Journal of the Operationalllesaan:h Society (199&) 47, 358-368 © 1996 Operational Research Society Ltd. All rights reserved.

d. 0160·5682/96 $12.00 ~

Case Study of Manpower Planning for

Clerical Operations
G. I. MOULD
School of Management, University of Stirling

This paper presents a case study of short term manpower planning in an agency which processed con-
tract clerical work. The problem originated when the agency was presented with an unexpectedly large
batch of work and a backlog quickly developed. Management took the obvious action of increasing staff
numbers but this had little effect on the throughput of work. The situation deteriorated and a rapid
response was required from management. This meant that only models that were readily accessible for
use by management could be considered. A decision support approach was adopted and a spreadsheet
based model developed. This was effectively a time tased simulation of the situation, but it gave manage-
ment the flexibility to evaluate alternative solutions, unaided. The model proved to be a highly successful
tool. It demonstrated that simply increasing temporary staff or working overtime were insufficient to
remove the backlog of work within a year. This encouraged management to search for alternative
solutions.

Key words: decision support systems, manpower planning, simulation, spreadsheet models

INTRODUCTION

Good management practice for planning and controlling work, requires a long term equilibrium
between the input of new work and thoughput of completed work. This can be achieved with the
aid of reliable forecasts of new work, evaluations of the current work outstanding and assessments
of achievable throughput based on projections of staff numbers and likely productivity. Most
manpower managers realise that achieving and maintaining an exact balance of work input and
output is impossible, but by building in flexibility, the situation can be kept in control. Flexibility
in work input can be achieved by allowing a small backlog of work to develop. Flexibility in
output is achieved by working overtime, or using temporary or clerical agency staff in the short
term and in the longer term, by training schemes, promotion and recruitment policies.
Traditionally manpower planning has been concerned with planning in the longer term, where
projections of future staff numbers can be made by applying the promotion and wastage rates
observed in the past 1 . This is encoded in Markov type models, which can be used to assess the
impact of personnel policies. A typical time scale for such a plan would be between 1 and 5 years.
The absolute accuracy of planning over this time horizon is often of little consequence as the main
objective is a broad assessment of future policy.
Manpower planning in the short term is about scheduling the available manpower resource to
meet current objectives. Typically this requires a much more pragmatic approach. Here the few
options available to management must be used with great care to bring about the desired
outcome. A relatively high degree of accuracy is required as there will be little opportunity to
correct inaccurate assumptions. Poor strategies may well compound the problem. Mathematical
programming techniques can be useful in deciding the allocation of scarce manpower resources
amongst competing activities. However it appears that management seldom use traditional man-
power planning techniques 2 • One reason for this may be the lack of understanding of the methods
used, which leads to a lack of confidence in the results produced. It is unlikely that management
will have the necessary skills to fully understand and implement the techniques themselves and
more pragmatically, they are unlikely to have the skills for running the necessary computer soft-
ware. A consultant is required to provide the technical expertise: formulating models and inter-
preting output. If a good working relationship is established the manpower planning process may

Correspondence: G. I. Mould, School of Management, University of Stirling, Stirlingshire, Scotland

Operational Research Society is collaborating with JSTOR to digitize, preserve, and extend access to
Journal of the Operational Research Society. ®
www.jstor.org
 G. I. Mould-Case Study of Manpower Planning for Clerical Operations 359

progress quite satisfactorily. However this is not always the case and management may lack the
confidence to use the model unaided and hence the manpower planning exercise will stop.
One of the main aims of this study was to produce a solution which was readily accessible to
management in an attempt to overcome implementation problems. Venema and Vessels 3 identi-
fied the benefits of adopting a decision support system approach, which gives the end user the
freedom to build and evaluate different scenarios of their choosing. Lee and Biles4 demonstrated
that a spreadsheet based manpower model could be implemented and used by management.

THE PROBLEM

The particular problem considered in this paper is that of an agency which undertakes contract
clerical work. The agency run a number of different contracts or schemes at any one time: these
can be very different in nature and subject matter. These run independently of each other with
clerical staff and management being dedicated to a particular scheme. Transfer of staff between
schemes is rare and must follow formal procedures of internal advertising of vacant posts. The
work of the scheme considered here is best described as the administration of claims relating to
redundancy or pension benefit. Each case relates to a claimant who is individually assessed on
their past employment history and current circumstances for their future entitlement to payments
of benefit. This assessment can be a complex process which requires detailed interpretation of the
rules of the scheme with respect to the claimant. The bulk of the work involves the processing of
new claims and answering queries relating to these. Once a claim has been processed it is put on
to the computer for routine handling. Some reprocessing may be required in the future if there is a
change in the circumstances of the claimant.
The schemes are operated by two grades of clerical staff: grade 1 and grade 2. Grade 1 staff are
usually inexperienced: all temporary staff are recruited to this grade. All grade 2 staff are experi-
enced in the work of the scheme. On average it takes two years for a person on grade 1 to gain the
necessary experience to be promoted to grade 2. There are three types of case which relate to the
current circumstances of the claimant. This can be employed, self employed or on a training
scheme. These will be referred to as types 1, 2 and 3 which are handled by specialist sections on
the scheme (also referred to as sections 1, 2 and 3). The bulk (83%) of the work is of type 1 with
around 13% of cases of type 2 and 4% of type 3. For each type of case there are two categories,
say A and B. These categories relate to the past circumstances of the claimant. In addition the case
may contain arrears, either because of the time it has been waiting for processing on the scheme or
because the applicant has been delayed in making the claim. This will add to its processing time,
as all details of the applicants circumstances during the intervening period must be fully investi-
gated. Category B cases can be more complex than category A, especially if the case contains
several months of arrears. A type 3 case is usually more difficult to process than types 1 and 2. In
fact, the processing of a type B3 case will generally require the attention of a well experienced
member of staff. In addition, if the case contains several months of arrears, it can take several days
to process. All category A, B1 and B2 cases without arrears will require little more than 30
minutes attention from a relatively inexperienced member of staff. The split of the work by type
and category plus a summary of processing rates is shown in Table 1, where the categories of
work have been simplified to type A, Band complex type B (type B cases with arrears). The nature
of the work is temporary in that it is envisaged that all cases will occur within the next two years.
After that the routine work will be largely handled by computer, with only a skeleton staff to
respond to ad hoc queries and instances of change in circumstances.
The agency is required to allocate its working time between category A and B work roughly in
proportion to the number of each type of case on file. At the present time approximately 60% of
cases are type A. However, this does not reflect the time required to process the cases. If this
balance of work were adopted then type A work would account for significantly more than 60%
of the processed cases owing to the expected shorter processing time. The agency has tended to
reverse this ratio, spending 60% of its time on category B work, with the resulting balance (A : B)
in terms of numbers of cases processed being approximately 3 : 2.
360 Journal of the Operational Rasearch Society Vol. 47, No. 3

TABLE 1. Split of work, processing rates and current backlog

Work category Section 1

A: simple B: simple B: complex

% of new cases 50 15 18
%of post 42 34
Proc. rate grade 2 staff* 21.5 21.5 4.0
Proc. rate grade 1 staff* 12.5 12.5
Current Backlog (in Mandays) 303.2 243.2 332.0

Work category Section 2

A: simple B: simple B: complex

% of new cases 10 2
%of post 11 9
Proc. rate grade 2 staff* 10.0 10.0 4.0
Proc. rate grade 1 staff* 7.0 7.0
Current Backlog (in Mandays) 22.1 14.7

Work category Section 3

A: simple B: simple B: complex

% of new cases 2 1
%of post 2 1
Proc. rate grade 2 staff* 5.5 5.5 4.0
Proc. rate grade 1 staff* 3.0 3.0
Current Backlog (in Mandays) 24.5 16.5 6.0

* Cases per day

The work has been progressing for a number of years and management has had the situation
well planned, and, for the most part under control. The agency has received reliable information
as to the number of new cases to expect each year. Estimates of the numbers of staff required for
each section have been good. Consequently the work of the agency has progressed smoothly with
a few delays. The backlog of work has been kept to a size that could be cleared by working
overtime for a week or two if it was so desired. In fact a small backlog of work was desirable as
this gave management some flexibility, and ensured that staff were always fully occupied.
Problems arose when there was a huge influx of unexpected work. At the start of the year work
was progressing normally, however the number of new cases began to increase in the early spring.
By April more than the total number of expected new cases for the year had been received, and
cases were still arriving at a considerable rate. A sizeable backlog of work had built up and
management felt they were losing control of the situation.

MANAGEMENT'S SOLUTION

Management's first action was to recruit more staff on short term contracts. It was possible for
the new staff to start work within a few weeks of the decision to recruit. These new recruits had to
be trained and this required the time of the more experienced staff, thus taking them off the job.
The situation deteriorated. This was partly due to the loss of the time of experienced staff, and
also due to the build up of arrears on the outstanding cases. Hence the new staff had little imme-
diate effect on the situation. However, within a few weeks the new recruits were able to process
routine cases of type At and type A2, but little could be done to reduce the backlog of type 3 and
category B work. The backlog of category B new cases was in fact increasing due to the reduction
in the available working time of grade 2 staff, who were now busy training the new grade 1
recruits. A second alternative of working overtime was tried, though this was not popular amongst
the staff, but it was a way of increasing the working time of grade 2 staff. It could provide at the
most a 10% increase in work time but, owing to its unpopularity, a 5% increase was more usual.
In time the new recruits became more effective, but many of them left, usually for other permanent
 G. I. Mould-Case Study of Manpower Planning for Clerical Operations 361

positions elsewhere within the organisation. When this became apparent management added a
condition to all new temporary contracts that they could not apply for other internal posts for the
first six months of their employment. Nine months after the original surge in work, the backlog of
work was sizeable and if anything is still increasing. It was at this point that management
requested help with staff planning.

DISCUSSION OF METHODOLOGY

Manpower planning problems are traditionally examined with the aid of Markov type models,
as discussed by McClean 5 • These models are appropriate for long term planning situations where
absenteeism, wastage and promotion rates can be consolidated and their effects monitored over a
number of years. A recent advancement has been the development of a bivariate model 6 which
allows the simultaneous inclusion of contributory factors such as seniority and performance. Short
term manpower planning is often seen as a manpower scheduling problem where the scarce man-
power resource is rationed amongst competing activities. Mathematical programming is the
classic approach to this situation. Krajewski et al. 7 identified three distinct types of manpower
scheduling problems: staff sizing, shift scheduling and the staff assignment problem. The problem
discussed in this paper is essentially staff assignment. Similar problems have been described by
Malhotra 8 , Franz9 and Willis 10 • These have all adopted a linear programming approach or goal
programming when multiple objectives could lead to an infeasible solution9 •
The problem in this case is perhaps best viewed as essentially a short to medium term problem
since the planning horizon is a year ahead, but plans will be updated on a regular three monthly
basis. The main objective was to find a means of reducing the backlog of work to a manageable
size. In addition there were two other considerations. The first was that a solution had to be found
very quickly since the situation was deteriorating daily. The second was to provide a solution that
was readily understandable and accessible to management for the evaluation of alternative future
scenarios. These requirements ruled out the use of Markov or LP type models as they were judged
to be too complex for the situation. An additional reason for rejecting these modelling methods
was that the multiple objectives of management were represented by attempting to quantify the
trade offs between the different alternatives 11 . This sort of methodology limited the potential of
management to investigate a full range of possible solutions. An LP approach embedded within a
decision support system 12 •13 had proved successful elsewhere, though this type of approach would
require considerable effort to develop a robust interface between the LP and the decision support
system. Time pressures effectively eliminated this approach. A system dynamics approach is often
useful for gaining a rapid overview of the situation but has the complication of difficulty of use for
evaluating numerous alternative scenarios, and is unable to provide the fine detail required for
developing feasible operational plans 14•
It was decided to adopt a decision support system approach for the reasons of flexibility as
identified by Verbeek 15 . An interactive model was built in a spreadsheet medium. This essentially
provided a simple simulation of the problem. The spreadsheet medium facilitated changes to the
data, and allowed the presentation of results in a readily comprehensible graphical form. Manage-
ment could assess the impact of different options, such as: varying the number of temporary staff
recruited; and changing the amount of overtime, etc. Though the solution methodology might be
thought of as 'quick and dirty' there is evidence to suggest that this methodology is actually
preferred by managementl 6 •

THE MODEL

All relevant data were collected. For the staff: details of numbers in each grade; attendance and
overtime data, which was available on a weekly basis for the past two years. Data on the number
of cases received and processed each week plus the routine post and queries, was available by case
category and type, also on a weekly basis. This was used to establish the expected processing rates
by the different grades of staff for each type of case. The analysis indicated that the number of
362 Journal of the Operational Research Society Vol. 47, No. 3

items of routine post and queries was roughly proportional to the number of cases on file, hence
estimates could be made for the number to be expected by each section for the coming year. Data
on a past large influx of category B work with arrears was available. From this it was possible to
estimate the time required to process cases with arrears. The current backlog of work was broken
down into categories and types of cases plus details of the delay as shown in Table 1, which
indicates that the situation was most critical for section 1. In addition, forecasts of the number of
new cases of category A and B were available for the year ahead. Using the past data it was
possible to estimate the number of cases that would go to each section.
With the aid of observations by management a number of assumptions were made about the
effectiveness of staff. It was assumed that a new grade 1 recruit took a month to become fully
effective. During this first month of employment they would require close supervision from a
grade 2 staff member. It was assumed that this training would take roughly 30% of the time of a
grade 2 staff member. Management also provided information on how they prioritised work and
details of how they assigned staff to particular types of work. The latter provides a simple rule
base to the model. This exercise was important for establishing the logic of any manpower model,
and was also a useful process for management as it helped them clarify the operational require-
ments of the scheme. Like many management activities, the prioritisation and assignment of staff
was undertaken on a daily basis without management being aware of exactly how they were
performing this task. The exercise of translating their daily activities into a rule base for the model
had the unexpected benefit ofmaking them aware of the rules they were routinely applying.

DECISION SUPPORT SYSTEM

A spreadsheet model was constructed to simulate the processing of claims over the next twelve
months. The original model was constructed using Smart. However more recent developments
have been undertaken using Quattro Pro, with the three dimensional aspect of the package
proving useful for presenting different groupings of data within the same model. Figure 1 shows a
screen image of the system showing the detailed work plan for the coming year. This also illus-
trates the use of different spreadsheet pages for the different groupings of data. There were essen-
tially four categories of data: model assumptions, model data, management decision variables and
management information reports. The first three of these cover the required data input and the
fourth is the data output from the model as shown in Fig. 2.
The model assumptions were established from the analysis of past work patterns, examples of
these are: processing rates; absenteeism rates; split of work by section. The model data consisted
of details of the current situation and forecasts of new work. Management's decision variables
reflected the tools that management had available to solve the problem. For example, the staff
plan (as shown in Table 2), the amount of overtime worked, the allocation of staff to a particular
section. The information contained in the three input data categories enable the calculation of

TABLE 2. Staff plan

Grade 2 Grade 1
Leaving Current New or Maximum Leaving Current New Maximum
promoted overtime overtime
Jan 15 5% 11 5%
Feb 15 5% 11 5%
Mar 15 5% 10 5%
Apr 15 5% 11 5%
May 14 5% 11 5%
Jun 15 5% 11 5%
Jul 15 5% 11 5%
Aug 15 5% 11 5%
Sep 15 5% 11 5%
Oct 15 5% 10 2 5%
Nov 15 5% 12 1 5%
Dec 14 5% 2 11 5%
 G. I. Mould-Case Study of Manpower Planning for Clerical Operations 363

FIG. 1. Screen image of the plan of work, worksheet.

outstanding workload and available staff time. The rule base simulates the staff allocation pro-
cedures as implemented in practice. Available staff time is allocated to the work on hand accord-
ing to the priorities specified by management. Hence the number of cases processed monthly of
each type can be calculated. For example, the available working time of grade 2 staff is calculated
from the data in the model. The proportion of working time allocated to complex B work is
calculated from the ratio specified in the management decision variables. Hence the number of
cases that can be processed is calculated and the outstanding work load is atljusted accordingly. If
there is insufficient work to fully employ these staff then the spare capacity is made available for
the processing of another type of work, in this case simple category B work. Any further spare
capacity of grade 2 staff would then be made available to category A work. Similarly for grade 1

Model Aaaumptlona Model Data

Current backlog
Absenteeism rates Forecasts of new work
Processing rates etc.
Split of work etc.

Mang. lnl. Report•

Work plan
Summary of backlog
staff usage etc.

FIG. 2. Structure of the manpower model.

364 Jaumal af 1M D....-•tianal ~rch Sociaty Val. 47. No.3

staff, these are allocated to category A or B work according to the proportions specified in the
management decision variables, with any spare capacity being reallocated to the other type of
work. Any unused staff time is recorded in the 'usage of staff' report.
The results from the model are displayed on the computer screen in the form of management
information reports. These include a detailed work plan, Figure 1, which gives details of the
backlog, new work and processed work, for each type of work, on a monthly basis; a monthly
summary of the backlog of work, expressed in terms of the numbers of cases outstanding and also
in terms of mandays. The backlog in terms of mandays is presented graphically in Figure 3, this
provides a better basis for comprehension and comparison. The final management information
report is the usage of staff by grade as shown in Table 3, this gives useful information on the
amount of overtime employed and any unused staff time.

Section1: Backlog (Mandays)

Staff: 15 grade 2 , 11 grade 1

2000

1500

1000

500

D 8 Complex~ 8 simple Ill A simplo

FIG. 3. Backlog of work, current staffing levels.

The model was designed so that an acceptable situation could be sought by changing the deci-
sion variables. Once this had been found, slight adjustments could be made to model assumptions
to test the robustness of the proposed solution. The forecasts of new work were revised every three
months, at this point the model data could be updated and, if so desired, the performance of the
model monitored over the preceding period.
The model was used by management to test the effects of different personnel strategies on the
backlog of work, the most usual being to look at the effects of different recruitment policies over
time. In addition, further runs were made altering the forecast work load, which could be
increased, decreased, brought forward or delayed. These alternatives were compared in terms of
the backlog of outstanding work. Once an acceptable solution had been found its robustness was
tested by setting processing rates at either a pessimistic, expected or optimistic level.

RESULTS

The model was well received by management who were eager to put it to use. It was first used
to evaluate the effect of recruiting different numbers of temporary grade 1 staff. The results shown
here are for section 1, the busiest of the three sections. Figure 4 shows the outstanding backlog of
work after a year, with different numbers of grade 1 staff, when a maximum of 5% overtime was
allowed. Increasing the number of inexperienced staff had no effect on the backlog of type B
complex work, although it did reduce the backlog of simple work. No further benefit was gained
from employing more than 15 grade 1 staff. If up to 10% overtime is allowed, then there is some
reduction in the backlog of complex type B work, but no additional benefit is gained from having
more than 14 grade 1 staff on the section. It is clear from these results that the number of tempo-
rary staff recruited does not have any effect on the size of the backlog of complex type B work.
 TABLE 3. Usage of staff

Section 1: Mandays used inc. overtime

Category A Category B Total Total
Routine New Routine New Claims Overtime Days
Queries Claims Queries Claims arrears used (Days) Redundant
---
grade 2 grade 1 grade 2 grade 1 grade 2 grade 1 grade 2 grade 1 grade 2 grade 2 grade 1 grade 2 grade 1
!'"'
Jan 124 94 5 0 38 141 0 0 154 15 11 0 0
Feb 102 79 5 0 32 118 0 0 129 13 9 0 0
Mar 119 90 5 0 37 135 0 0 149 15 11 0 0 ~
Apr 79 71 10 0 21 106 6 0 107 11 8 0 0 !.
May
Jun
79
88
67
78
10
10
0
0
21
24
101
116
6
6
0
0
107
117
11
12
8
9
0
0
0
0
=
Jul 113 87 5 0 35 130 0 0 142 14 10 0 0 j
Aug 78 65 5 0 24 98 0 0 99 10 8 0 0 a.
.....=
Sep 90 69 5 0 28 104 0 0 113 11 8 0 0 i:
Oct 116 96 6 0 35 144 1 0 146 15 11 0 0
Nov
Dec
99
68
91
56
6
6
0
0
30
21
136
84
1
1
0
0
126
89
13
9
11
7
0
0
0
0 I...
ii
=
=
;;·

.
"'if
if
;:
Cl
1:
3.
;;·
=
w
en
CJ"1
366 Journal of the Operational Research Society Vol. 47, No. 3

End of Year Backlog (Mandays)

Staff: 15 grade 2 , 5% overtime

2000

1500 - - - - - - - - - - - - - - - - - - -

1000

500

1
No. of grade 1 Staff

CJ B Complex ~ 8 Simple Ei:t A Simple

FIG. 4. End of year backlog of work with different numbers of grade 1 staff.

Some benefit can be gained from increasing the proportion of overtime worked, though it was
thought not to be possible to sustain an overtime rate of more than 5% for any significant period
of time. The only means of reducing the backlog of complex type B work was to substantially
increase the proportion of grade 2 staff time allocated to this task. Only when 95% of all grade 2
staff are allocated to category B work does the backlog decline, as shown in Figure 5. This
alternative is probably not feasible in practice, as grade 2 staff are also required to perform a
supervisory role for the routine work of the section.
The results were disappointing as the model demonstrated that a sizeable recruitment plan
would have little effect on the backlog. While the backlog of simple work could be easily eradi-
cated, the problem of the complex cases with arrears remained. The model demonstrated to man-
agement that the options under consideration were not capable of resolving the situation. Pressure
to find a solution quickly was increasing. The only option that had any effect, working overtime,
was unpopular with grade 2 staff. It was now obvious that a more radical plan was required.
The agency operated a number of similar schemes to the one described here, which were not
operating under such pressures at the time. One such possibility was to take grade 2 staff from the
other schemes working on different but related work and retrain them for this scheme. Meanwhile
the other schemes would be able to make effective use of some temporary grade 1 staff, if the work

Section 1:Backlog (Mandays)

Staff: 18 g1, 15 g2 (95% on Comp. B)

2000

1500 - - - - - - - - - - - - - - - - - - - - -

1000

500

LJ 8 Complex~ 8 simple IIJ A simple

FIG. 5. Backlog of work, grade 2 staff spending 95% of time on category B work.
 G. I. Mould-C.. Study af Man...-r Planning far Clericll O!*llians 367

was organised for them to handle the routine tasks. For implementation this option required the
cooperation of senior management as there was no precedent for the transfer of staff between
schemes in the agency. This option provided a far better solution than had been previously
obtained. With the additional effort of 3 grade 2 staff making a total of 18, it was possible to
eliminate the backlog of complex category B work by the end of the year. With this arrangement,
13 grade 1 staff were required to reduce the backlog of routine category A and B work to an
acceptable level. These results are shown in Figure 6. Further increases in temporary staff had
little effect on the situation. After obtaining this general set of results management were then able
to vary other model assumptions to test the robustness of the solution. The model proved to be
very useful to management who were eager to use it for themselves to evaluate a wide range of
scenarios.

Section 1: Backlog (Mandays)

Staff: 18 grade 2, 13 grade 1

2000

1500 - - - - - - - - - - - - - - - - - - - -

1000

500

0 B Complex~ B simple GiJ A simple

FIG. 6. Backlog of work with 3 additional grade 2 staff.

CONCLUSIONS

The case study illustrates a manpower planning problem where the classic techniques are of
limited value because of the speed of response required for a short term problem. In addition, it
was important to produce a model that would be readily accessible for use by management. These
conditions were satisfied by a spreadsheet based decision support system. Whilst the system would
not produce the optimum solution it would allow management to investigate the effect of their
different policy options. The techniques used to construct the model were readily understood, and
all assumptions were made highly visible. Management was encouraged to use the model them-
selves by changing decision variables and performing sensitivity analysis on model assumptions.
The simple operation, the instant response, the graphical presentation of results and the ease of
obtaining printouts, were all factors which contributed to the success of the model. As a result the
model has been extensively used by management. Similar models are being employed on other
schemes within the agency where the situation is not as critical as described in this paper, but
where benefits can be gained from such a planning tool.
The manpower model was essentially a simple arithmetic calculation of work input and
throughput over time. However this simple modelling process enabled management to gain a
fuller understanding of the situation. One useful result was that the model proved that the obvious
tactic of drafting in more temporary staff was not beneficial. Although in hindsight the solution of
using grade 2 staff from other schemes may seem obvious, at the time it did not, as there had not
been a tradition of moving staff between schemes in the agency. The implementation of the solu-
tion required the intervention of much higher level management than are normally involved in
short term manpower issues.
368 Journal of the Operational Research Society Vol. 47, No. 3

REFERENCES

1. S. VAJDA (1978) Mathematics of Manpower Planning. Wiley, Chichester.

2. M. J. FEUER, R. J. NIEHAUS and J. A. SHERIDAN (1984) Human resource planning: a survey of practice and potential.
Hum. Res. Plan. 7, 85-97.
3. W. Z. VENEMA and J. WESSELS (1988) Systematic modelling and model handling for manpower planning systems..
Decision Support Systems 4, 503-508.
4. A. S. LEE and G. E. BILES (1988) A microcomputer model for human resource planning. Hum. Res. Plan. 11, 293-315.
5. S. McCLEAN (1991) Manpower planning models and their estimation. Eur. J. Opl Res. 51, 179-87.
6. B. G. RAGHAVENDRA (1991) A bivariate model for Markov manpower planning systems. J. Opl Res. Soc. 42,565-570.
7. L. J. K~.AJEWSKI and L. P. RITZMAN (1977) Disaggregation in manufacturing and service organisations: a survey of
problems and research. Decis. Sci. 8, 1-18.
8. M. K. MALHOTRA, L. P. RITZMAN, W. C. BENTON and G. K. LEONG (1992) A model for scheduling postal distribution
employees. Eur. J. Opl Res. 58, 374-385.
9. L. S. FRANZ, H. M. BAKER, G. K. LEONG and T. R. RAKES (1989) A mathematical model for scheduling and staffing
multiclinic health regions. Eur. J. Opl Res. 41,277-289.
10. R. J. WILLIS and S. B. HUXFORD (1991) Staffing rosters with breaks-a case study. J. Opl Res. Soc. 42,727-731.
11. R. L. ACKOFF (1960) Unsuccessful case studies and why. Opns Res. 8, 259-263.
12. H. S. WEIGEL and S. P. WILCOX (1993) The army's personnel decision support system. Decision Support Systems 9,
281-306.
13. G. G. HEGDE and P.R. TADIKAMALLA (1990) Interactive workforce management on a PC. Eur. J. Opl Res. 48,275-280.
14. G. I. MouLD (1994) Short term manpower planning: time based simulation v system dynamics approach. Presented at
the 1994 International System Dynamics Conference, Stirling, July 1994.
15. P. J. VERBECK (1991) Decision support systems-an application in strategic manpower planning of airline pilots. Eur. J.
Opl Res. 55, 368-381.
16. S. ANTHONY and J. M. WILSON (1990) Manpower modelling using a spreadsheet. Omega 18, 505-510.

Received February 1994; accepted August 1995 after two revisions