• No results found

Master Thesis, MSc Technology and Operations Management

N/A
N/A
Protected

Academic year: 2021

Share "Master Thesis, MSc Technology and Operations Management"

Copied!
64
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

The Effect of Workload on the Trade-Off Between Quality and

Quantity: Individual Differences in Task Performance

Master Thesis, MSc Technology and Operations Management

Sabine Waschull s.waschull@student.rug.nl Student number: s2002485 University of Groningen

Date: 22.06.2015

First supervisor: MSc N. Ziengs Second supervisor: Dr. W.M.C. van Wezel

(2)

1

Abstract

The understanding of human behaviour is a critical factor for the successful implementation of operations strategies, techniques and tools by reason that in most non-standardized tasks, workers directly affect the performance of a system. Despite this, many operations models make assumptions which ignore real system behavior and therefore implications of the models and theories are not always clear. As previous research suggests, individual characteristics considerably account for the variation in the quantity and quality of the workers output because workers respond to operating conditions, yet little is known over the basic processes underlying the trade-off decisions. With the help of a controlled experiment, this paper investigated the effect of workload on the trade-off decision between quality and quantity in which the worker has discretion over the output of their work. Thereby, it aims to find an explanation for the individual differences, more specifically, it explores the

moderating role of the regulatory focus on the dependence between workload and

quality/quantity performance. Findings of this research suggest that workers trade-off quality and quantity depending on the congestion of the system, mostly by decreasing the quality of their service content as workload, and consequently the queue, increases. Moreover,

individual reactions cause variations in service times, however differences in terms of approaching the trade-off are not explained by the individual regulatory focus.

(3)

2

Table of Contents

Preface ... 4

Introduction ... 5

Theoretical framework ... 8

State dependency in discretionary tasks ... 8

Individual differences in reactions to workload ... 13

Conceptual model ... 15 Methodology ... 16 Participants ... 16 Experimental design ... 16 Experimental task ... 18 Procedures ... 20 Measures ... 21 Regulatory focus ... 21

Quality and Quantity. ... 22

Control. ... 23

Data analysis ... 24

Data characteristics ... 24

Treatment checks ... 24

Validity checks ... 24

Descriptive data analysis ... 26

Quality ... 26

Quantity ... 29

Mean difference analysis ... 35

Hypothesis 1 ... 35

Hypothesis 2 ... 37

Post-hoc analysis ... 45

Extreme cases. ... 45

Value vs. Time Curve analysis ... 46

Discussion and Conclusion ... 48

Appendix ... 60

Appendix A - Regulatory focus questionnaire ... 61

Appendix B - Items collected ... 62

(4)

3

Figures

Figure 1 Value vs Time in a Discretionary task ... 10

Figure 2 Value vs Time in a Non-discretionary task ... 10

Figure 3 Conceptual Model ... 15

Figure 4 User Interface Baseline Round ... 19

Figure 5 User Interface Treatment Round ... 19

Figure 6 Individual Number of Options Fulfilled ... 26

Figure 7 High workload – Number of Options Fulfilled over Time ... 27

Figure 8 Medium Workload – Number of Options Fulfilled over Time... 27

Figure 9 Low Workload – Number of Options Fulfilled over Time ... 28

Figure 10 Processing Time Baseline Round ... 29

Figure 11 Individual Processing Time Treatmentr round ... 29

Figure 12 Individual Processing Time per Option ... 30

Figure 13 Number Tries to set Sliders correct per Customer ... 31

Figure 14 Number of Tries to set per Slider ... 31

Figure 15 Comparison Baseline with Low Workload Group ... 32

Figure 16 Comparison Baseline with Medium Workload Group ... 33

Figure 17 Comparison Baseline with High Workload Group ... 33

Figure 18 High workload – Processing Time per Customer Round ... 34

Figure 19 Medium workload– Processing Time per Customer Round ... 34

Figure 20 Low Workload– Processing time per Customer Round ... 34

Figure 21 Processing Time - High Workload ... 45

Figure 22 Number of Options Fulfilled – High Workload ... 46

Figure 23 Processing Time – Low Workload ... 46

Figure 24 Number of Options Fulfilled, Low Workload ... 46

Figure 25 Value vs. Time Curve ... 47

Figure 26 Value vs. Time Curve – Promotion Focus ... 47

Figure 27 Value vs. Time Curve – Prevention Focus ... 47

Figure 28 Accuracy Measure of Observations ... 63

Tables Table 1 Summary of Treatment Conditions ... 17

Table 2 Concept Measures ... 23

Table 3 Experimental Control Measures ... 23

Table 4 Distribution of Participants Across Groups ... 24

Table 5 Summary Statistical Analysis, Hypothesis 1 ... 37

(5)

4

Preface

This thesis is submitted in fulfillment of the requirements for a Master of Science degree at the University of Groningen. It comprises an experimental study on the effect of workload on the trade-off between quality and quantity in an operations context. It aims to shed light on the underlying processes that lead to a trade-off and to explain individual differences in making it. I chose for this project due to my interest in behavioural operations management, concerned with the human factors that affect the performance of operations systems. Surprisingly, a large part of operations research models and theories assume that humans are machines or that they would all react alike specific situations. Everyone is an individual and I aimed to show that individual differences are present and should not be neglected in a task in which the worker has discretion over the outcome. Moreover, I intended to explain the differences of

individuals in deciding to focus on quality or quantity in a specific context, more specifically under different workload scenarios.

(6)

5

Introduction

A critical aspect of the realization of operating systems with regards to the implementation and success of tools as well as strategies is the understanding of human behaviour (Bendoly, Donohue & Schultz, 2006). Due to the reason that in many operations environments, tasks do not follow a standardized performance criteria but require individual subjective judgment for positive completion (Hopp, Yuen & Iravani, 2007), it seems

reasonable that workers directly affect how the system works with the quality and quantity of work they deliver. At the same time, it has been shown that workers react to operating

conditions (Bendoly & Prietula, 2007; Bendoly & Hur, 2007), and this has direct

consequences on their task performances (Schultz, Schoenherr & Nembhard, 1998; Batt & Terwiesch, 2012). However, a large part of research in the field of queuing systems assumes that tasks are well defined and service rates are independent from the state of the system with regards to workload (KC & Terwiesch, 2009; Powell & Schultz, 2004). Many operations models and theories assume fully rational and homogenous workers and aim to establish design parameters and policies that maximize operations performance. Consequently, implications of these models for the real world are not always clear, increasing the need to incorporate human’s deviations from rationality (Gino & Pisano, 2007). Therefore, this paper aims to get insight into individual workers reactions to workload and effects on performance, more specifically the decision to trade-off quality and quantity when workload increases.

Various empirical studies have already shown that people are affected by operating conditions and that there are variations in reactions to service times (Shunko, Niederhoff & Rosokha, 2014; Schultz, Schoenherr & Nembhard, 2010). It was found that workers vary their processing time with the state of the system by speeding up or slowing down (KC &

(7)

6

2012). This is achieved by reducing the service content, consequently trading off quantity over quality.

Even though studies have shown the importance of human aspects in operations management, most research related to state dependent behavior of workers has investigated the effect of the congestion of the system on the speed of the worker (Schultz et al., 2010; Batt & Terwiesch, 2012), failing to address if, and how individuals trade-off between quality and quantity in an operations context. Up until now, basic processes underlying the trade-off decisions are still poorly understood and the nature of dependence of service times on workload is not clear. This increases the necessity to further explore this phenomenon. As earlier research suggests, reactions to workload will defer due to different individual

(8)

7

serve all customers without compromising on quality. This leads to different motivations driving people’s behaviour during the execution of their job.

Therefore, this research explores the trade-off between quality and quantity under increased workload. Moreover, it is tested if the degree to which workers are susceptible to workload depends on an individual’s promotion or prevention orientation. Thus, the

moderating role of regulatory focus is examined with respect to the effect of workload on the quality-quantity trade-off. The main research question is:

What effect does workload have on how individuals direct their effort to quality or quantity?

The importance of this study lies in the fact that most analytical operations management models fail to anticipate individual’s characteristics and preferences. This research therefore adds to a greater understanding by obtaining insight into the decision processes related to the trade-off so that organizations can develop policies and operations designs that mitigate the risk of lower performance. To explore this question, a controlled experiment is conducted in which the workers performance is examined and tested for differences between promotion and prevention focused people.

(9)

8

Theoretical framework

State dependency in discretionary tasks. A large body of literature in queueing

theory assumes that service times are independent from the state of the system (exogenous). This enables the development of easier and simpler analytical models that are applied to analyze and evaluate the performance of operations systems (Delasay, Ingolfsson & Schultz, 2014). The majority of models focus on systems with well defined, so-called

non-discretionary tasks, in which the worker has no discretion over the value that is added to the product/service (Hopp, Irvani, Yuen, 2007). Yet, limited research has been done on

discretionary tasks, in which the worker can decide on the service time and service content. Generally speaking, these models assume that under increased workload, service times are not affected. Another prominent assumption taken in many operations research models, among others, is that workers are statistically identical, which means that they act homogenous, translated into service time distributions and arrival rates. This applies to the classical queuing models, such as the popular Erland C and B models and other variations of it (Brown et al., 2005). As these models therefore often make very strong assumptions regarding the nature of the statistical process, for example the service time distribution or the arrival process of customers, these models are only used to provide answers in very specific situations. This is because the assumptions rarely hold up in the real world (Kingman, 1961) and consequently, implications are not always clear. Therefore, this research aims to investigate the state-dependency of worker’s service times and workers homogeneity.

(10)

9

depending on the state of the system. The major goal of this type of research therefore lies in the development of policies regulating a feasible service rate that balances the cost of waiting with the cost of faster service.

Next to that, recent modeling work has focused on the development of models that aim to give insight into discretionary task completion, in which the fulfilment of tasks depend on the workers subjective standards (Hopp et al., 2007; Stidham & Weber, 1989; Ata &

Shneorson, 2006; George & Harrison, 2000). Consequently, these models move closer to developing analytic models that incorporate subjective judgement of workers with regards to the quality and quantity of their job, for example how fast they work and consequently the service they deliver. However, these models still assume that workers reactions to increased workload are identical and therefore do not take variations into account. This has direct implications for an operations performance, as in a discretionary task individuals differences can lead to divergent performance outcomes.

In a discretionary task, completion criteria is not standardized and known beforehand, for example a call agent in a customer service center can decide upon the length of a call, thereby deciding on the quality level of the task. Hopp et al., (2007), who introduced

(11)

10

In a discretionary task it is at the workers discretion to maximize value gained through trading of quality and quantity, hence adjusting quality levels in response to system

congestion. For example, as the congestion of the system increases, less time is spend on one job (customer) due to a change in the service content, not the service time per customer. This is also referred to as task reduction. Hence, an increase in the service quality leads to an increase in service time per customer and the other way around. The idea behind is simple: by using a given service time results in greater benefits when more customers are present

(through savings in holding costs), which means that a faster service time should be selected when there is a higher workload (holding costs originate from the jobs in the system). The position of the worker on the value time curve indicates the quality/quantity trade-off decision made. Hopp et al.’s findings show that queue length can be reduced by lowering product quality (service content). When faced with different levels of workload that affect the congestion of the system, the worker responds, and consequently moves along the value time curve shown in Figure 1. In a low workload scenario, a worker spends more time on a task, and obtains value from additional service, plotted at the upper end of the graph. When the workload increases, the trade-off position of the worker moves down the value time curve,

Figure 1 Value vs Time in a Discretionary task

(12)

11

resulting in task reduction. This means spending less time on a job achieved through task reduction.

In order to empirically validate the effect of workload in discretionary systems as described above, and to ground the structure of the next hypothesis to follow, the following hypotheses are first tested.

Hypothesis 1 Levels of workload affect individual service times and work quality

1a High levels of workload lead to a faster service time

1b High levels of workload result in task reduction

State dependent models offer a limited applicability to a real-life production system as they disregard worker differences and require a number of implicit assumptions that are not necessarily satisfied in a work environment. According to Delsay et al. (2014), other

(13)

12

Up until now, a few empirical studies have shown the speed-up phenomenon and the resulting trade-off which originates from workload (congestion). Edie (1954) showed that processing times at a toll booth are decreasing with increasing traffic, resulting in a lower waiting time when traffic is higher. Schultz, Schoenherr & Nembhard (2010) examined data from a car radio testing operating and show that workers react to the speed of their coworkers and individual reactions differed widely, both in speed and reactions. Relevant findings showed that workers do not all either speed up or slow down and that there is variability in workers responses. Batt & Terwiesch (2012) studied the effect of workload in which the servers have discretion over the duration of the task as well as the service content. They find both evidence for slow down and speed up due to load, whereby both speed up (rushing) and task reduction are used to accelerate and consequently result in a lower work quality. KC & Terwiesch (2009) showed in their research about the impact of work load on service time and quality in two different health care services that the processing speed of workers is influenced by the congestion of the system, meaning that its level of care capacity is adaptive to higher levels of workload. They find that as the load on the system increases, workers speed up for a certain time period, this effect however is not sustainable and after a long period of overload the service rate decreases. Furthermore, Tan & Netessine (2014) show that service times distribution follow an inverted u-shape when workload increases.

(14)

13

time is due to individual characteristics and resulting differences, currently ignored in most operations research models.

Individual differences in reactions to workload

The quality/quantity trade-off has been of extensive interest in an array of different fields, aiming to answer the question as to why and when people are fast or accurate (Foerster, Higgins & Bianco, 2003). It is known that differences in directing effort to either quality or quantity, which result in individual strategies used to achieve a specific outcome, indicate differences in self-regulation (Higgins, 1997). Therefore, this paper proposes a self-regulatory account of behaviour in discretionary tasks. From the author’s perspective, individuals

regulatory focus, defined as a promotion or prevention focus includes motivational concerns that influences the decision to focus on quality or quantity.

(15)

14

come closer to their ideal self, while security and safety applies to prevention focused people to reach their ought self.

In recent studies, it has been shown that that an individual’s regulatory focus affects the decision between quality and quantity (accuracy/speed) (Foerster et al., 2003). Findings show that people with a promotion focus work faster and consequently less accurate on a simple drawing task compared to prevention focused people, who worked slower but with higher accuracy. Crowe & Higgins (1997) showed in their research that when individuals are charged with a task to come up with as many work alteratives as possible, promotion focused people generate more different alternatives than prevention focused people, who show to be more repetitive to avoid mistakes. This study is in line with the evidence of other authors (Liberman, Moden, Idson & Higgins, 2001; Shah, Higgins & Friedman, 1998) showing that a promotion focus leads to eagerness motivation, as opposed to a vigilance motivation for prevention focused people. Therefore, one would expect promotion focused people that emphasize a risky processing style and which are focused to acquire successes to work faster as this will most likely lead to faster advancements. Prevention focused people will remain accurate as speeding up increases the likelihood of making mistakes. As such, in a situation in which individuals are presented with the goal of maximizing value through working accurate or fast (quantity vs quality dilemma), prevention focused people would exhibit a stronger focus on quality, and promotion focused people on quantity when forced to trade-off due to increased workload.

(16)

15

jobs/customer, serves as a trigger for the individual to direct their effort in a specific direction (work fast & become less accurate or stay accurate & remain at a reasonable speed). The degree to which an individual is susceptible to and triggered by increased workload will be different for promotion and prevention focused people.

Hence, this research tests the following hypothesis to investigate the moderating effect of regulatory focus on the strength of the relationship between workload and the trade-off.

Hypothesis 2 The level of workload will interact with the regulatory focus of an individual to influence the trade-off between quality and quantity.

2a In a high workload scenario, promotion focused people work faster and deliver

lower quality than prevention focused people

2b In a high workload scenario, prevention focused people maintain a high level

quality and work slower than promotion focused people

Conceptual model

The conceptual model presented in Figure 3 is developed based on the hypotheses described. There are three main concepts in this research, namely the levels of workload (low, high), the regulatory focus (promotion focus, prevention focus) and the quality/quantity trade-off.

Figure 3 Conceptual Model

Workload

Regulatory focus

(17)

16

Methodology

Participants

The sample consisted of 63 undergraduate and graduate students (42 male; 21 female) of the University of Groningen with an average age of 22.74 (SD=3.53). The sample was multicultural, however the majority of participants (42) were Dutch, the rest from other European countries and Asia. The subjects were randomly assigned to treatment groups and there was no reward for participation.

Experimental design

Recent success of laboratory experimental methods (Croson & Donohue, 2006; Sterman, 1989) has shown that behavioural experiments are an entrenched research methods to investigate human behaviour (Bendoly, Donohue & Schultz, 2006). It allows to control situational factors to establish conditions in which real behaviour to stimuli can be observed to test or predict theories. With the aim to investigate relationships, manipulated controlled treatments in an experiment are a useful method to identify the effect on other independent variables (Wakker & John, 1998). Therefore, to identify if and to what extent individual trade-off responses to workload differ, a controlled experiment was chosen. A computer simulated work environment was created in which subjects individually process incoming customer requests, comparable to a cashier, bank teller or a doctor. The computer task mimics a discretionary task, in which subjects can actively choose the quality and quantity of their work.

(18)

17

effect on performance, a controlled arrival rate was developed based on each subjects individual completion rates determined during a baseline round. The rationale behind it that this research discards homogeneity and assumes that every worker has a distinct individual work style during the execution of the job, due to different skills, external expectations as well as a drive for self-accomplishment (Bendoly & Prietula, 2007). Therefore, to ensure that the treatment groups experience the same pressure created through the arrival rate of incoming customers, a baseline estimate of the processing time per customer was determined for each subject during the first round. Consequently, the impact of different workload conditions on the behaviour can be investigated in the second round (see Bendoly & Prietula (2007) for similar procedure).

Setting the arrival process parameters of incoming customers has been proven as a suitable method to impact the workload of a system and is therefore an important factor in the experimental design (Bendoly & Prietula, 2008; Shunko et al., 2014). A pilot study was used to define the arrival rates in the three groups, where three people were exposed to the different arrival rates respectively. Based on the resulting congestion of the system and the effect on the testers, three rates were selected that would lead to a low, medium and high workload. Subjects were exposed to an arrival rate that was either 50 %, 75 % or 100 % as fast as their estimated baseline rate. To be able to obtain a comprehensive picture on the effect of different workload scenarios, three equal groups were formed and subjects randomly assigned to one treatment conditions, see Table 1 for an overview. The treatment groups are referred to as high, medium, low workload, respectively.

Table 1 Summary of Treatment Conditions

Treatment condition Arrival rate Name

1 50 % of baseline High

2 75 % of baseline Medium

(19)

18

The experimental task was executed through a computerized interface and consisted of two rounds of 30 incoming customers respectively in a baseline round and a subsequent treatment round. During the baseline round, a new customer (job) arrived as soon as the subjects was available again. Fast working individuals sometimes had to wait for jobs, and there was no queue (graphical queue as used during the second round was not visible then). During the treatment round, customers entered according to the arrival rate in the treatment group, and subjects received feedback on their performance through the queue. During rounds, subjects were required to answer a series of predefined questionnaires, described in the next section.

Experimental task

(20)

19

Figure 4 User Interface Baseline Round

(21)

20

The values of each option ranged between €1 and €12 and the sliders moved in increments of €0,10. During the baseline round, the subject was required to set all the sliders. During the treatment round, the subject had a discretion to decide what of the five options to fulfill, and consequently could decide to continue the task until all parts are checked or continue with the next customer in line. To do so he clicked on the submit button with leaving some sliders untouched. Depending on the speed of the subject, incoming new customers waited in queue if the subject was still busy. This did not apply to the baseline round.

Procedures

(22)

21

questions were asked to reinforce the important aspects of the instructions. Subjects were then instructed to wait when they were presented with the instruction page for the 2nd round. They were then informed about the modification to the first round, which included that customer now would arrive at random times and wait in queue until the subject is available to serve. In addition, the subject had to decide which one of the options to fulfill by moving the sliders onto the exact position representing the value. Moreover, they were told that the aim is to maximize the total value obtained through fulfilling customers’ requests in the given time and thereby they were led to believe that their performance would be measured and tracked. The total value they could gain was equal to the sum of all sliders set correctly and this value could be influenced by changing how accurate and fast they worked, and the amount of sliders they set. Next, an illustrative example was provided, e.g. if three of the five sliders with value 6, 5 and 7 are moved correctly on the specified position, the total value equals 18. After the completion of both rounds, subjects filled out a series of screens containing control questions aimed to provide additional feedback and to enhance validity with regards to the experimental design. These questions can be seen in Appendix B. Finally, subjects were debriefed, thanked and then dismissed. During the task, the system recorded various objectives including the value of the target slider, the number of tries to set the slider, the arrival and departure time of the jobs per customer and finally, the average time to process a customer.

Measures

Regulatory focus. The strategic orientation regarding the regulatory focus of the

subjects were measured before the execution of the experiment with the help of the

(23)

22

To determine each subjects score, the answers of items attributed to each subscale were averaged. The mean prevention scores were deducted from the mean promotion scores, as done by Faddegon et al., (2009), Faddegon, Scheepers & Ellemers (2008) and Sassenberg, Jonas, Shah & Brazy (2007). This resulted in a single scale measure on the regulatory focus of the subject. A higher scores suggested that the subjects exhibited a stronger promotion focus, a lower score a stronger prevention focus. Both subscales exhibited good internal consistency with a Cronbach Alpha .841 (prevention scale) .877 (promotion scale). To facilitate the identification of behaviour differences for people with a different regulatory focus, subjects were divided by a median split into promotion and prevention focused people.

Quality and Quantity. The unit of analysis is defined as the individual, calculated by

(24)

23

Table 2 Concept Measures

Variable Measures

Quantity Individual processing time

Processing time per round (customer) Individual processing time per option Tries to set slider per customer

Tries per slider

Quality Number of options fulfilled (individual)

Number of options fulfilled (per customer)

Control. The measures to control for any side effect that can have an effect on the

performance and consequently enhance internal validity are listed in Table 3. It also presents the results for the reliability analysis of control measures. The performance orientation items measured to what extent the individual valued quality and quantity during the task. The items assessed can be seen in Appendix B.

Table 3 Experimental Control Measures

Measure Reliability Analysis (Cronbach Alpha)

Age, Gender -

Workload effect: Over-challenged items

(25)

24

Data analysis

Data characteristics

The pool of subjects that participated in the experiment included 63 post-secondary students, distributed evenly over the experimental condition (21 for Experiment 1, 20 for experiment 2 and 21 for experiment 3). Subjects were able to click through customers without setting the exact value, thereby disregarding the instructions of the experiment. To control for the possibility that subjects did not participate seriously or not fully understood the task, an accuracy measure was devised by calculating the number of sliders that were set correctly on the specified target value, excluding the sliders that were purposely not chosen. The subjects that showed an accuracy of lower than 80 % (4 sliders set incorrectly per customer) were filtered out from the dataset, (see Shunko et al. (2014) for similar procedure), see figure 28 in Appendix C. Moreover, two subjects did not set any sliders during the experiment, which was identified during the examination of data. The number of participants per treatment condition distinguished by their regulatory focus are summarized in Table 4.

Table 4 Distribution of Participants Across Groups

High workload Medium Workload Low Workload

Promotion focus 8 8 14

Prevention focus 11 11 7

Treatment checks

Several checks regarding the experimental treatments were conducted including test for randomization, learning effect and manipulation check to increase internal validity.

Validity checks. To investigate if the randomization of subjects to treatment groups

(26)

25

the same experimental conditions. There was no significant difference between treatment groups, F(2.58)=2.415, (p=.099), suggesting that randomization was successful.

The second validity check examined the learning effect on the subjects performance. Therefore, the 30 customer that every subject had to process successively were divided into three groups. A One-Way Anova was conducted to compare average processing times between customer number ‘1-10’, ‘11-20’ and ‘21-30’. There was a significant effect of customer number on average processing times (F(2.276)=10.019, p=.00). A post-hoc test showed that the processing time for subjects serving the first ten customers (M=23.74, SD=5.14) is higher than for subjects that processed customer number 11-20 (M=20.82,

SD=4.09, p=.001) and number 21-30 (M=20.23, SD=4.37, p=.00). Furthermore it showed that there was no difference in processing times between customer number 11-20 and number 21-30 (p=.483). These results suggest that the average processing time was affected by a learning effect between number 1-10 and 11-20, but not in the preceding customers. As the learning effect affects the time it takes a subjects to process a customer and can lead to wrong

conclusions, the first ten rounds were omitted from the calculation of the average processing time of the baseline round. This is done to ensure that the arrival rate that the subject is treated with is based on an accurate average reflecting the skill and task performance of the

individual.

(27)

26

Alpha of .523. A One-Way Anova test was selected to compare the mean ‘workload

difficulty’ scores between the treatment condition high, medium low workload and it shows a significant effect of ‘treatment condition’ on ‘workload score’, F(2.58)=3.29 , p=.047.

Descriptive data analysis

Quality. Figure 6 presents the individual number of options fulfilled per customer. It

can be seen that the number of sliders set by subjects reduces with increased workload as subjects in the low, medium, and high workload group set five, four, and three sliders, respectively. These findings support hypothesis 1b, stating that a high workload leads to task reduction. Moreover, in all three workload groups, promotion focused people set more sliders than prevention focused people, suggesting that they focus on quality to a greater extent. This ultimately also affects their average processing time, as more time is spend on fulfilling all wishes of the customer. These findings are contrary to the hypothesis 2a and 2b, stating that in a high workload scenario, promotion focused people work faster and compromise on quality, whereas prevention focused people deliver higher quality and therefore work slower.

Figure 6 Individual Number of Options Fulfilled

In addition to the average individual decisions presented above, it is interesting to examine how options fulfilled per customer behave over time to obtain insight if the workload and regulatory focus effect on performance becomes more pronounced over time. Therefore, Figure 7 to 9 plot the averages for all subjects over the 30 customers for the three workload

0 1 2 3 4 5

Low Medium High

(28)

27

groups and per regulatory focus orientation. In the high workload condition (Figure 7), the distribution of promotion and prevention focused people is relatively similar, the number of options fulfilled range from 2.5 to 3.75 options for promotion and from 2.2-3.5 for prevention focused people, supporting the findings of Figure 6 that promotion focused people fulfill slightly more options.

Figure 7 High workload – Number of Options Fulfilled over Time

Figure 8 plots the average options fulfilled per customer for the medium workload condition. Options fulfilled range between 3.57-4.71 for promotion and 3.08-4.38 for prevention focused people.

Figure 8 Medium Workload – Number of Options Fulfilled over Time

In the low workload group (Figure 9) promotion focused people almost consistently fulfill slightly more options on average per customer over all rounds than prevention oriented people, ranging between 4.36 and 5 and 4 and 5 respectively.

(29)

28

The graph is relatively stable over time, which suggests that when faced with a low workload, reactions to customers are steady.

Figure 9 Low Workload – Number of Options Fulfilled over Time

The findings presented related to quality suggest that under growing workload, the quality differences between promotion and prevention focused people decrease compared to a low workload scenario. All people regardless of their regulatory focus become less focused on fulfilling all options for a customer, consequently decreasing quality. Moreover, promotion focused people generally set more sliders, however when workload increases this effect diminishes as the differences between individuals grows smaller. This finding is in contrast to the hypothesis, stating that promotion focused people deliver lower quality than prevention focused people when workload increases. Finally, the pattern of number of options fulfilled is stable over time in all three workload scenarios for both promotion and prevention oriented people, and the effect does not become more pronounced over time.

(30)

29 0 5 10 15 20 25 Baseline P roc essi n g ti m e Promotion Prevention

Quantity. This section presents the performance measures related to service time.

Figure 10 presents the average processing times of individuals during the baseline round. It can be seen that there are no apparent differences between promotion and prevention focused people in terms of how fast they process a customer. If differences would have been more pronounced during the baseline round, it could have been an indication that workload mitigates the effect of the regulatory orientation of the individual if differences diminish during increased workload. By reason that all subjects were instructed to set all sliders, the only speed-up mechanism investigated during the baseline round was service time.

Figure 11 shows the individual average processing times during treatment round. Individuals facing high workload show the highest average processing time, followed by the medium workload group. The low workload group displays the lowest average processing time, almost half as long as the high workload group. This suggests that time spend per

customer decreases when workload increases, as subjects work faster by setting less sliders. In addition, no differences can be found between prevention and promotion focused people.

Figure 11 Individual Processing Time Treatmentr round

0 5 10 15 20 25

Low Medium High

Pr o ce ssi n g tim e Workload Promotion focus Prevention focus

(31)

30

In addition, the time people spend for fulfilling one option (setting of one slider) can be seen in Figure 12. It does not show large differences between workload groups, which suggest that people do not speed-up by working faster (rushing). Moreover, prevention focused people spend slightly more time per option compared to promotion focused people, however this amounts to only a few milliseconds.

Figure 12 Individual Processing Time per Option

Figure 13 shows the individual total average number of tries to set the chosen sliders (per customer). The number of tries are decreasing marginally with an increased workload, approximately five tries in the low and medium workload group, and four in the high

workload group. With increased workload, prevention focused people use less total attempts to set the sliders correctly. Promotion focused people are using equal attempts in the low and medium workload group, and moderately less in the high workload group. This is most likely due to the fact that subjects decided to set less sliders when workload increased. Differences between regulatory foci are relatively small and diminish with increased workload.

0 1 2 3 4 5

Low Medium High

(32)

31

Figure 13 Number Tries to set Sliders correct per Customer

Figure 14 presents the individual average number of tries to set one slider and it shows that there is small difference between workload groups. The low workload groups exhibited the lowest number of tries per slider, whereas the high workload groups show the highest number. An explanation for this could be that because subjects set less sliders when workload increases, they take more time to set it correctly, thereby clicking on the slider more often. Moreover, increased pressure presented trough the workload could lead subjects to make more mistakes and needing more attempts to set the slider correctly.

Figure 14 Number of Tries to set per Slider

The results described above support hypothesis 1a, stating that a higher workload results in a lower time spend per customer. Moreover, within the workload groups, no large differences can be seen between promotion and prevention focused people, indicating partly that hypothesis 2a and 2b, stating that promotion focused people work faster than prevention focused people as workload increases, and that prevention focused people work slower than promotion focused people as workload increases, are not supported by the findings.

0 1 2 3 4 5 6

Low Medium High

Tr ie s Workload Promotion focus Prevention focus 1.05 1.1 1.15 1.2 1.25 1.3 1.35

Low Medium High

(33)

32

There are no evident differences between the groups. In addition, in all three experimental conditions, prevention focused people spend slightly more time per option compared to promotion focused people, for example 4.2 over 4.7 seconds in low workload group and 4 and 3.76 seconds respectively in high workload group. In addition to the finding that prevention focused people set less sliders, it was furthermore found that they slightly spend more time on setting one.

Moreover, to evaluate if pressure presented from the workload changes the effects of the regulatory focus on the performance measure quantity, the mean individual processing times for the baseline round (no queue), and the treatment rounds (queue & workload groups) are compared. This comparison gives insight if workload negates the effect regulatory focus has on the outcome of the task itself. It can be seen (Figure 15) that processing times are not considerably different in the low workload condition in both baseline and treatment condition and for both promotion and prevention focused subjects.

Figure 15 Comparison Baseline with Low Workload Group

In the medium workload condition (Figure 16), there is a moderate difference between the treatment and baseline round, the time to process a customer decreases with workload. Prevention focused people processed customer slightly slower during the baseline round, however no difference can be seen in the treatment round.

(34)

33

Figure 16 Comparison Baseline with Medium Workload Group

Finally, the comparison of the high workload condition (Figure 17) yields the largest differences in processing times. Customers are processed more than twice as fast in the treatment round. This means that an increased workload, leading up to a queue, and results in a faster overall processing time and less time spend per customer. With regards to the

regulatory focus, the comparison of the baseline round with the treatment conditions shows no large differences. In the baseline condition, differences are greater compared to the treatment condition, however there is no apparent trend visible. This supports the findings related to the effect of workload on the speed performance of an individual as presented above as well as hypothesis 1a.

Figure 17 Comparison Baseline with High Workload Group

Figure 18 to 20 show the processing times per customer as an average of all subjects per customer 1-30. In the high workload group (Figure 18), the processing time per customer over time is relatively stable. There is no difference between the average processing time for

(35)

34

both promotion and prevention focused people as the distribution over time is relatively stable and similar. It shows the lowest processing time for all customers compared to medium and low workload. These findings support hypothesis 1a and 1b. Furthermore, for both medium and low workload, no significant differences in processing time between promotion and prevention focused people can be seen (Figure 19 and 20). Possible individual differences with regards to processing time between promotion and prevention focused do not change over time.

Figure 18 High workload – Processing Time per Customer Round

Figure 19 Medium workload– Processing Time per Customer Round

Figure 20 Low Workload– Processing time per Customer Round

(36)

35

Mean difference analysis

Hypothesis 1

Individual Processing time. A One-Way Anova test was selected to test if there are significant differences between the individual average processing times in the three treatment conditions. Normality was checked by using a Shapiro-Wilk test and review of the test (High workload: p=.952, Medium workload: p=.358; Low workload: p=.384 ) suggest that

normality can be assumed. Furthermore, Leven’s test of Homogeneity of Variances yielded that equal variances can be assumed (F(2)=.90, p=.412). Review of the results shows that there is a significant effect of workload on the individual processing time, F(2.58)=43.50, p=.00. A post-hoc test exhibited that the processing time for people in the high workload group (M=11.28, SD=2.25) is lower than for people in the medium workload scenario (M=17.08, SD=2.79, p=.00) and in the low workload scenario (M=1989, DS=3.59, p=.00). Furthermore, there is a significant difference between people in the medium workload and low workload group (p=.004). This is also the case for processing time as a function of time (per customer). First, normality was checked with a Shapiro-Wilk test (High: p=.015, Medium: p=.81, Low: p=.007). A Kruskal-Wallis test was chosen to test the differences between workload groups due to normality assumption. Review of the results show that there is a statistically significant difference between average processing times per customer

between the workload conditions (x2(2)=75.94, p=.00).

Time per option. Results of the Shapiro-Wilk test for the measure of processing time per option (High: p=.00; Medium: p=00; Low: p=.02) show that normality assumptions are not met. The Kruskal-Wallis-test was selected as a suitable to test the mean differences

(37)

36

suggesting that people do not work speed up through working faster (rushing) but only by setting less sliders when workload increases.

Number of tries per customer/slider. Normality assumptions were also not fulfilled for number of total tries per customer (High: p=.006, Medium: p=.038, Low: p=.00) and per slider (High: p=.00, Medium: p=.00, Low: p=.00) . A Kruskal-Wallis test shows that the number of tries to set the slider correctly per customer significantly differs per workload group (x2(2)=8.403, p=.015), with a mean rank of 21.21 for high, 31.24 for medium, and 36.38 for low workload. Moreover, a Kruskal-Wallis test shows that there is no significant difference in tries per slider between the workload groups (x2(2)=1.618, p=.445). Subjects number of tries to set the slider per customer decreases, but the number of tries to set one slider does not change, as workload increases. This can be due to the fact that with growing workload, subjects actively decide to set less sliders, however the accuracy with how they set the sliders does not change.

(38)

37

In addition, to evaluate if subjects actually set less sliders due to the amount of people waiting in the queue, it is tested if there are differences in the number of options fulfilled between long and small queues. Normality assumptions were not met (Long queue: p=.07, Short queue: p=.00), therefore a Mann-Whitney U-test was conducted. There is a significant effect of queue size on the number of options fulfilled (U=223.5, p=.001) with a mean rank of 629.5 for a long and 1140 for a short queue. Therefore, a larger queue pressures people to set less sliders. Table 5 summarizes the findings. Increased workload results in a trade-off , people spend less time per customer, thereby comprising quality.

Table 5 Summary Statistical Analysis, Hypothesis 1

Hypothesis Measures (Averages) Significance level

1a High levels of workload lead to a faster processing time

Individual processing time One-Way Anova, p=.00 Processing time per

customer

Kruskal-Wallis, p=.00 Processing time per option Kruskal-Wallis, p=.36 Tries per customer Kruskal-Wallis, p=.015 Tries per slider Kruskal-Wallis, p=.445

1b. High levels of workload results in task reduction

Number of options (individual)

Kruskal-Wallis, p=.00 Number of options (per

customer)

Kruskal-Wallis, p=.00

Hypothesis 2

(39)

38

SD=4.06), t(57)=-.71, p=.484. This suggests that differences in time taken to serve a customer are not explained by the regulatory focus during the baseline round.

Next, an Independent sample t-test was chosen to compare the mean differences between prevention and promotion focused people during the treatment round. Normality was tested in the section above and the observations are independent. The analysis showed that there are no significant differences in processing times between promotion (M=16.82, SD=4.30) and prevention focused people (M=15.58, SD=4.95), (t(57)=1.028, p=.308).

Furthermore, to test for the effect of workload on the regulatory focus, a One-Way Anova was selected to assess if workload affects the processing time of promotion and prevention

focused people in the workload groups, respectively. Normality assumption was checked by using a Shapiro-Wilk test and review of the test (High: p= .356, Medium: p=.71, Low: p=.09) show that normality can be assumed. Review of the results suggest that there is a significant effect of ‘workload’ on ‘processing time’ of promotion focused people, F(2.29)=37.87, p= 00. A post-hoc test exhibited that the processing time for promotion focused people in a high workload scenario (M=11.04, SD=2.16) is lower than for people in the medium workload scenario (M=17.33, SD=2.48, p=.00) and people in the low workload scenario (M=19.84, DS=2.25, p=.00). Processing time of the medium workload scenario is also significantly lower than the low workload scenario (p=.02). The same applies to prevention focused subjects. Assumptions were checked and normality can be assumed (High: p=.48, Medium: p=.51, Low: p=.996). The One-Way Anova test shows that there is a significant effect of workload (F(2.28)=6.251, p=.006) on a prevention focused processing time.

(40)

39

One-Way Anova was conducted, yielding that there are no significant differences in ‘differences of processing times’ of promotion and prevention focused people between the three workload conditions (F(2,89)=2.326, p=.104). This suggests, that differences in processing times between promotion and prevention focused people do not grow smaller as workload increases over time..

To examine the interaction effect of workload and regulatory focus on processing time, a Two-way Anova was conducted. Assumptions of normality and independence of observations are assumed as stated earlier. Furthermore, variances can be assumed to be equal, identified by conducting Levene’s test of homogeneity (F(5)=2.163, p=.072). There was no statistically significant interaction between the effects of regulatory focus and workload (F(2.58)=0.097, p=.91). The main effect analysis shows that workload has a direct effect on the processing time, meaning that a high workload conditions results in shorter processing time than low workload conditions (p=.00). Furthermore, there was no effect of regulatory focus on the processing time (p=.96). By reason that normality and homogeneity assumptions are not fulfilled, the interaction effect of number of options fulfilled and regulatory focus cannot be presented.

Time per option. Next, the effect of regulatory focus on the measure time per option is evaluated. First, it is tested if there is a significant difference in time per option per regulatory foci, regardless of the workload group. Normality assumptions are not fulfilled (Prevention: p=.036, Prevention: p=.011, therefore a Mann-Whitney U test was conducted, showing that over all groups, there is no significant difference in time per option spend between promotion and prevention focused people (U=404, p=.638). Next, the mean differences between

(41)

40

assumptions are fulfilled. A One-Way Anova test yields that there is not a significant effect of ‘workload’ on ‘time per option’ of promotion focused people, F(2.29)=1.521, p=.237. Next, another One-Way Anova test was conducted for the time per option of prevention focused people. Normality assumptions were fulfilled (High: p=.52, Medium: p=.001, Low: p=.695. The test showed that workload also does not have a significant effect on time per option (F(2,28)=.599, p=.578) of prevention focused people. The interaction effect was not tested because equal variances cannot be assumed (p=.005). Next, it was tested if there is a

significant difference in time per option between promotion and prevention focused people in the workload groups. Normality assumptions were fulfilled for the groups high workload (promotion: p=.20, prevention: p=.517) and low workload (promotion: p=.598, prevention: p=.695) and not fulfilled for the medium workload group medium workload (promotion: p=.037, prevention: p=.001). An Independent sample t-test was selected to test the difference in time per option between promotion and prevention focused people in the low and high workload group. There is no significant difference found in the low workload group (F(19)=.-.982, p=.338) and high workload group (F(17)=.-487, p=.633). A Mann-Whitney U-test was selected to compare the time per option in the low workload group. There is no significant difference between promotion and prevention focused people (U=25, p=.117). This suggests that there is no difference in time per option spend between promotion and prevention focused people in the workload group.

(42)

41

difference in number of tries to set the slider for prevention focused people (F(2.29)=3.11, p=.051. A post-hoc test exhibited that the number of tries for people in the high workload group (M=3.97, SD=1.31) is not lower than for people in the medium workload group (M=4.8, SD=.85, p=.15), however it shows to be significantly lower than prevention focused people in the low workload group (M=5.59, SD=1.81, p=.017). There is no significant difference between the medium and low workload group (p=.223). This findings show prevention oriented people show less attempts to set the slider correctly with increased workload, this does not apply to promotion focused people, suggesting that they are less inclined to trade-off under increased workload than prevention focused people. Moreover, it was tested if there is a significant difference between number of tries to set the slider correct (per customer) between regulatory foci within the workload groups. Normality assumptions were fulfilled for the groups high workload (Promotion: p=.035, Prevention: p=.129), medium workload (Promotion: p=.163, prevention: p=.734) and low workload (Promotion :p=.037, Prevention: p=.196), therefore an Independent sample t-test was selected. There was no significant difference in tries per customer between promotion and prevention focused people in the groups high workload (F(17)=.331, p=.745), medium workload (F(17)=1.28, p=.217) and low workload (F(19)=-.496, p=.626). This suggests that promotion and prevention oriented people behave equally when faced with equal levels of workload.

Number of options fulfilled. First, it was tested if there is a general difference in number of options fulfilled between promotion and prevention focused people. A Shapiro-Wilk test yields that normality assumptions cannot be assumed (Promotion: p=.00,

(43)

42

promotion focused people generally set more sliders than prevention focused people.

Furthermore, it was checked if the number of options fulfilled for prevention and promotion focused people is different per workload condition. Normality assumptions were not fulfilled for the distribution of data in the promotion focused group (High: p=.009, Medium: p=.473, Low: p=.048). Therefore, a Kruskal-Wallis was selected, yielding that there is a significant effect of workload on the number of options fulfilled by promotion focused people

(44)

43

(45)

44

Table 6 Summary Statistical Analysis Hypothesis 2

Round Independent

measure

Group Processing time

(individual) Processing time (per option) Number of tries (per customer) Number of options fulfilled (per customer)

Baseline round Regulatory focus Promotion/

prevention Independent t-test, p=.484 - - Treatment round

Regulatory focus Promotion/ prevention One-Way Anova, p=.308 Mann-Whitney, p=.638 Mann-Whitney, p=.129 Mann-Whitney, p=.058

Workload Prevention One-Way Anova,

p=.006 One-Way Anova, p=.578 Independent t-test, p=.051 One-Way Anova, p=.00

Workload Promotion One-Way Anova,

p=.00 One-Way Anova, p=.237 Kruskal-Wallis, p=.404 Kruskal-Wallis, p=.005 Regulatory focus X workload

Interaction Two-Way Anova, p=.91

- - -

Regulatory focus High workload - Independent t-test, p=.633

Independent t-test, p=.745

Mann-Whitney, p=.856

Regulatory focus Medium Workload - Mann-Whitney, p=.117 Independent t-test, p=.217 Independent t-test, p=.689

Regulatory focus Low workload - Independent-test,

p=.338

Independent t-test, p=.626

(46)

45

Post-hoc analysis

Extreme cases. In addition to the analysis presented above, a post-hoc analysis of the

extreme values was conducted with the aim to obtain a better insight into the behaviour of people with a strong prevention or promotion focus. By looking at the extreme values, defined as the people that scored very high or low on the regulatory focus difference score, can lead to further suggestions and insight into the underlying concept. A high score indicates a strong promotion focus and a low score a prevention focus. Figure 21 and 22 present the extreme values for the average individual processing times and number of options fulfilled for the high workload condition and figure 23-24 for the low workload condition. It can be seen that in a high workload condition, the extremes that exhibit a strong prevention focus have a higher processing time while approximately fulfilling the same number of options compared to the two extreme case with a strong promotion focus. This could be an indication that strongly promotion focused people not only work faster through setting less sliders but also due to rushing. Despite the fact that the difference is not large, it suggests that highly promotion focused people spend less time per customer than strongly prevention focused people when faced with a high workload. In the low workload conditions, speed differences are not large.

Figure 21 Processing Time - High Workload

(47)

46

Figure 22 Number of Options Fulfilled – High Workload

Figure 23 Processing Time – Low Workload

Figure 24 Number of Options Fulfilled, Low Workload

Value vs. Time Curve analysis

The model presented by Hopp et al. (2007) assumes that in a discretionary task, the value generated is a function of time as seen in figure 2. Figure 25 presents the respective value vs. time curve for the findings of this research. As can be seen, the distribution of data is approximately an increasing function of time, however there are differences between

individuals and decisions are not uniform for all people. This shows that individual differences in terms of trading off do matter.

(48)

47

Figure 25 Value vs. Time Curve

Figure 26 and Figure 27 present the value-time curve for promotion and prevention focused people respectively. The comparison of both figures shows differences in the distributions. The majority of promotion focused people are located at the upper end of the graph, whereas prevention focused people are more widely scattered, and no association can be seen. This suggests that the findings in the descriptive section, stating that prevention focused people work slightly faster than promotion focused people because they set less sliders, is not applicable to all subjects with a prevention focus by reason that the processing time and options fulfilled exhibit quite large differences between prevention focused people.

Figure 26 Value vs. Time Curve – Promotion Focus

Figure 27 Value vs. Time Curve – Prevention Focus

0 1 2 3 4 5 0 5 10 15 20 25 Op tion s fu lfi lle d )

Time (processing time)

0 1 2 3 4 5 0 5 10 15 20 25 Op tion s fu lfi lle d

Time (processing time)

0 1 2 3 4 5 0 5 10 15 20 25 Op tion s fu lfi lle d

(49)

48

Discussion and Conclusion

The findings of this research on the effect of workload on the quality/quantity trade-off and the moderating role of the regulatory focus provide empirical contributions to existing

literature. This study explores the effect of workload of a service system, in which the worker has discretion over the duration of task completions as well as the number of options fulfilled for the customer (service content). Prior analytical research states that service times are dependent on the state of the system, limited empirical research has provided evidence of a speed up and slowdown mechanism and the mechanism behind it. This study is partly in line with prior research on state-dependent behaviour of worker as it finds evidence of a speed-up and slow-down mechanism, which is depending on the congestion of the system. In addition, individual differences in valuing quality and quantity have been found, which are still ignored in most operations research models. Aiming to explain these differences used to obtain a specific performance outcome, the role of the regulatory focus theory is investigated.

The findings provide additional empirical support for the effect of workload on the trade-off between quality and quantity and they support hypothesis 1a and 1b. In a human-paced service system, workers are inclined to speed-up if it is crowded (Chan et al., 2011; Batt & Terwiesch, 2012), compromising on quality through task-reduction or rushing. This paper shows similar findings, workers are speeding up by spending less time on processing one customer. This findings has been shown to be related to the size of the queue, as a trigger to speed-up. However, this is only achieved by reducing the number of tasks fulfilled for the customer and not by actually working faster. In all workload conditions, people

(50)

49

provided. These findings suggest when workload increases, people process customers faster, however when presented with the possibility, people choose to deliver a lower service content as opposed to actually working faster (rushing). This is not in line with the findings of KC & Terwiesch (2009) and Batt & Terwiesch (2012) that show that speeding up also occurs through rushing. Therefore, future research could expand on this phenomenon and identify under what specific circumstances in organizational environments or due to what individual differences people either speed up through rushing or decreased service content.

Decreased service content could potentially result in additional cost factors and offset the benefits gained through increased speed. The implication from a managerial standpoint is that required quality levels must be clearly defined and be incorporated into operational policies to ensure that workers do not reach a certain quality level when the congestion is high. Therefore, service policies must be developed to ensure that reduced service quality through task reduction is mitigated. Due to the limitations of the experimental design, it was not possible to identify if the trading off through reducing the service content resulted in a better or worse performance with regards to profit accumulated. It would be interesting to find out if subjects that reduced service content actually accumulated a higher profit within a defined time period as opposed to people that did not. Potential cost factors through lost future sales and waiting customers could be offset by the benefits gained through delivering less service per customer.

Moreover, this research shows the presence of individual differences with regards to trading off quality and quantity. However, the dependence of service time and content on workload cannot be explained by the strategic orientation related to the regulatory focus of an

(51)

50

resulting in a significant effect of workload on the processing time (quantity), the amount of sliders set per customer (quality). It should also be noticed that when looking at the

distribution of data, promotion focused people consistently fulfilled more options (tasks) for the customer in all workload conditions, and the difference in the amount of sliders set (disregarding workload) between promotion and prevention oriented people is marginally significant (p=.058), however no significant difference can be found looking at the distribution of groups within the workload groups, e.g. difference of promotion and prevention focused people in the high workload group (p=.856). These findings partly

contradicts the stated hypothesis that promotion focused are focused less on quality compared to prevention focused people and the findings of Foerster et al. (2003), in which they

generally show the quality/quantity preference for promotion and prevention focused people. An explanation for this finding could lie in limitations of the task definition of the

(52)

51

In addition, as this research could not validate the findings of Foerster et al. (2003) that promotion focused people work faster and prevention focused people more accurate, thereby compromising on the other trade-off dimension, additional research must be conducted to identify if regulatory focus theory actually leads to differences in worker behaviour in a discretionary task environment. As promotion and prevention focus are independent

dimensions, and people are both promotion and prevention focused, resulting in one person having high levels in both foci, one or neither (Higgins, 1997), this can also affect the direction of effort in terms of quality and quantity depending on the person. Therefore,

prevention and promotion focus should be assessed separately to identify, what factors lead to a preference for quality over quantity in a high workload scenario.

As shown in figure 25 individual differences in terms of trading off are present, resulting in different positions off the value-time curve. By reason that goal setting has been mentioned as a limitation of this research that could be an explanation as to why no

significant differences were found, additional research can focus on goal setting (maximizing value) and its effect on the direction of effort of promotion and prevention focused people. Additional research should test if goal setting diminishes the effect that the regulatory has on the direction of effort. A possible explanation for this effect could be that a possible

achievement of a goal can be interpreted by a promotion focused person as an achievement, and a prevention of failure by a prevention focused person. From a managerial standpoint this knowledge is valuable because goal setting could most likely mitigate a possible effect of the regulatory focus on task performance.

(53)

52

waiting in the queue, either consciously or unconsciously, and aim to decrease the congestion. However, the reactions to stimuli presented by increased workload are not equal, resulting in variations of the trade-off position on the value time curve as presented in figure 24 as well as 25 and 26. Hence, there are individual differences on how people trade-off quality and quantity, and consequently different perceptions on how to maximize value. The implication from a managerial standpoint is that in order to decrease costs, in the form of holding costs or lost customers goodwill, and to maximize value through increased throughput and additional service given, employees have to be aware on the individual differences, and management has to develop policies to guide employees to make comparable decisions in terms of their

position on the value-time curve.

Finally, this research has been conducted under classroom conditions, in which less experimental control could be executed and subjects were able to influence each other. This could have had an effect on the outcome of the results, as subjects were partly able to see other computer screens. Being able to see that other subjects already finished, can urge people to speed up or copy how other people worked. Also the mood of other people can be an influencing factor and affect how people execute the task. As such, control measures such as mood, motivation or task satisfaction were not included in the analysis of the effect of

regulatory foci and performance, as no differences between reactions indicated that this effect is mitigated by other factors. A second critical point includes the external validity, more specifically the usage of a student sample for data collection. However, this study aimed to establish causal relationships and therefore there is no reason to anticipate that the chosen sample would react differently to stimuli and tasks than others (Stevens, 2011).

(54)

53

results in a growing queue of incoming customers, people are triggered to process customers faster in order to reduce congestion. To do so, task reduction was used as a speed-up

mechanism, thereby lowering service content to reduce the congestion of the system. In this study, subjects did not work faster but fulfilled less options to the customer. In addition, individual differences in direct effort to quality or quantity were identified, however the prediction that individual differences to trade-off are a result of the regulatory focus of an individual has not proven to be true. Contrary to empirical findings in literature, no differences were found between promotion and prevention focused people and their preference for valuing quality or quantity when workload increases as well as when not presented with a queue during the baseline round.

This work contributes to explaining the dependence of workload and service time, which is relevant for the understanding on how people trade-off quality and quantity. By expanding the knowledge base on how individual differences affect task performances and system dynamics, operations research models and theories can become more relevant from a real system

perspective. Future research on behavioural factors is a prerequisite for this.

Referenties

GERELATEERDE DOCUMENTEN

this demand properly due to their highly innovative solutions at high prices. We currently see a trend that MDMs start investing in “reverse innovation” and by doing so take a

Only one other empirical investigation could be located that directly considered the link between the ethnic minority diversity of the board and financial performance

Thus, this offers opportunities for academics to revise the self-monitoring scale and study self-monitoring again in the context of engagement on social media, to

In order to achieve this, organisations should listen to the ideas of employees regarding training that is needed in order to develop skills and knowledge (bottom-up),

Thus, during this research, we would like to examine the influence of information and communication technology on the relationship between national culture and

In addition to the reward power, no statement is made of the possibility of other types of powers that may influence the governance mechanisms since coercive power or non-mediated

The buyers have different levels of information exchange.The buyer A and the supplier developed an open information exchange atmosphere in the interaction. The

100% outsourcing compared to the directing model: completely outsourcing all logistical activities has been rated higher on all core values compared to the positive ratings of the