Hierarchical planning in service industries: Utilizing capacity flexibility

Hierarchical planning in service

industries: Utilizing capacity flexibility

Master’s Thesis Dual Degree in Operations Management

University of Groningen: Faculty of Economics and Business

Newcastle University: Newcastle University Business School

Author: Martijn Dusseljee

S1981633

m.d.dusseljee@student.rug.nl

Supervisor: dr. M.J. Land (University of Groningen)

Co-assessor: dr. G. Pang (Newcastle University)

1

Abstract

Capacity decisions are essential in successfully operating service companies. In order to make capacity decisions in a large-scale planning problem a hierarchical planning model could be used, which allows for making decisions over different time horizons. When studying literature on hierarchical planning models, it is discovered that utilizing capacity flexibility is important. Namely, capacity flexibility simplifies matching capacity with demand, which directly influences customer satisfaction in service industries. Hence, a case study is undertaken at a large-scale dynamic hierarchical planning process in a service industry in order to provide more insights on the complexities of utilizing capacity flexibility.

By means of systematically examining the characteristics of hierarchical planning models (time horizon, scope of the planning activity, and the detail level of information) on their influence on the utilization of capacity flexibility, propositions could be formulated that could be translated into guidelines for utilizing capacity flexibility. These guidelines indicate the effect of demand peaks on the required staffing; the necessity of different start times to enhance the match of capacity with demand; the utilization of capacity flexibility in the long term by separating days off and shift scheduling; and the importance of robustness of the staffing in matching capacity with demand. These guidelines are of general interest for researchers and managers. Finally, limitations of the research and directions for future research are outlined.

2

1 Introduction

Within service industries most companies experience the criticality of adjusting capacity to demand since a service is produced and consumed simultaneously. Maintaining capacity based on satisfying the maximum of customer demand results in excess capacity costs. Hence, managers prefer to organize capacity in a manner that allows flexibility in reacting to changing customer demand. As shown by Armistead & Clark (1994), success or failure in managing capacity to match demand is of direct influence on customer satisfaction. In addition, Özlük et al. (2010) claim that, due to the occurrence of low and peak demand periods, capacity decisions are essential in successfully operating service companies.

Due to unpredictable fluctuations in demand, it is important to be able to adjust capacity in the short term (Sasser, 1976). This is generally done by adjusting capacity in terms of employees and rarely in terms of equipment and facilities (Crandall & Markland, 1996). Adjustment of employee capacity could be organized in a planning process by means of a monolithic or a hierarchical planning model. A monolithic planning model is a single, rigid, and uniform module, whereas in a hierarchical planning model the decision-making process is partitioned into subproblems or modules with a time horizon (Hax & Meal, 1975). According to Lasserre & Mercé (1990), due to its size and uncertainty, a standard strategy to handle a large-scale planning problem is to break it down into smaller parts which are more manageable.

In large-scale planning problems, effectively utilizing human resources yields potential benefits of lowering costs and improving productivity (Hao et al., 2004). Moreover, it is important to match capacity with demand since it directly influences customer satisfaction (Armistead & Clark, 1994). According to Van der Veen (2013), planning employees to match demand as efficiently as possible requires flexibility in capacity, particularly when demand fluctuates. Additionally, according to Harvey et al. (1997), maximizing capacity flexibility would simplify matching capacity with demand, whereas Idris (2012) states that flexibility is seen as a capability that needs to be utilized rather than an outcome. However, whereas several papers have described hierarchical planning processes in order to effectively utilize resources, none address the complexities of utilizing capacity flexibility.

3 analyzing the hierarchical planning characteristics (scope, detail level, and time horizon) in the subproblems on the utilization of capacity flexibility. The insights derived from the analysis are used to formulate a number of propositions that could be translated into guidelines for utilizing capacity flexibility.

4

2 Theoretical background

In this section the theoretical context of this study is discussed. Firstly, an overview will be presented of existing theory on hierarchical planning in general, which includes demonstrating essential characteristics in hierarchical planning models and discussing the importance of utilizing capacity flexibility. Secondly, six hierarchical planning subproblems are distinguished and elaborated. Moreover, this section states how this theoretical context will be integrated in this study.

2.1 Hierarchical planning

The first paper that describes the challenges of hierarchical planning is Hax & Meal (1975). When optimizing a planning and scheduling system for a production process, they found difficulties in the management involvement at the various stages of the decision-making process. Hence, they decided to use a hierarchical system, which makes decisions in sequence. They state that a hierarchical system is effective if it helps in establishing sub objectives at the various organizational levels that would be consistent with the management responsibilities at each level. In addition, it should allow for corrections to these sub objectives by the managers at each level, and for coordination among the decisions made at each level (Hax & Meal, 1975).

In order to decompose the overall problem into subproblems, the extent to which various stages of the decision-making process are coupled need to be examined (Hax & Meal, 1975). If two sets of decisions are independent, they can be separated and structured in a hierarchy of decisions. The overall hierarchy of decisions is thus partitioned into sets of decisions, also referred to as subproblems or modules, with different characteristics including the scope of the planning activity, the detail level of the required information, and the time horizon of the decision (Hax & Meal, 1975). Generally, the lowest level of hierarchy has the narrowest scope of the plan, the lowest management involvement, the most detailed information needed, and the shortest time horizon. Furthermore, each level has its own objectives and constraints in which decisions have to be made (Hax & Meal, 1975).

5 levels should not have a single direction, while also Schneeweiss (1995) notes that integration among hierarchical levels is essential.

Within many manufacturing industries, hierarchical planning is considered as suitable and practical since it allows for dealing with a set of decisions that could be made over different time horizons (Albey & Bilge, 2011). Also in a variety of service industries, such as healthcare, railways, banking, and aviation, hierarchical planning is used in order to solve complex planning problems. According to Hao et al. (2004), planning problems in service industries could be defined as resource allocation problems concerned with the satisfaction of operational goals by utilizing human resources effectively, while subject to certain constraints. Hao et al. (2004) argue that effectively utilizing human resources yields potential benefits of lowering costs and improving productivity in organizations. Therefore, these resource allocation problems have received much attention from both researchers and practitioners. A few interesting references are listed by Hao et al. (2004) and Ernst et al. (2004).

Moreover, service organizations, such as airlines, have to deal with highly complex planning problems. Due to the fact that a substantial part of the employees has to be scheduled in advance, fluctuations in demand cause variations in productivity and efficiency (Hao et al., 2004). Planning employees to match demand as efficiently as possible requires a high flexibility, particularly when demand fluctuates (Van der Veen, 2013). Upton (1994) defines flexibility as the ability to change or react with little penalties in time, effort, cost or performance. In addition, the main purpose of flexibility is to deal with fluctuations in demand, which requires the ability of companies to adjust capacity with an optimal flexibility (Harvey et al., 1997; Aranda, 2003).

6 2.2 Hierarchical planning modules

As mentioned, according to Hax & Meal (1975), a hierarchical planning model partitions the planning process into subproblems or modules with various hierarchical levels and different time horizons. Literature research demonstrates that common characteristics of hierarchical planning processes in service industries are its largeness of scale and its dynamic environment with high fluctuations in demand where the match of capacity with demand directly influences customer satisfaction. Additionally, literature research shows five studies that describe planning processes with these characteristics and distinguish hierarchical planning subproblems: Ernst et al. (2004), Herbers (2005), Stolletz (2010), Kyngäs et al. (2012), and Van der Veen (2013).

The first, Ernst et al. (2004), presents a review of staff scheduling and rostering in which it decomposes industry-wide rostering problems into six separate modules. Secondly, in the paper of Herbers (2005) several optimization problems at different planning modules are tackled. Moreover, Stolletz (2010) analyzes the workforce planning of a company that provides check-in services for different airlcheck-ines where the planncheck-ing process was partitioned check-into four modules. In addition, Kyngäs et al. (2012) describes an effective method for optimizing large-scale staff rostering instances. And Van der Veen (2013) addresses personnel planning and scheduling challenges that explicitly address preferences and characteristics of individual employees.

Although these five studies describe planning processes with similar characteristics, the authors distinguish different planning subproblems. Additionally, different terminology is used when referring to the subproblems. Therefore, this paper valued all subproblems identified in the studies and selected a hierarchy with six subproblems that are frequently mentioned in these studies. The hierarchy of six subproblems comprises strategic workforce planning, workload prediction, shift planning, line of work construction, task assignment, and task replanning. According to Hax & Meal (1975) and Stolletz (2010), it is essential to indicate the objective, time horizon, and constraints for each subproblem in the planning process. Hence, emphasis is placed on these aspects by highlighting them in a table for each subproblem.

2.2.1 Strategic workforce planning

Objective Time horizon Constraints

Making decisions focusing on the optimal size and mix of a workforce in the long term

Two to six months - Cost minimization

- Match of capacity with demand

7 According to Anthony (1965), strategic planning comprises decisions on organizational objectives, on changes in the objectives, on the resources used to achieve the objectives, and on the policies that control the resources. Although strategic planning and long-range planning are often used interchangeably in literature, these terms are not similar. Whereas long-range planning is concerned with the whole future, strategic planning is concerned with decisions and plans that significantly affect the course and character of an organization. Each strategic plan is developed for one aspect of the organization rather than the whole (Zeff, 1966).

Thus, when discussing strategic planning regarding workforce, decisions regarding the policies on human resources and the usage of these resources need to be considered. According to Ernst et al. (2004), strategic workforce planning comprises decisions focusing on the optimal size and mix of a workforce in the long term, also referred to as staffing decisions (Van der Veen, 2013). According to Van der Veen (2013), staffing decisions consider skill-mix decisions, such as which skills are necessary or should be trained among the workforce, and contract-mix decisions, such as which qualification group needs additional employees.

When regarding the time horizon of strategic workforce planning, Zeff (1966) argues that, essentially, strategic planning is performed as the need arises as well as in anticipation of needs. In addition, Stolletz (2010) mentions that this task is usually executed two to six months in advance. Furthermore, constraints in this module are the cost minimization of the staffing and the match of capacity with demand. In other words, the optimal size of the workforce is the size which allows for achieving an optimal match of capacity with demand while saving costs as much as possible.

2.2.2 Workload prediction

Objective Time horizon Constraints

Determining how many staff is needed for each time slot over a certain planning period

Varying No

Table 2-2: Objective, time horizon, and constraints of workload prediction

8 each work task requires one employee, this representation allows for an easy analysis of workforce requirements.

Demand curves are a common abstraction of workloads to be covered in the subsequent shift planning subproblem in which workloads are usually determined in time slots of fifteen or thirty minutes, see e.g. Brusco et al. (1995). Furthermore, Herbers (2005) mentions the importance of avoiding unnecessary demand peaks since a smoother demand curve will provide a better basis for shift planning. This problem of smoothing labor requirements is called the workload leveling problem (Herbers, 2005). Furthermore, since this module does not include a planning activity, no constraints are present.

Figure 2-1: Workload of an airport over the day (Herbers, 2005)

2.2.3 Shift planning

Objective Time horizon Constraints

Determining what shifts need to be worked in combination with assigning a number of employees to each shift necessary to perform all tasks

Varying - Number of different

shift lengths

- Number of start times - Current staffing levels

Table 2-3: Objective, time horizon, and constraints of shift planning

9 planning based on workloads as given by a demand curve (demand-level shift planning) or shift planning based on work tasks (task-level shift planning).

If the planning is carried out weeks before actual operations, demand-level shift planning will offer sufficient degrees of detail (Herbers, 2005). Demand-level shift planning aims at covering workforce requirements per time slot (e.g. fifteen or thirty minutes) by a cost-minimal set of shifts (Herbers, 2005). In order to find the cost-minimal set of shifts, a representative set of workloads could be taken as input (Bard, 2004). The assignment of actual work tasks may then be deferred to shortly before operations (Dowling et al., 1997).

Additionally, demand-level shift planning often allows for additional scheduling flexibility. In time slots with high workload, understaffing is common since companies rely on external employment agencies that provide temporary employees to cover these peaks in workload (Herbers, 2005). An example of demand-level shift planning is shown in Figure 2-2. On the horizontal axis the time over the day is shown whereas the vertical axis shows the workload (dark blue area) and the shift plan (light blue area). As visualized, all workload is covered by four shifts.

Figure 2-2: Demand-level shift planning (Herbers, 2005)

10

Figure 2-3: Task-level shift planning (Herbers, 2005)

According to Van der Veen (2013), shifts should respect a set of constraints and are supposed to cover current staffing levels, expressing the demanded number of employees in each time period, as efficiently as possible. This set of constraints comprises the number of different shift lengths and the number of start times per day and per week (Sabar & Zenjari, 2015). In addition to the demanded number of employees, staffing levels may also specify required competence levels (Van der Veen, 2013).

2.2.4 Line of work construction

Objective Time horizon Constraints

Assigning each individual to a shift or a day off for all days in the freeze fence

Varying - Number of different shift lengths - Number of start times

- Current staffing levels

- Minimum and maximum bounds on the number of consecutive days and on days off - Minimum rest times and start time

differences between consecutive shifts - Minimum, maximum and average working

hours per week

- Number of weekends off per period

Table 2-4: Objective, time horizon, and constraints of line of work construction

11 Day-off scheduling concerns determining the placement of days off, whereas shift scheduling deals with the assignment of employees to shifts. Scheduling both days off and shifts simultaneously is sometimes labeled tour scheduling (Kyngäs et al., 2012). These scheduling activities could be performed in cyclic or non-cyclic rosters (Herbers, 2005). In a cyclic roster, as shown in Figure 2-4, employees that are classified similarly all perform similar lines of work, but all start with a different line. Ernst et al. (2004) states that this roster type is most applicable for situations with repeating demand patterns. In addition, Herbers (2005) indicates that cyclic rosters are always anonymous. According to Van der Veen (2013), a cyclic roster may be specified for either all or a subset of the employees of a department. Subsequently, the final activity when using cyclic rosters is assigning individual staff to the lines of work (Ernst et al., 2004).

Figure 2-4: Cyclic roster (Herbers, 2005)

12

Figure 2-5: Non-cyclic roster (Herbers, 2005)

According to Herbers (2005), next to the aforementioned constraints of shift planning, union and legal regulations and company policies impose a number of constraints that need to be respected. The most frequent constraints include: minimum and maximum bounds on the number of consecutive days and on days off; minimum rest times and start time differences between consecutive shifts; minimum, maximum and average working hours per week; and the number of weekends off per period, see Bechtold & Showalter (1987), Lau (1994) and Dowling et al. (1997).

2.2.5 Task assignment

Objective Time horizon Constraints

Assigning tasks to a set of scheduled employees

One or two days - Required competence - Time window

- Walking time - Break placement

Table 2-5: Objective, time horizon, and constraints of task assignment

13 2.2.6 Task replanning

Objective Time horizon Constraints

Optimizing the planning with as few changes to the assignments as necessary

Maximum: a day - Required competence - Time window

- Walking time - Break placement

- Restricted changes in shift times

Table 2-6: Objective, time horizon, and constraints of task replanning

According to Stolletz (2010), ad hoc changes in the task assignments are necessary due to changes in the workload or unforeseen employees’ no-shows during the day of operation. These changes are usually carried out manually (Kyngäs et al., 2012). Moreover, Stolletz (2010) argues that the goal of this replanning task is to fulfill the requirements with as few changes to the assignments as necessary. In this subproblem, the aforementioned constraints of task assignment have to be respected again. Moreover, Stolletz (2010) mentions the constraint of making restricted changes in shift times during replanning, for example in overtime.

2.3 Conclusion

When studying literature on hierarchical planning models, it is discovered that utilizing capacity flexibility is important. Although it does not receive much attention, its value could be derived from a few papers: increasing flexibility will simplify matching capacity with demand (Harvey et al., 1997); flexibility is seen as a capability that needs to be utilized rather than as an outcome (Idris, 2012); and planning employees to match demand as efficiently as possible requires flexibility in capacity, particularly when demand fluctuates (Van der Veen, 2013).

Since matching capacity with demand directly influences customer satisfaction in service industries (Armistead & Clark, 1994), more insights on the complexities of utilizing capacity flexibility are needed in order to make successful capacity decisions. In order to derive these insights, aforementioned characteristics of hierarchical planning models will be analyzed on their influence on utilizing capacity flexibility. These characteristics include the scope of the planning activity, the detail level of the required information, and the time horizon of the decision (Hax & Meal, 1975). Hence, the research question considered in this study is:

14

3 Research methods

In order to provide an answer to the research question, a case study has been undertaken to analyze the characteristics (scope, detail level, and time horizon) of decisions in hierarchical planning models on their influence on utilizing capacity flexibility. The choice is made to focus on a single, exploratory case as this allows going into more depth when deriving insights on the complexities of utilizing capacity flexibility. According to Yin (2009), this is an appropriate choice when studying a new phenomenon in practice. The following subsection briefly describes the selection of the case before data collection and analysis procedures are outlined.

3.1 Case selection

It was important to select a large-scale hierarchical planning process in a service industry. The selected case, a planning process for airport ground staff, was ideal since it faces high fluctuations in demand and the match of capacity with demand is directly influencing customer satisfaction. The hierarchical planning process in the case study is the planning process for ground staff of the Passenger Services department of KLM Royal Dutch Airlines (further referred to as KLM). KLM has 33,000 employees in total and has Schiphol Airport near the Dutch capital Amsterdam as its main operational base. The airport works on a continuous basis with operations on 24 hours on seven days a week.

The Passenger Services department comprises four different ground staff departments (Airside, Landside, Lounges, and Ticket Office) and few other small departments, such as Operational Support. This Operational Support department is responsible for the planning of ground staff for all four departments. The planning differs among the ground staff departments since each of those departments has different patterns for fluctuations in demand and the size of the staffing differs much. Hence, the choice is made to focus on the largest and most dynamic department: Airside. The current staffing of the Airside department comprises approximately 800 employees of whom most are Gate Agent or Service Agent. Hence, only these qualification groups will be addressed in this paper.

3.2 Data collection

15 develop this understanding, several informants were consulted through open-ended interviews. Among the informants were Airside Managers, Capacity Planners, Resource Planners, Staff Planners, Operational Planners, and the Operational Support Manager.

Secondly, literature research was used to gain more knowledge of which activities could be present in ground staff planning environments. Subsequently, this knowledge was used for critically questioning informants on their responsibilities in the planning process, again by means of open-ended interviews. Examples of questions that have clarified these responsibilities are: “Over which time horizon is the decision made?” and “What is the influence of the decision on the available capacity flexibility”. In addition, observations were made during planning meetings and on the work floor. Finally, the informants served as a source for the verification of findings from software systems. Throughout the four month period, there was extensive contact with the Manager of the Operational Support department.

3.3 Data analysis

16

4 Case analysis

In the case study the Airside planning problem was researched by analyzing the hierarchical planning characteristics (scope, detail level, and time horizon) in the subproblems on the utilization of capacity flexibility. Different from earlier findings in literature, this case analysis identified an additional module. Hence, seven modules in the planning process are elaborated in this section in sequence of hierarchy. These modules and their objectives are outlined in short below after which each module is elaborated in a subsection. Moreover, each subsection ends with the formulation of a few propositions that demonstrate the insights derived by the analysis.

The highest hierarchy in modules is strategic workforce planning, where attention is paid to achieving a good match of capacity with demand in the long term. Additionally, three subproblems (workload prediction, shift planning, and line of work construction) are solved over a similar time horizon since the predicted workload and the produced shift plans are elements for creating rosters over a planning period of a flight season. Subsequently, staff planning, which is the additional module, is assessing the match of capacity with demand in the short term and is concerned with hiring temporary employees. Finally, task assignment and task replanning ensure this match on the day of operation.

4.1 Strategic workforce planning

According to literature, strategic workforce planning concerns so-called staffing decisions which comprise contract-mix and skill-mix decisions. Firstly, contract-mix decisions are discussed. Similarly to the argument of Zeff (1966), at Airside contract-mix decisions are made as the need arises as well as in anticipation of needs. The main contract-mix decisions in the previous years were made in 2013, when a strategic plan was composed based on expected developments concerning growth in flight production until 2020. In this strategic plan, the fulfillment of Gate Agent tasks as well as Service Agent tasks were discussed.

17 work eight, six, five, or four hours. Necessary to remark is that an agent can only work one shift length, which is documented in its contract.

As addressed by Alp & Tan (2008), utilizing temporary resources contributes to utilizing capacity flexibility. This is demonstrated at the Airside department by the fact that temporary employees can be used for shifts of three hours, but also of four, five, or six hours. This shorter shift length is highly important for utilizing capacity flexibility since Airside faces high workload fluctuations during the day with peaks of approximately one and a half hour. Figure 4-1 shows the workload over the day for Service Agent tasks by grey columns on time slots of fifteen minutes. Two types of tasks could be distinguished at Airside: fixed tasks (dark grey area in the columns) that are required throughout the day and flexible tasks (light grey area) that are required when aircraft departures occur.

Figure 4-1: Workload for Service Agent tasks over the day (case study KLM)

18 However, it could be argued that, considering the incentives, different decisions could be made. Whereas the demand peaks are based on flexible tasks, which require short shift lengths, fixed tasks throughout the day require long shift lengths. Since fixed tasks should always be fulfilled and require long shift lengths, these could be fulfilled by permanent employees in order to maintain the identity of KLM and the loyalty of employees to the company. The flexible tasks, which require short shift lengths, can then be fulfilled by temporary employees in order to maximize the capacity flexibility. When implementing a similar strategy for Gate Agents by means of hiring temporary employees to fulfill flexible tasks, the identity of KLM and the employee loyalty will be maintained while capacity flexibility will be utilized and costs will be saved.

Secondly, skill-mix decisions are discussed, which are, similarly to contract-mix decisions, made as the need arises as well as in anticipation of needs. In order to match capacity with demand while saving costs as much as possible, capacity flexibility should be utilized by means of training an optimal number of employees for skills. The current staffing concerning skills at the Airside department is visualized in Figures 4-2 and 4-3, which show that the agents possess zero, one, two, three, or four skills. The figures indicate a disproportionate distribution since a third of the Gate Agents and almost a quarter of the Service Agents do not possess skills. Moreover, since both Gate Agents and Service Agents can possess thirteen skills, the figures indicate that the staffing concerning skills could be understaffed.

Figures 4-2 and 4-3: The current staffing concerning skills for Gate Agents and Service Agents (case study KLM)

19 This understaffing is demonstrated by Table 4-1, which shows the measurement of the skills utilization throughout a week. Two variables are measured, namely the total skill coverage and the skill coverage per task in total. Firstly, the total skill coverage means the percentage of the skills that are in possession of employees divided by the total of demanded skills. Secondly, the skill coverage per task in total means that for all tasks it is determined if the demanded skills were in possession of the employee that performed the task. As demonstrated, although 91% of the demanded skills were in possession of employees, still only 69% of the tasks were assigned to employees that were in possession of the demanded skills. In order to achieve full skill coverage per task in total, excess capacity in skills is needed. Since training employees implies costs, this excess capacity results in excess costs.

Date Total skill coverage Skill coverage per task in total

01-08-2015 91,26% 67,70% 02-08-2015 90,79% 68,71% 03-08-2015 91,40% 69,39% 04-08-2015 91,61% 70,94% 05-08-2015 90,40% 67,55% 06-08-2015 92,37% 71,38% 07-08-2015 91,52% 69,85% Week average 91,34% 69,36%

Table 4-1: Performance measurement of skills utilization (case study KLM)

Besides the understaffing, the low skills utilization could be clarified by the disproportionate distribution of skills since tasks can be allocated to an employee that has more skills than a task requires. The assignment of tasks to employees that are not in possession of the skill required for the task is not preferred since satisfactory performance cannot be ensured. Nevertheless, Airside allows these task assignments since most tasks that require skills are not much more difficult than other tasks. This implies that reconsidering whether these skills are necessary would be valuable.

The following propositions demonstrate the insights derived from the analysis:

- When tasks of a qualification group cannot be fulfilled by temporary employees, excess capacity is required in cases of workload fluctuations and employee absence.

20 - In order to achieve full skill coverage per task in total, excess capacity in skills is needed. - When assigning tasks to employees that are not in possession of the skill required for the

task is allowed, it should be reconsidered whether the skill is necessary.

4.2 Workload prediction

The workload prediction subproblem needs to be solved in order to have an accurate match of capacity with demand in the line of work construction subproblem. Since the planning period of the line of work construction comprises a flight season, which could be either seven months for the summer season or five months for the winter season, workload is predicted seven months prior to the start of a flight season, which means twelve or fourteen months in advance of the day of operation. While this time horizon seems considerably long term, the Airside department succeeds in providing an accurate workload prediction which is almost similar to the workload on the day of operation, fourteen months later.

This achievement is due to the fact that the workload is mainly based on the number of aircraft departures throughout the day. Since exact aircraft departure times are documented in long term contracts with all cooperating airlines, the workload prediction can be highly detailed in the long term. Moreover, the number of aircraft departures is not subject to many adjustments once it is scheduled. This is demonstrated by Figure 4-4, which shows the adjustments in the number of departures on a specific day in the summer season (July 6, 2015) throughout a few months. It appeared that the historic number of departures proved to be similar to the number scheduled five months earlier.

Figure 4-4: Predictions of aircraft departures for a specific day in the summer season (July 6, 2015) (case study KLM)

21 throughout the years. Nonetheless, passenger and baggage figures and engagement standards can change due to, for example, the usage of new information systems or adjustments in the policies regarding boarding. Thus, since passenger and baggage figures and engagement standards only differ marginally, differences in the production of aircraft departures result in differences in the workload. The three aspects that are considered in determining the workload are illustrated in Figure 4-5.

Figure 4-5: Aspects considered in generating tasks (Herbers, 2005)

In order to derive the workload for a whole planning period, there is made use of a representative week for each season. Similarly to Bard (2004), a representative set of demands is taken by means of simulating a day of operation for each day in the representative week. The simulation allocates the aircraft to the gates, which results in tasks based on the aforementioned aspects. All tasks can be plotted in a demand curve for each qualification group as shown in Figure 4-6, which shows the required workload for each time slot of five minutes over the day. Although it is known which skills should be present in order to fulfill all tasks, the required skills are not indicated since these will be considered when solving the task assignment subproblem.

22 Taking a representative set of demands implies that the workload for each day is almost identical over the planning period. Hence, the workload derived from the simulation on a day (Monday for example) in the week accounts for all similar days (all Mondays) in the flight season. This means that the production of aircraft departures within a season is expected to be stable, whereas it is expected to differ between seasons. Figure 4-7 shows the differences between seasons visualized in the workload for all agents (Gate Agents and Service Agents). The workload is shown over the day but as an average of all days in the week. The orange toppings on the columns represent the increase of workload in the summer compared to workload in the winter, whereas the white toppings mean a workload decrease. The many orange toppings, especially on the demand peaks, imply a significant increase in production in the summer compared to the winter.

Figure 4-7: Comparison of workload of the summer and the winter (case study KLM)

The differences in workload within the seasons are visualized in Figures 4-8 and 4-9, which compare the workload for all agents between two weeks within the summer season. Again, the workload is shown over the day as an average of all days in the week. The production of aircraft departures increases at the start of the summer season, stabilizes after approximately two months and then decreases slightly. Since the summer season comprises seven months, three periods of time are distinguished: the early summer, the middle summer, and the late summer.

23 orange toppings are visible that indicate a workload increase in the middle summer. Whereas differences are visible, the demand peaks are at similar time slots throughout the season. This implies that the differences cannot be regarded as significant. To summarize, Figures 4-7, 4-8, and 4-9 have demonstrated that it is appropriate to use a representative week for each planning period of a flight season.

Figure 4-8: Comparison of workload of the middle summer and the early summer (case study KLM)

Figure 4-9: Comparison of workload of the middle summer and the late summer (case study KLM)

24 demand peaks in order to save staffing costs. Although this workload leveling problem is not addressed in this paper, the influence of demand peaks will be discussed in the next subsection.

The following propositions demonstrate the insights derived from the analysis:

- In order to derive the workload for a whole planning period, there can be made use of a representative set of demands if the planning period shows similar workload patterns. - The choice for a planning period depends on workload fluctuations.

- When in possession of highly detailed information concerning the workload, capacity can be planned accurately in the long term.

4.3 Shift planning

The Airside department makes use of demand-level shift planning since shift planning is based on workloads as given by a demand curve. The demand curves of the previous module are used, which means that shifts are planned for each qualification and that the time horizon comprises twelve to fourteen months. Moreover, the detail level is not adjusted, which means that shifts are planned over time slots of five minutes. By planning the cost-minimal set of shifts, shift planning forms the input for creating rosters that match capacity with demand while saving costs as much as possible. Shifts are planned in a software system that indicates how much shifts of each shift length are necessary to cover the predicted workload while considering the following constraints: the different shift length constraint, the start time constraint, and current staffing levels.

25

Figure 4-10: Shift plan for Service Agent tasks (only permanent employees) over the day (case study KLM)

As indicated earlier, in order to maximize capacity flexibility, all flexible tasks could be fulfilled by temporary employees. Nonetheless, since the length of the demand peaks is shorter than the shortest shift length of a temporary employee, excess capacity cannot be avoided. This demonstrates that short demand peaks have a large effect on the required staffing. The excess capacity is shown in Figure 4-11 where shifts for permanent employees are complemented with shifts for temporary employees. By means of using different start times, visualized by the blue increments, capacity flexibility is utilized since it enables to plan a shift such that it matches capacity with demand. Additionally, by means of using different shift lengths, the shortest shifts can be allocated to the time slots where demand peaks occur.

Figure 4-11: Shift plan for Service Agent tasks (permanent and temporary employees) over the day (case study KLM)

26 Gate Agent tasks can only be fulfilled by permanently employed Gate Agents. However, this is not visualized in the figure since the performance of Service Agent tasks by Gate Agents is considered in the shift plan for permanently employed Service Agents.

Figure 4-12: Shift plan for Gate Agent tasks over the day (case study KLM)

As mentioned, shift planning forms the input for creating rosters that match capacity with demand while saving costs as much as possible. The shift plans for each qualification are partitioned into shift plans for each work time percentage since rosters are made for employees with the same qualification and work time percentage. An example of a shift plan for Service Agents that work in shifts of six hours is shown in Table 4-2. The left column shows the start time of the shift and the shift length (050S means a shift of six hours starting at 5:00 AM), whereas the top row shows the days of the week. For each day a number of employees is scheduled on a particular start time. The right column shows the total of employees for each start time whereas the lowest row shows the total of employees for each day.

27 The following propositions demonstrate the insights derived from the analysis:

- When the length of the shortest shift length exceeds the length of the demand peak, excess capacity cannot be avoided. Short demand peaks thus have a large effect on the required staffing.

- By means of using different shift lengths, the shortest shifts can be allocated to the time slots where demand peaks occur.

- By means of using different start times capacity flexibility is utilized since it enables to plan a shift such that it matches capacity with demand.

4.4 Line of work construction

According to literature, the module line of work construction comprises preference scheduling, days off scheduling, shift scheduling, and staff assignment (Ernst et al., 2004) and results in the assignment of each individual to a shift or a day off for all days in the time horizon (Stolletz, 2010). In order to create lines of work based on a shift plan, the Airside department schedules days off and shifts simultaneously. Moreover, it makes use of cyclic rosters to cover the weekly repeating workload that is based on the representative weeks. Similarly to the shift plans, the cyclic rosters are made for employees with the same qualification and work time percentage. During the scheduling of days off and shifts, individuals are not yet assigned to the rosters.

28

Figure 4-13: Cyclic roster for Service Agents that work 100% (full-time) (case study KLM)

29 capacity can be planned accurately in the long term. Thus, the choice for the time horizon of planning capacity should be based on the detail level of the information on workload.

However, also the preferences of employees should be considered when making this choice. Namely, Nurmi et al. (2011) argues that it is preferable or even requisite for employees to know their working days a long time ahead, so they can plan their leisure activities. In addition, whereas shift scheduling is indicating the exact working hours, Van der Veen (2013) states that it is not essential for employees to know the exact working hours in the long term. Rong (2010) and Kyngäs et al. (2012) mention a way of allowing employees to plan their free time more conveniently while not losing capacity flexibility, namely when scheduling days off comprises a longer time horizon than scheduling shifts. Hence, separating days off and shift scheduling could enable utilizing more capacity flexibility in the long term while preserving employee’s preferences.

Not performing days off and shift scheduling simultaneously in order to utilize capacity flexibility in the long term could, possibly, be organized in several ways. Three practical ways are highlighted in this paper: night shift scheduling, shift type scheduling, and start time range scheduling. Firstly, night shift scheduling implies that employees are assigned to working days, night shifts, and days off in the long term (Van der Veen, 2013). Since night shifts negatively affect the biorhythm of humans, it could be convenient to inform employees on their night shifts in the long term. Subsequently, the exact working hours can then be announced in the short term.

Secondly, shift type scheduling implies that employees are assigned to days off and working days including a shift type in the long term. By, for example, making use of five different shift types including night shifts, employees have an indication within which hours of the day their shift will start. Similarly to night shift scheduling, the exact working hours can be announced in the short term. And thirdly, start time range scheduling implies that employees are assigned to days off and working days including a start time range in the long term. By, for example, making use of twelve different start time ranges, employees have an indication within which two hours of the day their shift will start. This could be shown by indicating 050S-070S in a roster line, which means that the exact start time will be 5:00, 5:30, 6:00, 6:30, or 7:00 AM. This exact start time can then be announced in the short term.

30 order to shorten the time horizon of scheduling shifts. Moreover, it is demonstrated that when choosing the time horizon of planning capacity, it should be based on the detail level of the information on workload while considering the preferences of employees.

The following propositions demonstrate the insights derived from the analysis:

- The choice for the time horizon of planning capacity should be based on the detail level of the information on workload.

- The preferences of employees should be considered when making the choice for the time horizon of planning capacity.

- Separating days off and shift scheduling could enable utilizing more capacity flexibility in the long term while preserving employee’s preferences.

- Night shift scheduling, shift type scheduling, and start time range scheduling could be used in order to lengthen the time horizon of scheduling days off or in order to shorten the time horizon of scheduling shifts.

4.5 Staff planning

This module is not addressed by any of the five papers that were used to describe a large-scale dynamic hierarchical planning process in a service industry in Section 2.2. However, due to the extensive time horizon of the capacity planning by means of rosters, the capacity is subjected to shift adjustments due to planned holidays, planned days off, request concerning shift exchanges, absence due to training, etc. Therefore, this additional module with a time horizon of four weeks is necessary in order to match capacity with demand in the short term. The objectives of this module are assessing the match of capacity with demand and, subsequently, hiring temporary employees to complement unfulfilled tasks.

31

Figure 4-14: Staff planning for all agents (only permanent employees) over the day (case study KLM)

Figure 4-15: Staff planning for all agents (permanent and temporary employees) over the day (case study KLM)

32 temporary employees until a day in advance allows for anticipating on unexpected workload increases.

The possibilities for anticipating on both unexpected workload increases and decreases imply a high utilization of capacity flexibility. The current time horizon of four weeks for this module is appropriate since, due to the large percentage of temporary employees that fulfill Service Agent tasks at Airside, shortening the current time horizon of four weeks could result in difficulties for the external deployment agency to fulfill the shifts and a short term perspective on the staffing for the Airside department. In addition, lengthening the time horizon would cause redundant work for both parties.

As mentioned, this additional module is necessary since the time horizon of the line of work construction of twelve to fourteen months and the planning period of a flight season do not allow for assessing the match of capacity with demand in the short term. When the time horizon of the line of work construction module would comprises four or two weeks with a similar planning period, evidently, this match is assessed when planning capacity by means of rosters. Whereas a long planning period of line of work construction results in additional capacity planning by means of this module, it saves much capacity planning since only twice a year workload has to be predicted, shifts have to be planned, and rosters have to be created. Since a planning period depends on workload fluctuations and the time horizon is related to the planning period, the time horizon is indirectly related to workload fluctuations.

The following propositions demonstrate the insights derived from the analysis:

- When the line of work construction module comprises a long time horizon, an additional module with a short time horizon is necessary in order to match capacity with demand in the short term.

- A structure of hiring temporary employees by means of an increasing percentage of the total required number of employees allows for anticipating on unexpected workload decreases by avoiding excess capacity in terms of temporary employees on the day of operation. - The possibility of hiring additional temporary employees until a day in advance allows for

anticipating on unexpected workload increases.

33 4.6 Task assignment

As addressed by Stolletz (2010), updates to the flight schedule and the number of passengers per flight as well as information about changes in employee availability are known one or two days before the day of operation. On the basis of this information, one day in advance fixed and flexible tasks are assigned to employees by a software system based on the competences (qualifications and skills) of the employees. Whereas at Airside the match of capacity with demand concerning qualifications is considered twelve to fourteen months in advance during previous subproblems, the match of capacity with demand concerning skills is just addressed one day in advance during the task assignment subproblem.

This could be clarified based on two reasons. First, the robustness of the staffing within the Airside planning process is low. Robustness of the staffing could be described as the ability to maintain the staffing like it was planned (Harvey et al., 1997). The low robustness is especially caused by the enormous amount of shift exchanges that Airside has to deal with. When the robustness would be high, it would be possible to consider skills during the line of work construction module. However, an additional robustness condition should then be implemented, which comprises the allowance of exchanging shifts only with employees with similar competences. Secondly, since temporary employees fulfill approximately 50% of the Service Agent tasks and can be hired until one day in advance, the match of capacity with demand concerning skills could only be made when all capacity is known.

34

Figure 4-16: Staff planning for all agents over the day including visualization of break placement (case study KLM)

Figure 4-17: Task assignment in the software system (case study KLM)

The following propositions demonstrate the insights derived from the analysis:

- The choice for the time horizon of planning capacity depends on the robustness of the staffing.

35 - Utilizing capacity flexibility in task assignment is possible by means of placing breaks.

Breaks should be placed to time slots when workload is low.

4.7 Task replanning

As addressed by Stolletz (2010), ad hoc changes in the task assignments are necessary due to changes in the workload or unforeseen employees’ no-shows during the day of operation. As mentioned by Kyngäs et al. (2012), adjustments to the task assignments are usually carried out manually. Since task replanning is similar to task assignment, the aforementioned constraints of task assignment have to be respected again. Moreover, Stolletz (2010) mentions the constraint of making restricted changes in shift times during replanning. At Airside it is allowed to change the shift times by means of overtime for a maximum of one and a half hour. This possibility of overtime provides additional capacity flexibility that is utilized when on the day of operation additional workload emerges due to weather issues for example. Moreover, in this subproblem, again the break placement is a possibility for utilizing capacity flexibility.

The following proposition demonstrates the insights derived from the analysis:

36

5 Discussion and conclusions

This paper addressed the question: How can service industries utilize capacity flexibility in hierarchical planning processes? In order to answer this question, a case study is undertaken at a large-scale dynamic hierarchical planning process in a service industry. This case study provided insights on the influence of the characteristics (scope, detail level, and time horizon) of hierarchical planning models on the utilization of capacity flexibility. From these insights implications for theory and management were formulated as well as limitations and directions for future research.

5.1 Implications for theory and management

Whereas literature indicated the importance of capacity flexibility in order to match capacity with demand as efficiently as possible, this paper demonstrated how to utilize this capacity flexibility in a large-scale dynamic hierarchical planning process in a service industry. By systematically examining the subproblems of the planning process, a number of lessons were uncovered that that could be translated into guidelines for utilizing capacity flexibility in a large-scale dynamic hierarchical planning process in a service industry and are of general interest for researchers and managers alike:

1. When demand peaks comprise a shorter time length than the shortest shift length, demand peaks have a large effect on the required staffing. Excess capacity in periods with low demand is unavoidable when shift lengths are larger than demand peak time lengths. However, using as much short shift lengths as possible minimizes excess capacity. This was addressed by Van den Bergh et al. (2013), which refers to the redundant overlap in shifts due to long shift lengths. The case study affirmed this by demonstrating redundant overlap in shifts in Figure 4-12.

2. A large number of different start times enhances the match of capacity with demand. Ernst et al. (2004) mentions the necessity of varying start times in order to have a good match of capacity with demand. Figures 4-10, 4-11, and 4-12 demonstrated that the different start times simplify covering the workload well.

37 4. The choice for the time horizon of planning capacity should be based on the robustness of the staffing. When robustness is high, which means that the staffing can be maintained like it was planned (Harvey et al., 1997), capacity can be planned over a long time horizon. However, when robustness is low, the time horizon should be short in order to maintain the planning that is matching capacity with demand without being hindered by adjustments. In addition, this implies that when comprising a long time horizon and a low robustness, the robustness should be increased in order to have an efficient capacity planning.

5.2 Limitations and future research

A number of limitations to this research should be noted. Firstly, this analysis is based on only one large-scale dynamic hierarchical planning process in a service industry. This process is subject to high workload fluctuations throughout the day but low workload fluctuations between weeks. Future research could examine the guidelines for utilizing capacity flexibility by, preferably, analyzing an environment with more significant workload fluctuations between weeks. This could, possibly, lead to different insights regarding utilizing capacity flexibility. It is recommended to choose for an in-depth case study as this can provide the necessary in-depth insights.

38

6 References

Albey, E. & Bilge, Ü., 2011. A hierarchical approach to FMS planning and control with simulation-based capacity anticipation. International Journal of Production Research, 49(11), pp. 3319-3342. Alp, O. & Tan, T., 2008. Tactical capacity management under capacity flexibility. IIE Transactions, Volume 40, pp. 221-237.

Anthony, R. N., 1965. Planning and control systems: a framework for analysis. Cambridge: Harvard University Press.

Aranda, D. A., 2003. Service operations strategy, flexibility and performance in engineering consulting firms. International Journal of Operations & Production Management, 23(11), pp. 1401-1421.

Armistead, C. G. & Clark, G., 1994. The coping capacity management strategy in services and the influence on quality performance. International Journal of Service Industry Management, 5(2), pp. 5-22.

Bard, J., 2004. Staff scheduling in high volume service facilities with downgrading. IIE Transactions, 36(10), pp. 985-997.

Bechtold, S. E. & Showalter, M. J., 1987. A methodology for labor scheduling in a service operating system. Decision Sciences, Volume 18, pp. 89-107.

Bertrand, J. W. M., Wortmann, J. C. & Wijngaard, J., 1990. Production control: a structural and design

oriented approach. New York: Elsevier.

Bitran, G. R. & Hax, A. C., 1977. On the design of hierarchical production planning systems. Decision

Sciences, 8(1), pp. 28-55.

Brusco, M. J. et al., 1995. Improving personnel scheduling at airline stations. Operations Research, 43(5), pp. 741-751.

Crandall, R. E. & Markland, E. R., 1996. Demand management - Today's challenge for service industries. Production and Operations Management, 5(2), pp. 106-120.

Dowling, D., Krishnamoorthy, M., Mackenzie, H. & Sier, D., 1997. Staff rostering at a large international airport. Annals of Operations Research, Volume 72, pp. 125-147.

Edmondson, A. C. & McManus, S. E., 2007. Methodological fit in management field research. Academy

of Management Review, 32(4), pp. 1155-1179.

Ernst, A., Jiang, H., Krishnamoorthy, M. & Sier, D., 2004. Staff scheduling and rostering: a review of applications, methods and models. European Journal of Operational Research, Volume 153, pp. 3-27. Hao, G., Lai, K. K. & Tan, M., 2004. A neural network application in personnel scheduling. Annals of

Operations Research, Volume 128, pp. 65-90.

Harvey, J., Lefebvre, L. A. & Lefebvre, E., 1997. Flexibility and technology in services: a conceptual model. International Journal of Operations & Production Management, 17(1), pp. 29-45.

Hax, A. C. & Meal, H. C., 1975. Hierarchical integration of production planning and scheduling. In: M. Geisler, red. Studies in Management Sciences: vol. I, Logistics. New York: Elsevier, pp. 53-69.

39 Idris, F., 2012. Achieving flexibility in service operations using the rigid flexibility framework: an exploratory study. International Journal of Business and Society, 13(3), pp. 279-292.

Karlsson, C., 2009. Researching Operations Management. In: Researching Operations Management. New York: Routledge, pp. 6-41.

Kyngäs, N., Nurmi, K. & Kyngäs, J., 2012. Optimizing Large-Scale Staff Rostering Instances. Hong Kong, Proceedings of the International Multiconference of Engineers and Computer Scientists.

Lasserre, J. B. & Mercé, C., 1990. Robust hierarchical production planning under uncertainty. Annals

of Operations Research, Volume 26, pp. 73-87.

Lau, H. C., 1994. Manpower scheduling with shift change constraints. In: Proceedings of the 5th

International Symposium on Algorithms and Computation, LNCS 834. sl:Springer Verlag, pp. 616-624.

Nurmi, K., Kyngäs, J. & Post, G., 2011. Driver rostering for bus transit companies. Engineering Letters, 19(1), pp. 125-132.

Özlük, Ö., Elimam, A. A. & Interaminense, E., 2010. Optimum service capacity and demand management with price incentives. European Journal of Operational Research, Volume 204, pp. 316-327.

Rong, A., 2010. Monthly tour scheduling models with mixed skills considering weekend off requirements. Computers & Industrial Engineering, Volume 59, pp. 334-343.

Sabar, M. & Zenjari, A., 2015. Impact of Flexibility and Employees' Preferences on Shift Scheduling: a Simulation Approach. International Journal of Economics and Management Engineering, 5(1), pp. 1-8.

Sasser, W. E., 1976. Match supply and demand in service industries. Harvard Business Review, 54(6), pp. 133-140.

Schneeweiss, C., 1995. Hierarchical structures in organisations: a conceptual framework. European

Journal of Operational Research, 86(1), pp. 4-31.

Stolletz, R., 2010. Operational workforce planning for check-in counters at airports. Transportation

Research Part E, Volume 46, pp. 414-425.

Upton, D. M., 1994. The management of manufacturing flexibility. California Management Review, pp. 72-89.

Van den Bergh, J. et al., 2013. Personnel scheduling: A literature review. European Journal of

Operational Research, Volume 226, pp. 367-385.

Van der Veen, E., 2013. Personnel Preferences in Personnel Planning and Scheduling. Gouda: Egbert van der Veen.

Voss, C. A., Frohlich, M. & Tsikriktsis, N., 2002. Case research in operations management.

International Journal of Operations and Production Management, 22(2), pp. 195-219.

Yin, R. K., 2009. Case study research: design and methods. Los Angeles: SAGE.

Zeff, S. A., 1966. The Accounting Review. Planning and Control Systems: A Framework for Analysis