Analysis of real-world personnel scheduling problems

Patrick De Causmaecker, Peter Demeester, Greet Vanden Berghe, Bart Verbeke

KaHo Sint-Lieven, Information Technology
Gebroeders Desmetstraat 1, 9000 Gent, Belgium
Email :{Patrick.DeCausmaecker, Peter.Demeester, Greet.VandenBerghe, Bart.Verbeke}@kahosl.be
Tel: +3292658610, Fax: +3292256269

Abstract. This paper provides a classification of real-world personnel scheduling prob-

lems. Various companies have been examined and their way of scheduling is described ac-
cording to three independent parameters: time, personnel and duties. From this classifica-
tion, four different different types of the scheduling problems can be identified: perma-
nence, mobility, fluctuation and project centred planning. We point out that all of the stud-
ied cases exhibit additional properties and complexities when compared to similar cases,
previously published in literature. Additionally, the companies’ current solutions and fu-
ture demands for scheduling are discussed.

Keywords: Employee Timetabling, Distributed Timetabling Systems, Practical Experi-

ences

1 Introduction

Recently we started a research project to develop prototype software for real-world distributed
personnel scheduling [14]. In most of the studied cases, the central problem is to find (rightly
qualified) personnel for assistance in a department that is temporarily understaffed, due to e.g.
illness or leave of personnel. This kind of planning often leads to time-consuming weekly nego-
tiation meetings. At these meetings, each head of department informs the other heads about the
current attendance status of the personnel. This information helps the attendees to decide which
department to temporarily award extra personnel. In none of the studied cases, there was an
objective measuring instrument available to determine under- or overstaffing of a department.

Every company has its own needs and desires concerning staffing. This diversity on the one
hand and the need to be generic on the other hand is most conveniently met by a framework of
independent components. The aim is to classify the encountered planning problems and to gen-
erate a basic structure in which the different planning systems can be implemented.

A particular focus of the project is distribution. We found that the departmental structure within
companies larger than a critical size tends to consist of islands within which fine-grained opti-
misation is feasible. This isolation can and should not be complete. The departments need to
interact in order to avoid the limitations of local optimisation. We will consider an agent model
to handle this interaction. The ultimate aim of the project is thus to develop an agent-based
framework that can handle distributed personnel scheduling.

In this paper we report on the different personnel planning problems that we encountered within
the companies involved. Although the number of companies is not large enough to make quan-
titative statements, we use them as a starting point for the qualitative classification of the per-
sonnel scheduling problems.

Section 2 describes the objective of this research and briefly discusses other contributions to
classifying personnel scheduling problems. In Section 3, we define a set of independent pa-
rameters to describe personnel scheduling problems. We rank the studied companies according
to our personnel scheduling classification in Section 4. Starting from this characterisation that is
based on real-world problems, we will develop a basic model that can be applied to most situa-
tions.

2 Personnel Scheduling Problems

The objective is to classify personnel scheduling problems based on a small but representative
sample of Belgian companies. We visited eleven different companies and interviewed them
about their experience with personnel scheduling.

The investigated companies are:

- A small local police division (+/- 60 policemen)
- A hospital in a large city (+/- 700 employees)
- The largest Belgian health and safety board (+/- 500 employees)
- The largest Belgian home health care organisation (+/- 1100 employees for each of the five
counties)
- A small warehouse for books and post cards (+/- 20 employees)
- One of the largest Belgian employment agencies (with 40000 available temporary employ-
ees)
- Two of the largest software consultancy and development organisations (with about 1000
employees each in Belgium)
- Three smaller software consultancy and development companies (number of employees
ranging from 20 to 50). All of them deal with numerous problems in labour intensive com-
panies and they are very well aware of various types of personnel scheduling problem.

In order to make the sample more representative, we added fast food restaurants and call centres
to the list. Those are interesting because of certain specific short-time constraints. Including
these, we consider the variety of studied problems sufficient for the purpose of this research.

Ernst et al. present an elaborate review of applications, methods and models in personnel
scheduling in [7]. Ten different personnel scheduling domains are covered in the overview, of
which three are considered economically much more relevant than the rest: crew scheduling,
call centre scheduling and health care personnel scheduling. Crew scheduling differs from the
other personnel scheduling problems in that locations play a significant role in the problem
description. All aspects of transportation related problems have been widely examined over the
years [3,8] but we have decided not to go into crew scheduling in this research.
Also the call centre sector has been the subject of many research papers. A review by Grossman
et al. [9] describes the problems and solutions in call centers for outbond and inbound systems.
The health care sector is also well analysed throughout the literature. Especially nurse rostering
in hospitals is a complex problem for which automated decision support is still not common.
Burke et al. describe various aspects of the nurse rostering problem in a recent review paper [4].
They classify models and approaches and formulate challenging directions for future research.
A simplified case on police services, with specified shift sequences, is studied by Taylor and
Huxley [12]. It considers service types with 4 days of 10 hours or 5 days of 8 hours each fol-
lowed by 3 consecutive days off. However, real-world problems often are much more complex
(see Section 4.1). Laporte [11] describes constraints concerning rotating schedules, which dif-
fer from the constraints that we have identified in the emergency and security services (Section
4.1.1).
Tien and Kamiyama [13] present a common framework with which they classify approaches to
manpower scheduling. The framework consists of five stages: temporal manpower require-
ments, total manpower requirement, recreation blocks, recreation/work schedule and shift
schedule. Bradley and Martin [2] on the other hand distinguish three stages in the personnel
planning process: staffing (long-term process of determining how many personnel must be
employed by the organisation to provide a predetermined level of service), personnel schedul-
ing (the process of converting expected daily work force requirements into precise scheduling
assignments) and allocation (actual assignment of scheduled personnel to work sites). The plan-
ning horizon for the three stages ranges from years and months (in the case of staffing), over
weeks (in the personnel scheduling case) to days and hours (in the allocation phase). The latter
phase is the most relevant for this research.

Apart from the literature review papers, the published approaches and solutions are very domain
specific [1,6]. In this paper, we aim at a generic framework to support fast development of such
applications.

3 Scheduling space and constraints

Most scheduling problems have three independent variables, which we call dimensions: person-
nel (P), time (T) and the task or duty to be carried out (D). Personnel consist of all employees
involved in the scheduling problem. They are divided in groups, according to their capabilities
and their assigned tasks. A simple example is a group of employees and a manager. In more
complex cases there are many different groups and departments, each having a manager. The
time dimension can be subdivided according to the time period to be scheduled. In call centres
the smallest time concept, for example, relates to seconds and the scheduling period stretches
over a couple of minutes. In case holidays need to be assigned, the smallest time division can
range from one hour to one day, while the scheduling period often is one year. The task, or
required duty, can range from very specific and non-varying (answering a phone call), to more
complicated, (answering a phone call about a special promotion of a mobile operator in English,
French or Dutch).

In some cases, such as school timetabling involving different groups of students, classrooms,
lecturers and a time division in hours, there is a fourth dimension that we will call the location
(L). The groups of students correspond to the duties, the teachers are the personnel and the
classrooms are the locations. We refer to [5] for a survey of practical course timetabling. We
notice that in case of transportation problems the task and location are quite similar and there-
fore no fourth dimension is required. Each task is characterised by its starting time and location
and its finishing time and location [14]. In this classification, we will restrict the research to
three dimensional cases with P (personnel), T (tasks) and D (duties).

The three independent axes (P, T and D) define a three dimensional solution space. In this space
many areas that are defined by constraints, represent infeasible solutions. We distinguish two
types of constraints: hard and soft constraints. Hard constraints should not be violated. An ex-
ample of a hard constraint is that an employee can only answer one telephone call at a time.
Soft constraints often are economically less important, but if fulfilled, the staff’s satisfaction
level will increase. ‘Monica prefers not to work before 9 o’clock’ is an example of a soft con-
straint.

More complicated constraints are: after working three consecutive night shifts, a rest period of
at least one day is required; no night shift after a free day. Every constraint defines a relation
between the three axes P, T and D. Many personnel scheduling problems are over-constrained.
Infeasibility is very common but the organisations typically have ways of coping with that. The
framework that we develop should provide means to implement these - often implicit - coping
rules.

In the next section we will describe all the scheduling problems in terms of (P, T and D) and
their mutual relationships ((D,T), (P,T) and (P,D)), as presented in Table 1.

Table 1. Parameters for the classification of personnel scheduling problems

P Division of personnel into groups

D Division of different tasks
(D,T) Duty D has to be executed at time T
(P,T) Employee P has to work at time T
(P,D) Employee P has to perform duty D
Employee P has to perform duty D at time T
(P,D,T)
(the actual schedule)

In the examples we discuss the dynamics of the axes case by case.

4 Application and classification

We interviewed all the involved companies and organisations. In this paragraph, we briefly
discuss and classify the needs we encountered in this small but diverse sample of companies.

We classify the companies according to the personnel scheduling problem. Police services and
hospitals face the problem of permanent assistance. The number of personnel needed is defined
in advance (except for emergency cases). Their personnel scheduling problem is permanence
centred. On the other hand, there are companies that need to schedule duties involving transpor-
tation. These companies have a mobility centred planning. Note that we exclude pure transpor-
tation companies. Examples considered in this paper are health and safety board and home
health care. Next to permanence and mobility centred planning, we distinguish personnel plan-
ning based on fluctuating demand (fluctuation centred planning). Four different examples are
discussed: a warehouse or distribution centre, an employment agency, a call centre and a fast
food restaurant. Finally, we discuss companies that divide the work into projects, to which they
assign different groups of employees. They aim is to accomplish different projects with differ-
ent groups and therefore we call this scheduling problem `project centred planning’. Typical
examples are software development and consultancy.

4.1 Permanence centred planning

In permanence centred organisations, a minimum personnel coverage is required at all times.
Personnel work in shifts (typically three, but in some cases more).

4.1.1 Security and emergency services

Belgian police is divided in zones. A zone consists of a set of adjoining cities or villages. Every
zone has 3 major units: intervention, criminal investigation and traffic control. Only the inter-
vention unit is in continuous (24/24) stand-by. In case of severe problems such as multiple
collisions on the motorway, assistance can be obtained from a neighbouring zone. This does not
frequently occur.

Scheduling the personnel of the intervention unit is carried out in reference periods of 2 months.
In each period every individual is granted 4 weekends free. In a ‘book of wishes’ individuals
can express their weekend preferences. The planner will first attempt to schedule the free week-
ends. There are also numerous legal constraints on the workload, for example:
- maximum 10 hours per day,
- maximum 50 hours per week,
- on average 38 hours per week,
- at least 11 hours of rest are required between two shifts. Intervention unit staff are therefore
scheduled in cycles: on the first day of a cycle, they work an early shift; on the second day a
late shift and on the third day a night shift; they rest on the fourth day,
- there are mandatory training sessions like shooting practice, truncheon exercises; individuals
are also allowed to take promotion related courses,
- after 10 consecutive working days, minimum two days of rest are required,
- maximum 55 night shifts per year; maximum 10 nights per reference period (2 months).

Difficulties arise when an individual reports sick. Adaptive scheduling, which tends to consume
a lot of effort, is then required. In order to guarantee permanency, some of the statutory rules
sometimes need to be broken. As a consequence, most individuals work more than they are
supposed to and this brings along that the police staff has to be indulgent when people ask for a
leave (give and take).
Currently, the planning is still carried out manually on a large sheet of paper. The most impor-
tant goal is to minimise labour costs by reducing overtime, as is presented in the analysis (see
Table 2).
 Table 2. Parameters for personnel scheduling in security and emergency services
Civilians (mainly office workers), inspectors, head inspectors, police
P
commissioners, zone chief
Reception, intervention, criminal investigation, traffic control and office
D
work
Intervention is permanent, other tasks are mostly carried out during day
(D,T)
time
The department intervention works in different shifts, other departments
(P,T)
generally only during day time. There are many legal constraints.
Civilians only do reception, inspectors and head inspectors can carry
(P,D)
out most tasks, commissioners and zone chief have specialised duties.

4.1.2 Hospitals
In general, hospitals assign nurses to different departments at an organisational level. The num-
ber of nurses that is assigned to each department is based on the workload. Hospitals also deal
with a need for permanency. As in the police case, a lot of statutory regulations need to be met.
For example:
- maximum 10 hours work per day,
- maximum 9 hours for night shifts,
- minimum 36 hours rest per week,
- minimum 11 hours rest per day (with 2 exceptions per week on which 8 hours and 10 hours
are sufficient),
- if working time exceeds 5.5 hours, a rest period is required: half an hour or 15 minutes twice,
- maximum 5 consecutive night shifts,
- maximum 35 night shifts per 13 weeks,
- the average over 4 weeks should be at most 50 hours per week,
- the average over 13 weeks should be at most 45 hours per week.

Each department has to be staffed at all times by a head nurse and a number of regular nurses.
In case there is no head nurse available, a nurse with relevant experience can replace the head
nurse. Apart from the legal regulations, there are different types of leave (extra holidays for
nurses older than 45, compensation for overtime, (un)paid vacation for exceptional occasions,
social service leave, maternity leave, paternity leave,...).

In order to handle absence of personnel, the studied hospital uses a ‘mobile team’ that supports
the departments that are understaffed. At weekly meetings the staff determines which depart-
ment gets an extra nurse. The particular hospital that we studied ran an experiment to determine
the workload of every department in an objective manner, in order to discover the understaffed
departments. Due to the resulting mass of paperwork, the experiment failed.

The hospital uses an Excel spreadsheet to roster individual nurses. The sheets are huge (about
12 MB for one month and one deparment), and awkward to handle. The primary purpose of
these sheets is to organise holidays: a printed version of the sheet is put on a bulletin board that
nurses can consult if they want a day off. The main goal of the personnel rostering is to mini-
mise labour costs by reducing overtime.
 Table 3. Parameters for personnel scheduling in hospitals
A number of nurses is required in each department: junior
P
nurses, nurses and head nurses
Nursing concerns very diverse tasks (bathing, medical
D
care,… ) but details are not required in the schedules.
There is a permanent work schedule. The number of nurses
(D,T) needed is captured in weekly (or longer) periodical
schemes, which are divided in different shifts.
(P,T) There are many legal constraints on working time
Based on experience, qualification and individual agree-
(P,D)
ments, some nurses can replace a head nurse.

Notice that since the specific tasks are not to be scheduled (see D in Table 3), the problem is
rather two- than three-dimensional.

4.1.3 Common characteristics

In general, permanency is guaranteed by cyclical schedules, but the cycles are very often dis-
rupted due to absence of personnel. Shift moves are thus unavoidable.

4.2 Mobility centred planning

When mobility is involved, the main focus of personnel scheduling is on reorganising employ-
ees’ assignments during daytime. Delays cannot be avoided and they regularly occur. Conse-
quently, assignments depend on the progression of work and transport. This is particularly valid
in pure transport cases. In the health and safety board case the coach that is used to examine
people on-site, must be booked by one of the local divisions. In the case of the home health
care, the mobile nurses can independently plan their route provided that they visit all the pa-
tients.

4.2.1 Health and safety board

Health and safety organisations deal with medical examination, prevention and protection on
the work floor. Their clients are companies (without a medical board) for which they perform
medical inspections in-house. Other subjects handled by these boards are violence, stress,
safety,... investigations.

The sample organisation has approximately 500 employees (doctors, nurses, engineers, com-
puter scientists,...) divided in nine local departments, which serve about 40.000 companies.
Companies are visited a few days per year. In large companies, the health and safety organisa-
tion has a permanent office. There are eight coaches available, equipped with medical instru-
ments. Each company has special demands like available days and hours for examination,
whether they want the same team as the previous year or not, local facilities for examination,
average time needed for an examination,... All employees have to be examined and this is not
always easy to organise. The schedule changes regularly because of unforeseen delays or accel-
erations. Different departments of the health and safety board support each other when there is
local overpressure. In this case as well, there is need for an objective, agreed upon, measure-
ment of the workload.

The visited health and safety board currently uses a Java application for personnel scheduling.
There is no automated decision support involved. The goal is to facilitate the work of the plan-
ners, since at present approximately 10% of the people are involved in drawing up the personnel
schedule. Table 4 presents the classification of the health and safety board sector.

Table 4. Parameters for personnel scheduling in health and safety board

P Doctors, nurses, engineers, computer scientists
D Examinations, driving from A to B, reporting
(D,T) Continuous rescheduling of duties
(P,T) Normal day schedule for everybody, with holidays
(P,D) Continuous rescheduling and reassignment of duties

4.2.2 Home health care

Home health care organisations employ nurses who visit the patients at home. Most patients are
visited on a regular basis, ranging from three times a day to once a week. The central office
assigns the nurses to a department and to a district. A district can be a number of streets or a
small village. There are two nurses for each district and for every three adjoining districts, an
extra nurse is assigned. This extra nurse can replace one of the six regular nurses in case of
leave, illness…

Most nurses work four hours in the morning and four hours in the afternoon; other nurses work
from 6 pm till 10 pm on weekdays, and in the weekends during the morning and the evening.
The maximum allowed work time is nine hours per day and 50 hours per week. The central
office has a team to guarantee permanency at noon (when the regular nurses have a lunch
break), in the evening and at night. The pool consists of seven people, two which always have
to be stand-by. There is a mobile team and a pool of temporary nurses to assist departments in
case of understaffing. Neighbouring departments can step in after a phone call. Notice that
nurses and patients get attached to each other, implying that changes are in general not appreci-
ated. The schedule is generated by the central office. According to the contract, each nurse gets
a standard schedule that is repeated every two weeks.
The schedule is manually constructed with an Excel spreadsheet, but the organisation recently
started to automate the process. The goal is to reduce overtime and thus minimise labour costs,
as is presented in Table 5.

4.2.3 Common characteristics

The studied organisations are arranged in local departments that are reluctant to exchange per-
sonnel. Although we classified these two organisations as mobility centred, we experienced, to
our surprise, that mobility was not a major issue. In the the home health care organisation, the
nurses are allowed to define their own route.
 Table 5. Parameters for personnel scheduling in home health care
P Nurses, assistant nurses, dieticians and office workers
D Nursing, helping to get dressed,...
(D,T) Most tasks are a periodic, but there can be emergencies
Many different work schedules, many people work half-
(P,T)
time
Nurses are allowed to carry out most tasks, assistant nurses
(P,D)
have more restricted duties (washing and clothing people)

4.3 Fluctuation centred planning

The required number of personnel and/or the assignment of tasks depends on a continuously
changing demand. Most companies with fluctuating demands, maintain a history record for
demand forecasting.

4.3.1 Distribution/Warehouse
Individuals working in warehouses are qualified for a variety of tasks. Some warehouses organ-
ise job rotation: each individual has another task every month (for which he/she is qualified) in
order to maintain skills. The most common tasks are: handling incoming goods, putting them at
a location, picking up goods, and preparing for shipment. In order to guarantee a flexible as-
signment of tasks, at least three qualified people are required for each task. For some important
tasks, there are only three of these people available, which can hamper the scheduling. Certain
posts in the warehouse need to be permanently manned during working hours in order to guar-
antee a sound operation. The amount of work in the warehouse depends heavily on fluctuating
demand. By experience, the planner can predict more or less which periods of the year will be
busy. When the workload is too high, temporary employees are hired. About one week of train-
ing is required before new people can start on the job. Therefore, the planner needs to decide in
advance when to hire temporary employees.

The visited company sets the personnel schedule up manually. It makes use of a ‘polyvalency’
table that keeps track of each employee’s capabilities and experience. The goal of scheduling is
to reduce overtime and the number of temporary workers (see Table 6 for a classification).

Table 6. Parameters for personnel scheduling in warehouses

Staff is extended with temporary employees who need
P
training in advance
Goods coming in, classification, taking orders, picking,
D
goods going out, administration,...
Incoming goods and incoming orders determine the
(D,T) (re)organisation of tasks. Some duties are always required,
others are less urgent
The required number of employees is derived from previous
(P,T)
information about the same time of the year.
Some people can carry out many tasks, others are restricted
(P,D)
to certain activities.
4.3.2 Employment agency
Employment agencies try to find jobs for people and to find qualified people for jobs. The vis-
ited agency consists of about 80 national offices. These have about 40.000 temporary workers
available, of which, 10.000 are employed every day.
Companies offering jobs have selection criteria for the workers. This means that the agency has
to keep many parameters for each person. Examples are experience, seniority, transport facili-
ties, physical condition, driving license, availability,...

Jobs frequently require training in advance. Most companies that recruit work forces from the
agency, prefer to hire temporary workers who have worked for them before. Some companies
only want to work with a pool of people, specifically trained on the spot.
Each office tries to supply the local demand. When a local office is in need of more workers, it
contacts the neighbouring offices.

There is a need for a well-structured database with distributed facilities and planning tools to
optimise the delivery of work forces.

At the visited agency, the local offices work independently. The goal is to maximise the client
satisfaction (see Table 7 for a classification).

Table 7. Parameters for personnel scheduling in employment agencies

P Temporary workers that can be summoned
D Delivery of temporary workers
Demand for short or long periods and known a short or long
(D,T)
time in advance
(P,T) Available hours/days are different for each person
(P,D) Workers have personal demands and qualifications

4.3.3 Fast food

The scheduling problem of fast food centres is in the name: when a customer orders a meal,
delivery is expected within a very short time interval. Depending on the time and the day of the
week, there is a different number of customers and a different demand. Fast food centres main-
tain history records on an hourly basis. This allows them to calculate the average demand at
each moment and to start preparing meals in advance. The tasks are almost scheduled per min-
ute (e.g. prepare 20 hamburgers by 11h53), according to the predicted patterns. Personnel de-
mand is based upon this pattern.
The fast food companies have developed their own software programs for personnel scheduling.
The goal is to serve food as quickly as possible (maximising customer’s satisfaction) with a
minimum number of personnel. The characteristics are listed in Table 8.

4.3.4 Call centres

Call centres are hyperactive organisations where lots of phone calls have to be answered (e.g.
mobile phone operators, power companies,… ). Before a call reaches a human operator, an
automatic operator asks questions concerning the preferred language, the topic of the call,...
According to these data, the phone call is diverted to a human operator, who is able to answer it
in the appropriate language. Each employee has a computer screen that shows how many phone
calls are in the operator’s queue. Some call centres strive to answer every phone call within 3
minutes. All incoming calls are more or less equally divided among the employees and are
rescheduled according to the time it takes to answer a call. Some calls are labelled more urgent
since they come from privileged or impatient clients: there is an unwritten rule that impatient
clients need to be helped as soon as possible.

Table 8. Characteristics for personnel scheduling in fast food restaurants

P Employees and local manager
D Cleaning, cooking, serving
(D,T) Varying, with local peaks
The number of people employed depends on the history of
similar days and hours. This changes at every possible time
(P,T)
step (e.g. half an hour) using different regimes and working
schedules for everybody
(P,D) Most people can perform all the tasks

Important call centres have programs that deal with the problem described above. The goal is to
answer all the phone calls in time and to minimise the labour costs since call centres are not
profitable (see Table 9).

Table 9. Characteristics for personnel scheduling in call centres

P Operators and local manager
The task is to answer incoming phone calls. Each phone
D
call is distributed according to language, topic, ....
The number of phone calls is changing all the time. Each
(D,T)
phone call should be answered within 3 minutes
The number of people employed depends on the history of
similar days and hours. It changes at every possible time
(P,T)
interval (e.g. half an hour) using many different regimes
and working schedules
Each employee speaks specific languages and knows spe-
(P,D)
cific topics

4.3.5 Common characteristics

A general feature of these problems is the importance of accurate forecasting.

4.4 Project centred planning

A project can have different groups of employees assigned to it. Each group can work on differ-
ent projects at a time. According to the demand, the groups and assigned projects are resched-
uled.
4.4.1 Consulting and software development companies
In many consulting and software engineering companies, the activities are organised in projects.
The consulting company makes a proposal and if it gets accepted, the contract has to be exe-
cuted in a certain time window. The execution is normally carried out in a team consisting of a
project leader and some programmers. After visiting the client and interpreting the problem, the
development cycle (analysis, design, implementation,… ) starts. A provisional version of the
software is demonstrated to the customer. The appearance of extra information, the participation
of clients who did not attend the initial meetings, etc. may cause delays and pressure on other
projects (e.g. by transferring people from other projects). In one of the software development
companies studied, each project team contributes to 3 projects: a project that is in its starting
phase, one in its final state and another one that is fully running. If one of these projects is run-
ning out of time, it involves rescheduling the other 2 projects.

Currently, the planning is a manual process on a white board. The goal of this type of personnel
scheduling problem is to improve the project management (see Table 10 for a classification.)

Table 10. Characteristics for personnel scheduling in consultancy companies

Consult, build (develop) and run (support) for different
D
clients
In small groups, each group has clients to work for. A de-
P velopment group consist of a project leader and program-
mers
(D,T) For each client: consult, build and run in respective order
(P,T) No shifts, work by day (9-5)
(P,D) Specific qualifications for all the team members

4.4.2 Common characteristics

Project centred planning looks quite similar to fluctuation centred planning, but we opted to
make a distinction since the projects are executed in teams. Forecasting could be helpful but it
seems very difficult to realise because every project has its own needs (and a different grouping
of personnel).

5 Conclusions

An overview of the studied personnel planning for the end users is presented in Table 11. The
personnel scheduling in all the studied organisations and companies happens to be still a manual
process. Some use software tools (e.g. Excel in the hospital), however, none of the packages
support automatic planning. They are just a more or less elegant way to present the roster to the
employees. In some cases these tools are only used as a holiday organiser: employees can first
check the roster to find out if a leave is appropriate. The documents used in the hospital are so
huge that they are impractical: they contain so many rules and data that the limits of Excel’s
capabilities are reached.
In most organisations, much effort and time is spent on creating personnel rosters. Even more
time is spent when the roster has to be changed due to illness or holidays. We have anecdotal
evidence from hospitals and police divisions for this statement. In the case of the health and
safety board, the personnel scheduling involves almost 10% of the organisation’s personnel.

Another striking point is that people expect the manual planner to know everything: how the
company is organised, how certain tasks are executed, the capabilities of each employee, the
interrelationships between the employees, the contracts of each employee,… Thus, the general
responsibility puts considerable pressure on the planner.
When comparing the goals of every organisation, we detect some commonalities. One of the
largest costs in Belgian production is the labour cost. Most organisations therefore want to
minimise the labour costs by reducing overtime or by reducing the number of temporary work-
ers.

Some of the visited software companies pretend to sell personnel planning software. But, when
asked if they schedule their own personnel automatically the answer was negative. However,
many commercial tools exist and obviously some companies use them. Still, it was very diffi-
cult to obtain demonstrations and the few that we have seen were not fit for flexibly solving
different problems.

Although distribution was a particular focus of the study, it was difficult to detect in the visited
organisations. Exchange of personnel requires mutual empathy, which is not always obvious.
There is lack of objective measures to determine work (over)load in the departments. Another
reason is probably that distribution is not easy to organise in the case of manual scheduling.
Consequently, the planners are not aware of the possibilities that distribution of personnel can
offer.

During the visits, we identified the following important parameters for the personnel planning
problem:
- the number of available employees,
- the amount of overtime,
- fixed or fluctuating demands,
- the work contracts (full time, part time,...),
- the robustness of the planning,
- the possibility of employees to mutually swap tasks,
- the flexibility of the schedule.

In order to determine the criteria and quality measures of a good schedule, the following pa-
rameters can be helpful: the quality of the schedule for the individuals, the quality of the sched-
ule for the organisation, the difference between schedule and execution, the incorporation of
personnel preferences, the impact of a change, the cost of executing a schedule,...

From a company policy’s point of view, it is interesting to know how many people are required
to satisfy certain needs, how long in advance a schedule can be fixed and if the customers are
satisfied. There was a recurring request for a simulation tool in nearly every visited organisa-
tion. It should allow to experiment with the number of required personnel and with holiday
periods.
We conclude with one last remark. In general there is no wish for a fully automated planning.
Instead, support will be strongly appreciated. Most planners prefer evaluation tools that indicate
possibilities and that warn for mistakes such as double bookings. Furthermore, people always
prefer to be able to make schedules that ignore certain predefined constraints.

References

[1] U. Aickelin, K. Dowsland: Exploiting problem structure in a genetic algorithm approach

to a nurse rostering problem, Journal of Scheduling, Vol. 3, No. 3, (2000), 139-153
[2] D.J. Bradley, J.B. Martin: Continuous personnel scheduling algorithms: a literature re-
view, Journal of the Society for Health Systems, Vol. 2, No.2, (1990), 8-23
[3] H.-J. Bürckert, K. Fischer, G. Vierke: Transportation scheduling with holonic MAS - the
TeleTruck approach, Proceedings of the Third International Conference on Practical Ap-
plications of Intelligent Agents and Multiagents (PAAM'98), (1998), 577-590
[4] E.K. Burke, P. De Causmaecker, G. Vanden Berghe, H. Van Landeghem: The State of the
Art of Nurse Rostering, to appear in Journal of Scheduling, (2004)
[5] M.W. Carter, G. Laporte: Recent Developments in Practical Course Timetabling. In: E.
Burke, M. Carter (Eds.): Practice and Theory of Automated Timetabling II, LNCS 1408,
(1998), 3-19
[6] P. Cowling, G. Kendall, E. Soubeiga: A Hyperheuristic Approach to Scheduling a Sales
Summit, In E. Burke and W. Erben (Eds.), Selected Papers of the Third International Con-
ference on the Practice And Theory of Automated Timetabling PATAT2000, LNCS 2079,
(2001), 176–190
[7] A.T. Ernst, H. Jiang, M. Krishnamoorthy, D. Sier: Staff scheduling and rostering: A re-
view of applications, methods and models, European Journal of Operational Research 153
(2004), 3-27
[8] M. Gamache, F. Soumis: A method for optimally solving the rostering problem. In: G. Yu
(Ed.): OR in the Airline Indystry, Kluwer (1998), 124-157
[9] T. Grossman, D. Samuelson, S. Oh, T. Rohleder: Call centres, In: S. I. Gass and C. M.
Harris (Eds.): Encyclopedia of Operations Research and Management Science, 2nd Edition
(2001), 73-76
[10] J. Hueter, W. Swart: An integrated labor-management system for taco bell, Interfaces,
28(1) (1998), 75–91
[11] G. Laporte: The art and science of designing rotating schedules, Journal of the Operational
Research Society 50 (1999), 1011-1017
[12] P. Taylor, S. Huxley: A break from tradition for the San Francisco police: Patrol officer
scheduling using an optimization-based decision support system, Interfaces 19 (1) (1989),
4–24
[13] J.M. Tien, A. Kamiyama: On Manpower Scheduling Algorithms, SIAM Review, Vol. 24,
No. 3, (1982), 257-287
[14] http://ingenieur.kahosl.be/projecten/DINGO/
 Table 11. Schematic overview of personnel scheduling problems in companies and organisations

Type of Hard
Dimensions of Minimise/
personnel con- Distributed? Extra
the problem maximise?
planning straints
Different
Minimum divisions
Agents coverage (intervention Very rigid legal
Minimise
Police (+qualifications), + alternated regulations
overtime
shifts qualifica- with criminal (in theory)
tion investiga-
tion,… )
Minimum Individuals Minimise
coverage (personnel overtime
Ware- Employees
+ can carry out and tempo-
house (+qualifications)
qualifica- different rary em-
tion tasks) ployees
Minimum
Nurses occupa- Very rigid legal
Mobile teams Minimise
Hospital (+qualifications), tion + regulations
(departments) labour cost
shifts coverage (in theory)
of nurses
Availabil-
ity of the Different
Maximise
Employ- Temporary em- temporary pools of em-
customer
ment ployees employ- ployees from Large database
satisfaction
agency (+qualifications) ees + which to
qualifica- choose
tion
Satisfy the
Occupation wishes of the
qualifica- Reduce the
Health and of the coach customers to the
Personnel tion + work of
safety (teams are not highest possible
(+qualifications) minimum human
board fixed) + in extent + taking
coverage planners
space legal constraints
into account
Nurses qualifica-
Home Teams of Minimise
(+qualifications), tion +
health care nurses labour cost
shifts coverage