahi et al., 2019; Kaine & Josserand, 2019; Kellogg et al., 2020; Pichault & McKeown, 2019). Further research reveals more negative aspects of gig work and calls for action of regulators and other involved stakeholders (Goods, Veen, & Barratt, 2019; Kellogg et al., 2020). They focus on the wellbeing and job conditions of workers or the quality of the jobs (Goods et al., 2019). Another study even examined how the relationship between the gig workers and the algorithm can be explained (Kellogg et al., 2020). However, the explicit investigation of algorithmic control and their impact on platform workers’ experiences would be desirable (Griesbach, Reich, Elliott-Negri, & Milkman, 2019) to not only increase the effectiveness of platform work but also enhance worker satisfaction (Lee et al., 2015). Especially during the war for talent, companies must encourage loyalty and maintain a high retention rate of their workers (Bhattacharya, Sen, & Korschun, 2008). This is only possible if employees are satisfied with their jobs (Bhattacharya et al., 2008).

Research has shown that workers who can customize their job show an increase in wellbeing and job satisfaction (Tims, Bakker, & Derks, 2013). This results from job crafting activities where workers can redesign aspects of their job according to their needs and abilities (Tims et al., 2013). Abilities to craft the job have already been examined in several industries and contexts (Tims et al., 2013; Wrzesniewski, LoBuglio, Dutton, & Berg, 2013). However, little to nothing is known about job crafting possibilities in the gig economy, which is one of the increasing alternative forms of work arrangements (Lazazzara et al., 2020). Researchers even

(10)

4

claim that job crafting research in the gig economy would be desirable due to the different workplace characteristics of the gig workers (Lazazzara et al., 2020). In particular, to the best of the authors’ knowledge, gig workers who are managed by an algorithm have not been investigated yet concerning job crafting. Among others, this is the case of workers of the food delivery platform Lieferando. Thus, after reviewing previous research, the authors identified job crafting theory as being applicable to this context. The paradox between algorithmic control as being more opaque and more instantaneous than any other form of management and job crafting as a tool to modify and redesign a job makes this problem worth to study (Kellogg et al., 2020; Tims et al., 2013). Even though, this thesis focuses on algorithmic management in the food delivery industry the findings might be partly applicable to other areas, where workers are controlled by an algorithm (Jarrahi et al., 2019).

1.3 Purpose

As the authors could identify the biggest research gap in existing literature within food delivery platforms, this thesis focuses on this industry as one of the growing sectors of the gig economy (Goods et al., 2019). Literature distinguishes between two different kinds of food delivery platforms: platform-to-consumer delivery and restaurant-to-consumer delivery. Since platform-to-consumer is estimated to be the leading delivery service, generating 86.0 billion US dollars in revenue worldwide by 2024 (Statista, 2019a), this thesis focuses exclusively on this kind of platform. Furthermore, the main attention is drawn to the company Takeaway.com and specifically to their German brand Lieferando. The decision to investigate specifically Lieferando and the German market was predominantly made for two main reasons. First, Germany is one of the biggest and most promising markets for Takeaway.com, when looking at gross merchandise value (Takeaway.com, 2020). Hence, the German market is of high interest in this research stream. Second, the authors are both from Germany and therefore already demonstrate a good knowledge about the market. In addition, accessibility to the research context exists.

In brief, this thesis examines, the abilities of German food delivery riders of the company Lieferando to perform job crafting while being managed by an algorithm. Hence, the research question of this study is as follows:

RQ: How do riders of the food delivery company Lieferando craft their job while being managed by an algorithm?

(11)

5

Deriving from the purpose, this thesis exerts a worker’s perspective and, in particular the perspective of algorithmic-managed food delivery riders of Lieferando in Germany. Thus, riders of other food delivery services as well as all other employees of Lieferando are not of interest to this research. Nevertheless, the pilot testing of the interview was done with a former worker of the food delivery platform Foodora, which was acquired by Lieferandos parent company Takeaway.com in December 2018 (Ksienrzyk, 2019).

1.4 Contribution to Theory and Practice

Overall, this thesis aims to test job crafting theory in a new context. This can be argued as the researchers aim to find out how food delivery riders of the company Lieferando can perform job crafting. Although this theory has been utilized in several contexts and industries, little research exists about job crafting abilities of algorithmic-managed workers and how they experience it. Thus, this thesis contributes to the research stream in the field of algorithmic management as well as job crafting since they have never been examined in combination to the authors’ best knowledge. It can further help to adjust the job crafting framework in the context of algorithmic management. The results of this study can provide deeper insights into the paradox of autonomy and flexibility and how this is perceived and handled by the workers. This study can help to better understand to which extent algorithmic-managed workers can craft their jobs according to their needs and abilities although supposedly following a rigid process that cannot be influenced or modified. Furthermore, this research can be highly relevant for the company Lieferando as well as other food delivery companies that manage their riders through an algorithm. It may help these companies to better understand how they can support their workers to enhance their job crafting abilities. Based on that knowledge, positive outcomes such as job satisfaction, employee retention and productivity can be achieved simultaneously for the companies and their workers. To conclude, this thesis aims to make contributions to both, theory and practice.

(12)

6

Literature Review

____________________________________________________________________________________

The purpose of this chapter is to provide the theoretical background for this topic. First, the authors give an overview of algorithmic management, followed by a presentation of previous research about the gig economy. Subsequently, deeper insights about the theoretical background of job crafting follows.

______________________________________________________________________

2.1 Approach to Literature Review

The authors first conducted an extensive literature research, in order to write an abundant literature review. To access a wide range of relevant and high qualitative literature, the research was performed on three online databases (Primo, Google Scholar and Web of Science). Keywords such as “algorithmic management”, “algorithmic management” AND “food delivery platforms”, “algorithmic management” AND “gig economy”, “algorithmic management” AND “job crafting”, “gig economy”, “job crafting” AND “gig economy”, “job crafting” AND “food delivery platforms”, and others, were used to find relevant articles (see Appendix 1: Article Search, for detailed list).

2.2 Algorithmic Management

2.2.1 Overview from a Business Perspective

Algorithms are computer-programmed procedures, that convert input data into the desired output in a way, that is supposed to be more comprehensive, instant, interactive and obscure as former technological systems (Gillespie, 2018; Kellogg et al., 2020). Algorithmic management, also called “Scientific Management 2.0”, describes a new way to track and manage workers by optimized decisions regarding their tasks (Schildt, 2017). Thereby, algorithms take full control over the workers, schedule their work and assign tasks (Duggan et al., 2020). Algorithmic management can therefore be defined as “management of labor by machine” (Kaine & Josserand, 2019, p. 493). These workers are no longer managed by a human but by technology (Schildt, 2017) and specifically by algorithms, who start to take up leadership roles (de Cremer, 2019). In particular, algorithmic management relocates the power from managers to professionals with high data analytics skills (Schildt, 2017). Although algorithmic management over labor is not new, it has gained momentum in recent years after a series of bankruptcies in the end of the last century (Schildt, 2017). As a result, organizations are redesigned, and workplaces are innovated (Schildt, 2017) and become

(13)

7

increasingly quantified as well as monitored by algorithms (Leicht-Deobald et al., 2019). The main purpose of using algorithmic management is to optimize processes and calculate predictions, thus leading to more efficiency (Schildt, 2017).

Several recent papers look at algorithmic management from a critical perspective (Griesbach et al., 2019; Kellogg et al., 2020; Veen, Barratt, & Goods, 2019; Wesche & Sonderegger, 2019). One of the main investigated topics is the new form of control through algorithms (Kellogg et al., 2020; Wesche & Sonderegger, 2019). From a historical perspective, algorithmic management recreates many characteristics of older procedures of work control (Griesbach et al., 2019; Veen et al., 2019). Especially, previous research on the labor process, observes strategies of how employers control their employees (Griesbach et al., 2019). Kellogg et al. (2020) as well as Wesche and Sonderegger (2019) mainly look at the decision-making process of an algorithm and how it is perceived. The primary difference in the decision-making processes is that with algorithmic management, decisions are based on data (Fountain, McCarthy, & Saleh, 2019). Through an automated process an algorithm comes to a conclusion based on the input data. Previous literature already analyzed and compared automated algorithmic decisions and human decision-making (Bader & Kaiser, 2019; Jarrahi et al., 2019; Kellogg et al., 2020; Wesche & Sonderegger, 2019) and suggests, that a decision might be perceived differently depending on who made it (Lee, 2018).

The authors Kellogg et al. (2020) as well as Bader and Kaiser (2019) claim that decisions made by algorithms are more accurate compared to human made decisions. According to Leicht-Deobald et al. (2019), algorithmic decision making can help human resources departments to monitor employees more accurately, but simultaneously be problematic in an ethical manner. It can harm the personal integrity of employees because it may produce blind trust in processes and rules, thus excluding human sense-making as algorithms lack the capability of moral imagination (Leicht-Deobald et al., 2019). Additionally, Tambe et al. (2019) highlight that a big advantage of human based decisions is that they can be adjusted, which is still a huge challenge for decisions from machines.

2.2.2 Overview from a Worker’s Perspective

Algorithmic management not only changes the workplace from a business perspective but has additional implications for workers. In particular, they are no longer managed by a human being but by a machine instead (Kaine & Josserand, 2019).

Several studies claim that the algorithm has a mainly negative effect on the job quality and livelihood for the workers (Goods et al., 2019; Kellogg et al., 2020). It gives managers a new

(14)

8

form of control over their workers, which reshapes the relationship between each other (Kellogg et al., 2020).

Even though algorithmic management gives the workers some sort of flexibility, they do not fully understand how it works (Jarrahi et al., 2019; Kellogg et al., 2020). As a result, workers cannot predict the algorithms’ processes and behavior (Sun, 2019). Researchers describe that it is not transparent and individuals who are managed by it, cannot fully understand the decision-making process because companies do not want to disclose it (Jarrahi & Sutherland, 2019; Lee et al., 2015). This is due to the circumstance that organizations, which use algorithmic management, do not want its workers to outsmart the system (Jarrahi & Sutherland, 2019). However, studies show that workers make an effort and try to make sense of algorithmic management (Jarrahi & Sutherland, 2019; Lee et al., 2015). Human workers even show the phenomenon of “algorithmic aversion”, describing the preference of human input over algorithmic input, even if the algorithm is more exact (Dietvorst, Simmons, & Massey, 2015). Dietvorst et al. (2015) explain this aversion with the loss of control over the job, especially if the algorithm is perceived as being opaque. Furthermore, literature explains that workers who are managed and controlled by an algorithm question its decisions (Kellogg et al., 2020; Lee et al., 2015). This behavior leads to obstacles and workers start to create workarounds and manipulate the technology to escape the control mechanism (Jarrahi & Sutherland, 2019; Kellogg et al., 2020; Lee et al., 2015). A decent amount of the reviewed literature questions the quality and fairness of decisions made by algorithms (Bader & Kaiser, 2019; Kellogg et al., 2020; Lee et al., 2015; Tambe et al., 2019). According to Lee (2018), workers would build more trust if the algorithm was more transparent, which results in better cooperation.

2.2.3 Areas of Application

Work organized and controlled through digital platforms is an increasing phenomenon around the globe (Altenried, 2020). Especially in the gig economy, algorithmic management is emerging (Altenried, 2020) and becoming a strategic management tool for the digital economy (Cheng & Foley, 2019). Organizations within this industry use data driven algorithms to make decisions without involving humans (Schildt, 2017). Thanks to algorithmic management, a large number of workers can be managed at the same time (Lee et al., 2015). As an enabler for business models of the gig economy, algorithmic management is a crucial part for the development of these services (Jarrahi & Sutherland, 2019; Lee et al., 2015).

(15)

9

Increasingly, algorithmic management is also becoming more popular in different work contexts beyond the gig economy (Altenried, 2020), such as within delivery or logistics. For example, Amazon uses wristbands for warehouse workers around the world to surveil order respondence and packaging (Yeginsu, 2018). Another example for the utilization of algorithmic management beyond the gig economy is UPS. The parcel delivery service is using an algorithm for the optimization of delivery routes and the routing of packages (Rosenbush & Stevens, 2015).

2.3 Gig Economy

2.3.1 Characteristics

The gig economy is a new way of work, which is reimbursed on a pay-as-you-go rate, meaning workers are only paid for the actual working hours (De Stefano, 2016). Literature distinguishes different types of platforms in the gig economy as already described in the background. Digital platforms like Uber, Amazon MTurk, Upwork and Takeaway.com are just a few examples, where an algorithm is managing employees through a vast collection of data (Jarrahi et al., 2019; Kaine & Josserand, 2019). Furthermore, algorithmic management is even perceived as one of the core innovations, that made these services possible (Lee et al., 2015). It seems, that algorithmic management practices offer employees more flexibility, autonomy, task variety and complexity (Wood et al., 2019). However, there is no consensus about this view. In contrast, Kellogg et al. (2020) claim that workers, who are controlled by an algorithm lack autonomy.

Work in the gig economy has four basic characteristics: unsteady work schedules due to the dependence on customer demand, workers need to supply any kind of capital such as bikes or smartphones, the work is paid as a piece wage and lastly, work is distributed via platforms (Stewart & Stanford, 2017). Usually, the work is described as micro tasks that do not last very long (Kaine & Josserand, 2019; Kellogg et al., 2020). Workers can pick assignments relatively flexible on a first-come, first-served basis and therefore determine the schedule on their own (Jarrahi et al., 2019; Kellogg et al., 2020). In connection with other criteria, especially the described flexibility leads to various opinions about the status of gig workers among researchers (Jarrahi et al., 2019; Kaine & Josserand, 2019; Kellogg et al., 2020; Ravenelle, 2019; Sun, 2019). Furthermore, one of the main advantages but also the biggest discussion points in previous literature is the autonomy of the workers in the gig economy (Kellogg et al., 2020). According to Möhlmann and Zalmanson (2017), working in the gig

(16)

10

economy and in particular on a digital platform differs from traditional forms of work especially in terms of employment. Jarrahi et al. (2019) describe gig workers as entrepreneurs since they can decide autonomously when and on what they would like to work, which comes close to running their own business in the authors’ opinion. Additionally, Ravenelle (2019) describes how the gig economy can foster entrepreneurship by offering flexibility and autonomy to its workers. In contrast, Kaine and Josserand (2019) as well as Rosenblat and Stark (2016) call them independent contractors as they claim that companies of the gig economy try to circumvent employment-related liabilities by not hiring workers as employees. This opinion is also supported by Kellogg et al. (2020).

Previous literature has already expressed some controversies regarding the gig economy. Workers usually do not have the possibility to interact much with their coworkers or managers (Kellogg et al., 2020; Tassinari & Maccarrone, 2020). Literature highlights that these workers are often isolated from social groups (Kellogg et al., 2020; Tassinari & Maccarrone, 2020). Therefore, they start to build communities where they exchange information and try to overcome isolation (Jabagi et al., 2019; Tassinari & Maccarrone, 2020). Other literature suggests that new business models hire gig workers to reduce labor costs (Kaine & Josserand, 2019). Therefore, literature criticizes how companies reduce their liabilities towards workers (Griesbach et al., 2019), which usually leads to low payments and exhaustion (Kellogg et al., 2020; Wood et al., 2019). This is also supported by Kaine and Josserand (2019) who note that the risk shifts from employers to the workers in the gig economy. In addition, the surveillance has engendered massive criticism from former employees and media in turning humans into machines (Yeginsu, 2018). The negative perception of algorithmic management on digital platforms has led to resistance and some workers even subvert it according to several authors (Kaine & Josserand, 2019; Kellogg et al., 2020; Sun, 2019).

2.3.2 Food Delivery Platforms

The concept of food delivery services is not new but has experienced some changes due to the development of online platforms (Veen et al., 2019). Online food delivery platforms bring restaurants, workers and customers together in their digital ecosystem (Veen et al., 2019). Food delivery riders support restaurants through the algorithmic management method (Griesbach et al., 2019; Tassinari & Maccarrone, 2020). Based on their location, riders are assigned to jobs when they are “signed in” on the platform (Tassinari & Maccarrone, 2020).

(17)

11

However, the riders do not get the chance to choose between tasks but can either accept or reject an order (Griesbach et al., 2019).

Previous studies have already examined the control of workers within the food delivery industry and particularly how it is experienced by them (Griesbach et al., 2019; Veen et al., 2019). Griesbach et al. (2019) point out that workers are mainly limited by technical systems in the form of algorithms. However, Sun (2019) shows how workers of food delivery platforms make sense of the algorithm and even try to reshape it.

Tassinari and Maccarrone (2020) as well as Griesbach et al. (2019) have taken a closer look at the solidarity between workers. While the former shows that solidarity is fostered by algorithmic control, Griesbach et al. (2019) highlight that it rather has a negative and competitive effect on workers’ collaboration.

Several researchers describe that there is an information asymmetry for workers on these platforms as the mechanisms of job distribution is not revealed (Sun, 2019; Tassinari & Maccarrone, 2020). Additionally, literature regarding food delivery platforms criticizes the same issues as papers regarding the gig economy in general. This includes, but is not exclusive, the before mentioned information asymmetry, minimization of labor costs and pushing the liability to the riders as contractors (Kellogg et al., 2020; Tassinari & Maccarrone, 2020; Veen et al., 2019).

Literature shows that there is a paradox between the opaque control through new technologies and information asymmetries as well as simultaneously creating a trigger for mobilization (Gandini, 2019; Shapiro, 2018; Tassinari & Maccarrone, 2020; Veen et al., 2019). Griesbach et al. (2019) describe it as freedom, which is limited through algorithmic control.

Some of the previous literature claim that food delivery companies follow the logic of capitalism as they are still missing regulations (Sun, 2019; Veen et al., 2019). Most of the previous studies are qualitative research (Goods et al., 2019; Griesbach et al., 2019; Sun, 2019; Tassinari & Maccarrone, 2020; Veen et al., 2019). Nonetheless, no previous qualitative research on food delivery platforms has investigated job crafting abilities of riders.

(18)

12

2.4 Theoretical Background

2.4.1 Job Crafting

Jobs are not only designed by organizations in accordance with requirements, but also (re)designed by the workers themselves (Bruning & Campion, 2018; Niessen, Weseler, & Kostova, 2016). The process of proactively shaping the job is called job crafting. Although the concept of job crafting was mainly shaped by Wrzesniewski and Dutton in 2001, the notion was already mentioned more than 20 years before by Kulik, Oldham, and Hackman (1987). Job crafting describes the changes that individuals make to their jobs according to their needs, abilities and preferences (Berg, Dutton, & Wrzesniewski, 2008; Wrzesniewski & Dutton, 2001). Thus, individuals shape the boundaries of tasks or relations and can balance their job demands and job resources with their personal abilities and needs (Tims & Bakker, 2010). In contrast to following a top-down approach to foster work engagement, job crafting follows a bottom-up approach, where employees proactively enhance their job (Tims, Bakker, & Derks, 2012; van Wingerden, Bakker, & Derks, 2017). The central characteristic of job crafting is the fact that employees alter their job self-initiated (Tims et al., 2012; Wrzesniewski & Dutton, 2001). It is important to distinguish job crafting from other bottom-up forms of job redesign (Tims et al., 2012). Furthermore, job crafting is referred to proactive work behavior. According to Parker and Collins (2010), proactive work behavior is initiated by an individual either by acting in advance of a future situation and/or by taking control and causing change. However, self-initiation is a crucial part of both (Parker & Collins, 2010). In fact, job crafting can be described as a specific form of proactive work behavior (Tims et al., 2012).

In addition, job crafting is not onetime (Berg et al., 2008). Instead, Berg et al. (2008) suggest job crafting as being a process that individuals practice over time. The beginning of this process is the individuals’ motivation to craft their job, which derives from three individual needs (Wrzesniewski & Dutton, 2001). First, workers want to maintain control over their work to avoid estrangement. Even if in small ways, by taking control or crafting some factors of work, employees can make the job their own and fulfill the basic human need for personal control (Wrzesniewski & Dutton, 2001). Second, workers want to create a positive self-image in their work. As individuals crave to create a positive sense of self in their own eyes and also recognized by others, changing tasks or relations in their jobs is an important motive (Wrzesniewski & Dutton, 2001). Third, meeting basic human needs for connection to others. To introduce meaning to their lives, employees alter their connections to others to fulfill

(19)

13

their need for human connection (Wrzesniewski & Dutton, 2001). If the current job of an employee does not fulfill these needs, motivation for job crafting most often results from this situation (Wrzesniewski & Dutton, 2001).

More recent literature begins to distinguish between different perspectives on the theory of job crafting. In their paper, Bruning and Campion (2018) describe that one should differentiate between the role-based theory coined by Wrzesniewski and Dutton (2001) and the resource-based theory, which was initiated by Tims and Bakker (2010). Although Bruning and Campion (2018) argue that there are coinciding characteristics of the two perspectives, their differences should not be covered up. On the one hand, the role-based view analyzes how the needs-supply is balanced with the work (Bruning & Campion, 2018). On the other hand, the resource-based view describes how demands are met by searching for resources (Bruning & Campion, 2018). Another major difference highlighted in their study is the fact that resource crafting results in improved efficiency while role crafting improves personal enrichment (Bruning & Campion, 2018; Lazazzara et al., 2020). In addition, Lazazzara, Tims, and de Gennaro (2020) explain that Tims and Bakkers’ (2010) perspective does not consider cognitive crafting. Tims and Bakker (2010) use the Job Demands-Resource model, which is a job design framework, to analyze job crafting activities. Furthermore, the resource-based perspective is mostly used by quantitative studies as it targets to measure variables (Lazazzara et al., 2020). In contrast, the role-based perspective is mainly used for qualitative research as it tries to reveal motivations, underlying reasons, or personal opinions, which is in line with the purpose of this research. Thus, this thesis is based on the job crafting theory coined by Wrzesniewski and Dutton (2001).

2.4.2 Types of Job Crafting

It is important to note that job crafting does not describe the redesign of a job as a whole (Berg et al., 2008). In fact, job crafting involves the change of certain aspects of the job within the boundaries of specific tasks. According to Wrzesniewski and Dutton (2001), employees can craft their job through three strategies: crafting tasks, changing the relationship with others as well as modifying the cognition about the job. The three strategies are not alternatives, but instead they might co-occur as a combination (Wrzesniewski et al., 2013). Task crafting can be defined as changing the task boundaries of a job (Wrzesniewski & Dutton, 2001). In particular, employees can accomplish task crafting by altering the number, scope, or type of their tasks at work. Thus, employees are redesigning the original tasks of their job concerning the time and effort spent on the different features of the task (Lin, Law,

(20)

14

& Zhou, 2017). As an example, Lin et al. (2017) mention improving the procedure of tasks or increasing their complexity. Furthermore, Wrzesniewski and Dutton (2001) mention the action of cutting tasks or adding more or different tasks that can also create a different job than the one prescribed. A teacher spending more time learning new classroom technology to fulfill a passion for information technology is an example of task crafting stated by Wrzesniewski et al. (2013).

The second job crafting strategy mentioned by Wrzesniewski and Dutton (2001) is relational crafting, which can be described as the changing of relational boundaries of a job. This includes either altering the quality or the amount of interactions with other individuals at work or both. For example, the development of new relationships, reducing and avoiding relationships with others or increasing the time spent with preferred colleagues are mentioned as changes in regard to interpersonal relationships at work by Wrzesniewski and Dutton (2001). Wrzesniewski et al. (2013) suggest, that employees from different departments can form a relationship to better understand the impact of their work on other departments. As an example, they name the relationship of a marketing analyst with an employee of the sales team to come to a real understanding of the impact of his work on the salesperson.

Cognitive crafting as the third job crafting strategy describes the changes in cognitive task boundaries of a job (Wrzesniewski & Dutton, 2001). These cognitive changes can appear in several forms. For instance, employees can perceive and interpret parts of their jobs (e.g. tasks or relationships) or the job as a whole different than prescribed to alter its significance (Wrzesniewski & Dutton, 2001). Thus, by changing the way of thinking about the job, individuals modify how they approach their jobs (Lazazzara et al., 2020). As an example, Wrzesniewski et al. (2013) mention a custodian as thinking about himself as an enabler of education by making clean and distraction-free classrooms available for students. However, the activity of cognitive crafting of jobs is seen as controversial (Niessen et al., 2016). Tims and Bakker (2010) claim that cognitive crafting is not about actively shaping the boundaries of the work. Instead, individuals do only revise the perspective on the work circumstances that are not in accordance with their needs, abilities and preferences. Thus, it only involves coping with the given work conditions, which is not corresponding with the definition of job crafting by Wrzesniewski and Dutton (2001). Nevertheless, in line with Niessen et al. (2016), the authors of this thesis believe that cognitive crafting is as important as task and relational crafting to achieve a better job fit.

(21)

15

More recent studies have widened the limitation of changes to task, relational and cognition and suggested possible other forms. For instance, Lyons (2008) explored the engagement in self-initiated skill development of salesperson in their research. Furthermore, Grant, Fried, Parker and Frese (2010) described that employees working in a service job avoided serving unfriendly customers. Lazazzara et al. (2020) further distinguish the three crafting strategies task, relationship and cognitive crafting by looking at the avoidance and approach of each of them. Additionally, they describe crafting in other domains like location or temporal crafting as another dimension. Nevertheless, this study mainly touches upon the three suggested strategies of job crafting by Wrzesniewski and Dutton (2001), which lay the foundation for job crafting theory.

2.4.3 Enablers of Job Crafting

Research has found specific factors that enable job crafting. Several enablers have already been identified and evaluated especially by the study of Tims and Bakker (2010), which differentiates between characteristics of the work environment and personal characteristics of the workers that influence job crafting abilities.

By looking at the work environment and in particular, the fit between the demands of a job and the individual skills, knowledge and needs, Tims and Bakker (2010) claim that a misfit could enable workers to perform job crafting activities. They assume that workers know what leads to the misfit and can therefore balance it by engaging in job crafting. This is also supported by Frese and Fay (2001) who suggest that proactive behavior like job crafting is a result of job dissatisfaction or misfit. Especially nowadays, where expectations on a job go beyond economic factors, workers are interested in opportunities, which make their job more meaningful (Rosso, Dekas, & Wrzesniewski, 2010).

Furthermore, several authors highlight that autonomy is a driver of job crafting (Kim, Im, & Qu, 2018; Tims & Bakker, 2010; Wrzesniewski & Dutton, 2001). Only if workers feel that they have the opportunity and freedom to change something they will be triggered to engage in job crafting (Wrzesniewski & Dutton, 2001). However, it is important to notice that only the perceived degree of autonomy is relevant, the actual freedom does not matter in this case (Lyons, 2008). In contrast, Wrzesniewski and Dutton (2001) point out that workers who are limited in their freedom and creativity might find motivation in this constraint to modify their job. However, they recommend future research to confirm this assumption.

Another relevant aspect that enables workers to craft their job is the possibility to perform a task independently (Tims & Bakker, 2010). A worker whose individual performance is not

(22)

16

dependent on the success of others is more likely to adjust the job as the changes would not affect another worker (Tims & Bakker, 2010). Since job crafting is carried out to increase the personal benefits and changes might have negative effects on a coworker, it is usually not executed if the job tasks involve more than one worker (Wrzesniewski & Dutton, 2001). Next to the environmental factors there are also some individual attributes, which make it more likely for workers to engage in job crafting. As job crafting is considered to be a proactive behavior, it is more likely that workers do job crafting if they have a proactive personality (Crant, 2000; Tims & Bakker, 2010). Moreover, self-efficacy plays an important role when it comes to job crafting and individual traits (Tims & Bakker, 2010). This is due to the belief that workers who trust their abilities and skills to change the work environment are more likely to act and examine job crafting (Tims & Bakker, 2010).

2.4.4 Obstacles to Job Crafting

Despite the beneficial nature of job crafting and even though enabling job characteristics for job crafting are present in all jobs nowadays (Tims, Derks, & Bakker, 2016), not every individual has the opportunity to engage in it (Brickson, 2011; Tims & Bakker, 2010). Literature has already examined existing obstacles to job crafting.

Wrzesniewski and Dutton (2001) claim that contextual factors may limit the opportunities of employees to craft their jobs. According to them, employees may be hindered from job crafting when their jobs are explicitly defined and controlled. This is especially the case in the modern workplace as technology has enabled an increasing control of organizations and supervision (Wrzesniewski & Dutton, 2001). Furthermore, Wrzesniewski and Dutton (2001) see constraints to job crafting depending on the level of an employees’ task interdependence. In particular, the freedom to craft the job decreases with a higher level of task interdependence as the timing and tasks of others limit individual changes.

Other authors mention hierarchical levels as an obstacle to job crafting (Berg et al., 2010). More precisely, Berg et al. (2010) suggest that the employees’ perceived freedom to do job crafting does not automatically reflect their level of power and autonomy. Instead, employees at lower ranks find themselves having more possibilities to craft their job due to easier adaption to work environments than higher-ranked employees with greater autonomy and power as they feel more constrained. Although both, employees in low and in high ranks engaged in job crafting, they faced different barriers. While lower-ranked employees had to change others’ expectations, higher-ranked employees had to change their own expectations.

(23)

17

Kim, Im, Qu and Namkoong (2018) mention isolated work environments as a barrier for especially relational crafting. However, other types of job crafting might still be possible such as task crafting, according to them. Furthermore, they suggest that managers should carefully watch their employees to identify other possible hindering factors to job crafting (Kim, Im, Qu, et al., 2018).

Deriving from personal experiences while engaging in job crafting practices, Brickson (2011) mentions three different sources of obstacles: professional values, job structure rigidities and publishing barriers. However, Brickson (2011) is solely focusing on job crafting in scholarship. Thus, although important insights into existing barriers to engage in job crafting are provided, they might not be fully applicable in other work contexts.

2.5 Algorithmic Management and Job Crafting

Job crafting is not limited to specific jobs or industries (Nielsen & Abildgaard, 2012). Instead, it occurs in various work contexts as previous research shows. Some research can be found in the consumer goods sector (Lyons, 2008), area of health care (Dutton, Debebe, & Wrzesniewski, 2000), hotel business (Kim, Im, & Qu, 2018), mining and manufacturing industry (de Beer, Tims, & Bakker, 2016) or the service sector in the airline industry (Hur, Shin, Rhee, & Kim, 2017). Furthermore, Tims et al. (2012) are using a sample of employees from various sectors, such as health care, service or education in their research. However, to the authors’ best knowledge, there exists no research regarding job crafting possibilities of algorithmic-managed workers. In particular, the workplace of algorithmic-managed workers is characterized by a new form of control (Kellogg et al., 2020). Due to the increased surveillance, this does not seem to promote the ability to proactively shape the job according to personal needs and capabilities (Wrzesniewski & Dutton, 2001). Thus, to the understanding of the researchers, the employee’s possibilities to exert influence on the job remains limited in the first place. However, Wrzesniewksi and Dutton (2001) claim that even in the most constrained and routine jobs, workers can wield some influence on the nature of work. Therefore, job crafting might be a possibility for the algorithmic-managed workers to regain control over the work that is perceived as opaque (Dietvorst et al., 2015; Jarrahi & Sutherland, 2019), and to fulfill “the basic human need for control” (Wrzesniewski & Dutton, 2001, p. 181).

(24)

18

Methodology

____________________________________________________________________________________

The following chapter describes the research philosophy and applied research approach. It continues with the research purpose, research design and strategy and provides a structured guideline of the conducted investigation and its sampling method.

______________________________________________________________________

3.1 Research Philosophy

The research philosophy is presented to help the reader understand the researchers’ interpretation of the surrounding by clarifying necessary assumptions (Saunders, Lewis, & Thornhill, 2009). These assumptions further justify the research strategy and research methods, thus leading to a conclusively designed research. By thinking about philosophical assumptions of the study the researchers enhance their creativity and expect a positive impact on the quality of research (Easterby-Smith, 2015).

Most debates among philosophers about the relationship between data and theory relate to ontology and epistemology (Easterby-Smith, 2015). The basic assumptions about the nature of reality and existence are explained by ontology, while epistemology gives the researchers an understanding of the nature of knowledge (Easterby-Smith, 2015). Social science, to which this thesis contributes, differentiates between three different viewpoints: internal realism, nominalism and relativism (Easterby-Smith, 2015). In this thesis, the researcher’s ontological position is relativism. The authors understand that different riders may have various perceptions of the control by an algorithm, which implies that there is no single truth. This is in line with the relativist position which assumes, that multiple truths exist as individuals hold different views and have different experiences (Easterby-Smith, 2015). Given the fact that the researchers are conducting interviews with a certain number of riders working for the food delivery platform Lieferando, they accept the existence of different experiences, which encourages the relativist position. There is no universal truth as the considerations and perceptions of the riders differ (Easterby-Smith, 2015). The research does not focus on any measurable characteristics but on the riders’ capabilities, which further supports a relativist viewpoint (Easterby-Smith, 2015). Furthermore, the researchers interact with the participants through interviews rather than just looking at them from the outside, like the realist ontological perspective does. This helps the authors to get an in-depth understanding of the participant’s opinions.

(25)

19

After being clear about the ontology, the epistemology is discussed to build a consistent comprehension of the researchers’ theory of knowledge (Easterby-Smith, 2015). Evolving from the relativist ontological perspective, the researcher’s epistemology is constructionism. Among other reasons, this choice has been made due to the correspondence between the two positions. Instead of assuming that reality is external and objective as used by positivist studies, the researchers view reality as being determined by people and their different experiences, which encourages the constructionist position (Easterby-Smith, 2015). This study aims to understand how riders make sense of their world by listening to their experiences, which reflects the idea of social constructionism (Easterby-Smith, 2015). In addition, the researchers rate a small number of research participants in-depth as more relevant than conducting a survey with a large number of participants like in positivist research. In particular, the researchers are more interested in individual’s experiences and stories rather than objective measures.

There are, unfortunately, some weaknesses of the constructionist position that need to be taken into consideration. For example, data collection can be highly time and resource consuming as well as data analysis and interpretation can be difficult (Easterby-Smith, 2015). However, the researchers rely on the suitability of this position for the thesis since the beforehand mentioned strengths outweigh the weaknesses.

3.2 Research Approach

Saunders, Lewis and Thornhill, (2012) distinguish two research approaches for qualitative research: deductive and inductive. When using an inductive approach, researchers formulate their research question based on an already existing theory (Saunders et al., 2012). In this thesis the research is based on the previously introduced job crafting theory. Using an inductive approach helps the authors to analyze the data gathered from in-depth interviews with the food delivery riders concerning the underlying theory. This enables the researchers to establish a connection between the new findings and the already existing studies. Before analyzing the data, the authors examined the job crafting theory regarding its components, issues and themes. The examination of theory forms the base for an analysis of the data (Saunders et al., 2012). After a great understanding of the theory is ensured, the authors conducted their empirical study to analyze and evaluate job crafting in the new context.

(26)

20

3.3 Research Purpose

The purpose of a research can be exploratory, descriptive or explanatory (Saunders et al., 2009). However, it is also possible, that the research follows more than one purpose. An exploratory study is used, when a researcher wants to elucidate the understanding of a problem, e.g. through a search of literature or conducting interviews with “experts” or focus groups (Saunders et al., 2009). Thus, the research question usually scrutinizes how a phenomenon occurs (De Massis & Kotlar, 2014), which correlates with this study. The research purpose of this thesis is to explore how riders of the food delivery platform Lieferando can craft their jobs while being managed by an algorithm. The empirical study aimed to explore the different aspects of job crafting in the context of the gig economy. Therefore, the authors conducted interviews with riders of the food delivery platform Lieferando in Germany. The conduct of interviews leads to more flexibility and might trigger the unconscious mind of the interviewee (Sreejesh, 2014). Thus, the research purpose of this thesis is exploratory.

3.4 Research Design

The research design clarifies, what the researchers observe and how it is observed (Easterby-Smith, 2015). It can either be quantitative, qualitative or a combination of both in a mixed-method approach (Saunders et al., 2009).

The authors chose to do a qualitative research by conducting technology-mediated interviews. In contrast to a quantitative research, the answers are not limited to a given number, but openly designed (Sreejesh, 2014). It ensures that rich data is generated as not only the answers themselves are analyzed, but in addition, the respondents can be observed while answering questions. This helps the interviewer to get a better understanding of the respondents' perspectives and believes (Sreejesh, 2014). Lazazzara et al. (2020) suggest the use of qualitative research to gain more profound insights into job crafting theory and in this case understand the thoughts of delivery riders within the context of algorithmic management. This type of research further helps to understand how workers craft their job to match with their skills and needs to increase their work identity and meaning of work (Lazazzara et al., 2020). The authors decided to pursue a single case study design as the aim is to explain the social phenomenon of job crafting in-depth (Yin, 2014).

(27)

21

3.5 Data Collection Method

To collect data, the authors decided to conduct qualitative interviews. According to Tracy (2020), interviews offer opportunities for mutual discovery, understanding, reflection and explanation as well as outline lived experiences and viewpoints in a subjective way. Hence, interviews provide researchers with information that is otherwise difficult or even impossible to observe (Tracy, 2020). This is exactly the case in the present research as the researchers want to explore the individuals’ personal feelings and experiences while being managed by an algorithm.

The interviews were semi-structured, meaning that the researchers guided the interviewees through primarily selected topics, but did not follow a strict list of predetermined questions. In particular, a previously prepared topic guide (see Appendix 5: Topic Guide with Example Interview Questions (English) and Appendix 6: Topic Guide with Example Interview Questions (German)) with prepared questions provided the loose structure of the interviews to ensure the coverage of all relevant topics. In preparation for the topic guide the researchers consulted the study by Berg et al. (2010) to benefit from the experience of previous research. However, the order of the questions and following-up questions was determined based on the interviewee’s answers. This helped to focus on topics, which are relevant for the research and simultaneously gave the interviewers flexibility to some extent (Sreejesh, 2014). Furthermore, it supported the authors to guide the conversation, but still left enough room for individual, tailored questions, depending on the conversation. However, it is important to note that these interviews needed to be prepared in advance to prevent bias as much as possible (Sreejesh, 2014). In addition, the questions were open-ended, which allowed an exploratory approach (Easterby-Smith, 2015). To gain deeper knowledge through asking the questions, the researchers employed the technique of laddering. Thus, interviewees were asked to provide reasons rather than statements of facts or descriptive accounts by asking “why” questions (laddering up) as well as to give examples of specific situations (laddering down).

As the study concerns the German market and the interviewers are German, interviews were mostly conducted in German language (see Appendix 4: Overview of Interviews). However, interviewees could choose beforehand, if they wanted to conduct the interview in German or English language. In this way the researchers ensured to avoid any constraints due to language barriers of the interviewees. To appeal to the interviewee’s responsibility in answering truthfully, they were informed about the research and more specifically, that this

(28)

22

research is part of a master thesis at Jönköping International Business School in Sweden. However, the researchers only provided basic information about the research topic preliminary to the interview to guarantee unpersuaded answers from the interviewees. Nevertheless, the participants were informed right at the beginning of the interview to receive more information about the research afterward. For example, the underlying theory as well as the research question were mentioned subsequently. In addition, the interviewers followed the recommendation by Gioia, Corley, and Hamilton (2013) and avoided the use of theoretical terminology such as “algorithmic management”, “job crafting” or “task crafting” to not miss out on any key aspects by asserting a predetermined understanding on the topic.

Furthermore, the researchers planned to conduct face to face interviews to have the best chance to collect, understand and interpret the information that reflects the view and understanding of the interviewee (Easterby-Smith, 2015). Unfortunately, due to the global Covid-19 crisis, the researchers had to adapt their strategy and switched to technology-mediated interviews in order to follow the prevailing safety distancing measures and to avoid any physical contact. Nonetheless, this change to technology-mediated interviews resulted in more flexibility for the researchers as well as for the research participants regarding scheduling. The interviews were conducted via the video conferencing services Skype and Teams. Furthermore, the interviews were held individually and not within focus groups. This decision was based on the fact that the research question is very subjective and investigates the personal opinion and perception of individuals. Therefore, the authors did not believe that a focus group would deliver representative results as participants may not talk that openly in a group.

Due to the nature of the interview as being semi-structured and open-ended rather than having a strict structure with pre-coded questions, the entire interaction between the interviewers and the interviewee was recorded. Thereby, no information could be missed by enabling the researchers to listen to the interviews again (Easterby-Smith, 2015). In addition, the authors took digital notes during the interview to better internalize what was said by the interviewee.

It is important to note that even though the researchers did not analyze work identity and meaning of work in-depth the two topics were not excluded from the interview. This is because the participants gave valuable input regarding their job crafting abilities when they were asked questions about these topics. However, looking into work identity and meaning of work in more detail would have exceeded the scope of this thesis.

(29)

23

3.6 Pilot Testing

To ensure that the topic guide and the included questions are relevant and valid, the authors did a pilot testing. It was conducted with a former rider of the food delivery platform Foodora, who reacted to the post in the Facebook Group “Messengers of Germany”. Although the rider did not possess the requirements for being an eligible research participant considering the research purpose of this thesis (e.g. rider at Lieferando), basic requirements were given (e.g. working under the constraints of an algorithm at a food delivery platform) to be eligible for the pilot test to test the interview structure as well as the content of the topic guide. For example, the authors realized the analogy of some questions which resulted in duplicated answers from the interviewee. As a result, some questions were removed, and additional questions were added to the topic guide which seemed to be expedient. However, the overall structure worked well and enabled a smooth interview process. Furthermore, the authors were able to gain highly relevant, additional insights from the pilot test. For example, it became clear that there were major differences between Foodora and Lieferando such as shift distribution and work assignment. In particular, these differences were the reason why the rider stopped working as a food delivery “messenger” when Lieferandos parent company Takeaway.com acquired Foodora. At Lieferando, riders have to report their availability for shifts one week in advance and shifts are distributed based on this information. In contrast, at Foodora, riders were able to pick the shifts by themselves on a day by day basis.

3.7 Sampling Strategy

As the aim of this research is to study the job crafting possibilities of algorithmic-managed workers in the online food delivery industry in Germany, it made sense to only interview riders from one company, in particular Lieferando. The riders who are working for the company can deliver the best insights into their job crafting possibilities under the circumstances of algorithmic management as they can share their own work experience. To acquire a sufficient number of interviewees, a multi-tiered participant recruitment strategy, similar to Goods et al. (2019), was deployed.

This thesis relied on online recruitment (Mendelson, 2007) to recruit Lieferando riders. Interview partners were picked using a random sampling strategy to ensure a high probability to reflect a representative sample (Easterby-Smith, 2015). Initially the authors wanted to recruit research participants by street intercepts (Herzog, 2012). Therefore, the interviewers planned to visit the meeting points or offices in several cities in Germany, but mainly in

(30)

24

Munich, where Lieferando riders start and end their shifts. This sampling decision was based on the platform’s operational area and where the authors of this thesis were located. The plan was to directly conduct the interview on-site if the rider had the time and agreed to it. Otherwise, the authors would have collected the contact details and scheduled an interview session at a later point. However, due to the global pandemic, the authors mainly focused on online recruitment strategies through social media to avoid unnecessary and prohibited social contact. In total, eleven interviewees were recruited via the online platforms Facebook, Instagram and Xing as well as the messenger WhatsApp. Furthermore, the authors made use of the snowball sampling technique (Biernacki & Waldorf, 1981). Therefore, all interview participants were asked if they could refer colleagues who would also like to take part in an interview. Interview participants were told not to reveal any further information about the research to guarantee unbiased answers. The researchers were able to recruit another two interviewees with this technique, resulting in a total sampling of thirteen interviewees. More detailed information regarding the sampling procedure can be found in the Appendix (see Appendix 3: Interview Sampling).

3.8 Analysis of Data

As a first step and preparation for the analysis, the interview recordings were transcribed. Therefore, the researchers used the transcription software from trint.com and sonix.ai for the interviews held in German language, and otter.ai for the interviews held in English language. In particular, the software was used to accelerate and facilitate the transcription process. Nevertheless, to ensure the correctness and completeness of the transcripts, the researchers listened to the interview recordings while going through the transcripts from the transcription software. For the authors, the preparation of the data for the analysis was especially important to get an overview of the data and identify parts of particular interest (Easterby-Smith, 2015). Additionally, the sampling has been adjusted before further analysis. In particular, the researchers removed the pilot test (interview 1) and only used twelve out of the thirteen interviews (interview 2 to interview 13) for further analysis. More information regarding the interviews is portrayed in the Appendix (see Appendix 4: Overview of Interviews).

The approach to qualitative data is closely connected to the research philosophy, thus, certain assumptions regarding the analysis process can be made (Easterby-Smith, 2015). Linked to the relativist ontology, where multiple realities exist and the constructionist epistemology of this research, the analysis process rather develops cyclically in cooperation with the research

Managed by a Machine: Workers&apos; Job Crafting Abilities in the Case of Lieferando Riders in Germany