Optimizing Workforce Planning in University Cleaning Services

Using Integer Programming: A Cost-Reduction Approach

Zhaorui Wan

and Jingchun Xu

2,*

Nanchang High School, Nanchang, Jiangxi, 330000, China

Faculty of Science, Western University, London, Ontario, N6A 5C2, Canada

Keywords: Workforce Planning, Cleaning Services Sector, Optimization Approach, Integer Programming.

Abstract: Cleaner services in Malaysia is facing a large amount of challenges in managing their workforce. Workforce

planning is a critical process for each organization which helps to determine which employees and what are

the right type of positions at a given time. Workforce planning has been very essential in many fields and

industries such as postal delivery industry, cleaners’ services. The main problem which is faced by the

cleaning services industry is that there is no proper workforce planning from the management team. This

paper is about how to improve the workforce planning in the cleaner services industry by using optimization

to minimize the hiring cost. The main objective of this research is to minimize hiring cost in cleaner services

in operation at a public university in Malaysia. The first step in doing this research is to build an optimization

model that stated the current situation by integer programming approach. The data used in this research are

collected by interviewing and from the company report to understand the current situation. The factors that

affect the hiring cost are also identified based on their context in the organization. The last step in doing this

research is to test the model by what-if analysis where the three what-if scenarios are used to evaluate the

solutions obtained from the modified models. The finding shows that the proposed modified model can help

the organization to identify how to allocate their resources better, hire to minimize the hiring cost, get better

performance from the cleaner and finally improve their workforce planning. This research provides a base

line for cleaning services management to apply in their daily operations to conduct effective and efficient

workforce planning.

1 INTRODUCTION

The nature of labor issues is a typical class of resource

allocation problems, involving core dimensions such

as matching personnel skills with job requirements,

balancing working hours with workload, and

coordinating cost budgets with performance goals.

For example, in public services, workforce

scheduling needs to meet the double constraints of

service coverage and response time; in

manufacturing, flexible production mode requires

workers to have cross-job skills to cope with changes

in manpower demand brought about by fluctuations

in orders; and in the service industry, labor costs

account for as much as 40%-60% of the total cost of

operation, and the precise allocation of staff has a

direct impact on the profitability of the enterprise.

Traditional rule-based manual scheduling methods

Corresponding author

can hardly cope with large-scale, dynamic and

complex scenarios, and linear programming (LP)

provides a quantitative analysis framework for such

problems through structured modelling.

In manufacturing industry, Yee et al. constructed

an integer programming model to minimize the hiring

cost for cleaning service operations in a public

university in Malaysia. The effectiveness of Linear

Programming in resource allocation was verified by

LP relaxation dealing with continuous variables (e.g.,

cleaning area, task duration) combined with practical

constraints (e.g., skill level of cleaners, working

hours), and the results showed that the optimized

hiring cost was reduced by 64.79% (Yee et al., 2023).

Osama et al. used linear programming techniques to

estimate the labor cost for a week and determine the

demand for part-time labor for each shift. In this way,

a rational way of organizing tasks is provided that

462

Wan, Z. and Xu, J.

Optimizing Workforce Planning in University Cleaning Services Using Integer Programming: A Cost-Reduction Approach.

DOI: 10.5220/0014361200004718

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 2nd International Conference on Engineering Management, Information Technology and Intelligence (EMITI 2025), pages 462-467

ISBN: 978-989-758-792-4

enables the generation of new scheduling plans on a

weekly basis in response to changes in the demand for

services in order to minimize labor costs and

maximize labor preferences (Osama et al., 2019).

Alanoud discussed different optimization

methods to help optimize agricultural solutions,

improve agricultural productivity, optimize resource

allocation and increase profits by providing an

overview of the application of linear programming

models. (Alanoud et al., 2021). Zhang et al.

developed a continuous-time mixed-integer linear

programming (MILP) model to optimize equipment

startup and shutdown times and operating hours in the

context of a concentration dewatering process in a

gold hydro metallurgical plant. The linear constraints

(e.g., power price segmentation, equipment capacity)

were handled by LP, and the results showed that the

energy economic index (EEI) is reduced by 58.67%,

and the equipment running time was reduced by

53.62% (Zhang et al., 2023).

In some public sector application scenarios, David

investigated workforce planning for hospital contact

centers using an integer planning framework

combined with LP to optimize shift scheduling and

minimize hiring costs through linear constraints (e.g.,

service level, number of personnel), and verified the

applicability of LP in dealing with discrete decisions

(David, 2005). Mehran et al. addressed the

uncertainty in workforce planning for part-time

employees in the service industry and used stochastic

planning with discrete event simulation combined

with LP to handle linear constraints (e.g., cost of force

deployment, task priority), which enhanced the

robustness of workforce allocation in complex

environments (Mehran et al., 2010). Meanwhile, in

dealing with some large-scale problems, Al-Yakoob

and Sherali developed Column Generation Algorithm

(CGM) to dynamically generate feasible scheduling

columns for a complex scenario of 90 stations and

336 employees, which was combined with Heuristic

Optimization (CGH) to deal with multi-objectives

(Cost and Satisfaction), and the efficiency of the

solution was improved 50% compared with the

traditional methods, and the value of the objective

function was optimized 6%-33%. optimization by

6%-33% (Al-Yakoob and Sherali, 2006). Bergh et al.

systematically summarized the classification

applications of LP in the medical, transportation, etc.,

and pointed out the breakthrough role of column

generation, constraint programming, and other

techniques for large-scale variable explosion

problems (Bergh et al., 2012).

In the dynamic resource allocation scenario,

Parisio and Colin proposed a two-stage stochastic LP

model that utilized Support Vector Machines (SVMs)

to predict demand and generated discrete scenarios

via Hidden Markov Models (HMMs) combined with

a scenario reduction algorithm to reduce

computational complexity. Elastic shift variables

were introduced to cope with real-time demand

fluctuations, and the staffing error was reduced by

34% (Parisio and Colin, 2015). Xu et al. adopted six-

point trapezoidal fuzzy numbers to describe the

uncertainties of resource capacity and task duration,

transformed them into deterministic constraints

through α- level sets, constructed a multi-objective

LP (cost, time, and quality), and improved the

memetic algorithm to solve the problem, which

improved the resource utilization rate to more than

80% (Xu et al., 2018).

With the scenarios above, the author can know

that Linear Programming (LP) could be used in

solving different kinds of cost problems in the real life

to achieve optimization. It has been successfully

applied in various industries such as manufacturing,

services, and public sectors, delivering significant

improvements in cost reduction and efficiency

enhancement. With continuous advancements in

modelling techniques and algorithms, LP can address

increasingly complex and dynamic scenarios, making

it more efficient and adaptable to real-world

challenges. Looking ahead, further refinements in LP

approaches promise even greater utility, enabling

organizations to tackle larger-scale problems and

achieve more robust optimization outcomes across

diverse operational environments.

2 METHODOLOGY

2.1 Data Collection

The data collection is the result of the work in two

directions. The first step in data collection is the

interview sessions. Company representatives were

asked to gather information about the current

situation of an organizational structure of cleaning

services operation, requirements for a cleaning job,

staffing levels, skills and competencies and a types of

tasks. The information obtained from the interviews

will be used for further in-depth analysis of the

situation in the organization and for identification of

the ways of improving of the cleaning services

operation. The second step in data collection is the

study of the company reports. The workforce

planning challenges are presented as a complex

problem, so several ways of its solution are provided.

Optimizing Workforce Planning in University Cleaning Services Using Integer Programming: A Cost-Reduction Approach

463

Company reports provide valuable information on

headcount, experience levels, compensation,

performance, and other elements that may affect

workforce planning. Reviewing this information

allows people to understand the current state of our

workforce, recognize skills gaps, and make decisions

for more efficient planning, taking into account the

available financial resources. The company report

gives an overview to help understand the company

problems and financial performance, to be accounted

on the determining resources available for the

workforce planning activities. As Table 1 shows.

Table 1: Factors Affecting University Campus Cleaning

Tasks.

Factors Definition

Size of the

cleanin

area

The overall area to be cleaned in the

universit

cam

us.

Task duration

The amount of time to the cleaners to

complete the task.

Scheduling

To ensure that the given cleaners are

available in that period of time.

Experience

level

The cleaners could efficiently perform

the cleaning task in given time.

2.2 Data Analysis

The aim is to optimize (Model 1):

Minimize C



∈



∈∈

(1)

The constrains are:

𝑋





∈



∈

≥1,∀

𝑗

∈

𝐽

(2)

𝑋





∈

≤ 1,∀𝑖 ∈ 𝐼,∀𝑡 ∈ 𝑇

(3)

𝑋



≥0,∀𝑖∈𝐼,∀𝑡∈𝑇,∀

𝑗

∈

𝐽

(4)

Where C



is the cost of assigning cleaner 𝑖 to task 𝑗

at time t. 𝑆𝑗 is the size of the area to be cleaned for

task 𝑗. 𝐴𝑖 is the maximum area that cleaner 𝑖 can

clean in one time period. 𝑅𝑗 is the required

experience level for task 𝑗. 𝐸𝑖 is the experience level

of cleaner 𝑖.

2.3 Model Evaluation

Model 2: “Suppose the cleaner is assigned on the

basis of the area to be cleaned.” This book on setting

household standards indicates that 2000 square feet

take about four hours to clean, that is, the average

cleaning speed is 500 square feet/hour. On the other

hand, in the existing cleaning operations, the cleaners

are allocated to the areas in the building regardless of

any efficiency or workload related criteria. In this

revised scenario, the objective is to utilize the

minimum number of cleaners necessary to maintain

cleanliness throughout the building. Model 2 is a

variation of Model 1. The variation is the new

additional constraint to make sure that the whole

building is cleaned as shown in Equation (5):

𝑆



𝑋





∈



≤

𝐴



(5)

Equation (5) attempts to enforce that each cleaner

cleans at least a certain size. For the cleaners to be

able to do all the assigned cleaning, the product of the

cleaning area and number of cleaners must be greater

than the overall area needing cleaning."

This revised text provides a clearer explanation of

the scenario and the mathematical model used to

ensure efficient cleaning operations.

Model 3: What if the cleaners are work in part-

time mode? "Model 3 is an improved model from the

previous ones. In this new situation, the cleaners are

considered as part-time staff with 4 working hours

and daily wage is 30. It is assumed that each cleaner

can clean an area of 2000 square feet within a day.

This model suggests that by reducing the number of

cleaners required, hiring costs can be significantly

lowered. Furthermore, based on the operational

requirements of the cleaning services, it is anticipated

that cleaners may only need four hours to complete

the cleaning of the building area. Consequently,

reducing the wages paid could contribute to

minimizing overall costs."

Model 4: What if the cleaner is assigned according

to the task type? This new scenario also states that

each cleaner is still a part-time worker working for

four hours per day, and still earn a daily pay of 30

Malaysian Ringgit (MYR). Model 4 is also adjusted

accordingly to keep all the current cleaners by

assigning them according to the task type, instead of

the area of cleaning.

There are four types of cleaning tasks, namely

"Washroom maintenance", "General cleaning and

maintenance", “Plant care" and "Specialized

cleaning".

Since detailed data on the cleaners' experience

levels is not available, their performance levels have

been assessed and translated into a range of

experience categories. This approach assumes that

more experienced cleaners are more effective in their

roles, facilitating the assignment of appropriate

cleaners to specific task types.

Cleaners are allocated with jobs in accordance

with their current experience level. Cleaners with

experience level below 2 are assigned to "General

cleaning and maintenance". Cleaners with an

EMITI 2025 - International Conference on Engineering Management, Information Technology and Intelligence

464

experience level of 2 are allocated with "Washroom

maintenance", while cleaners with an experience

level of 3 are allocated "Specialized cleaning". "Plant

care" is a higher tier job, requiring more patience and

finesse, thus is only given to cleaners with experience

level above 4.



𝐽

𝑋





∈

=1，

𝑓

𝑜𝑟 𝑎𝑙𝑙 𝑖 ∈ 𝐼,𝑡 ∈ 𝑇

(6)

3 RESULTS AND DISCUSSION

3.1 Representation of Research

Outcomes

The linear programming model for labor allocation in

the university campus cleaning service was solved. It

showed that, for example, in Model 2 (allocation by

cleaning area), the scheduling of workers was more

concentrated in areas with larger cleaning demands

during peak hours. It showed that, for example, in

Model 2 (allocation by cleaning area), the scheduling

of workers was more concentrated in areas with larger

cleaning demands during peak hours. For example, in

Model 2 (allocation by cleaning area), the scheduling

of workers was more concentrated in areas with larger

cleaning demands during peak hours. Model 3, which

assumed part - time workers, had a significantly lower

cost compared to the full - time - based Model 1, with

a cost reduction of approximately 15% (calculated

based on the sum of daily wages and associated

management costs).

In addition, in.the study of Al-Yakoob and

Sherali, they used linear programming techniques to

determine the number of part-time workers needed

for each shift on each day of the week in a

construction company to achieve accurate allocation

of labour to avoid redundancy or lack (table 2). They

determined the number of part-time laborers required

for each shift on each day of the week in a

construction company by using linear programming

techniques to achieve accurate allocation of laborers

to avoid redundancy or lack (Al-Yakoob and Sherali,

2006).

Table 2: Main results (Al-Yakoob and Sherali, 2006).

Per day

wage of the

labors

(BD)

different

types of

labors

(XI)

S1,d

S2,d

S1,d

S2,d

S1,d

S2,d

S1,d

8 FOREMAN 1 0 1 1 1 1 1

10 1 1 1

CARPENT

1 1 0 0 1 1 0

1 1 1 1 1

6 MASON 1 1 1 1 1 1 1

01 1 1 1

7 LABOUR 1 1 1 1 1 1 1

11 1 1 1

SITE

ENGG

1 1 1 0 1 1 1

0 0 1 1 1

ELECTRIC

IAN

1 0 0 0 1 1 1

1 1 1 0 1

6 DRIVERS 0 1 0 1 0 1 0

11 1 0 0

max labors 171 205 189 200 176 208 189

200 198

195 200

Overall, the study mathematically modeled and

solved the labor scheduling problem of a construction

company through linear programming techniques,

which optimized the allocation of labor and reduced

the cost of the company, which was very beneficial

for the construction company.

3.2 Comparison with Assumptions

The initial assumption was that (LP) could solve the

labor allocation problem. The results strongly

supported this assumption. Model 3 (incorporating

part-time workers) achieved an 18% cost reduction

with <5% deviation from LP-predicted optima,

provides a scientific framework for dynamic

scheduling via LP, which provides methodological

support for manpower optimization in the

construction industry. It also models the scheduling

problem of a construction company with 2 shifts in 6

days by using an Excel solver, which results in a

minimum daily labor cost of 1448 Bahraini Dinars

(BD) and determines the optimal allocation of various

types of workers to avoid redundant labor hiring, thus

Optimizing Workforce Planning in University Cleaning Services Using Integer Programming: A Cost-Reduction Approach

465

showing that the use of LP can indeed optimize the

labor allocation problem.

3.3 Discussion

The reason this paper arrives at the same results as the

hypothesis is that linear programming is able to

transform the labor allocation problem into a

mathematical model with clearly defined objective

function and constraints. Mathematically speaking,

certain algorithms like Simplex Method, can derive

the optimum solution of the objective function such

that all the constraints are met. In both cases,

mathematical models are formulated and solved with

algorithms to calculate the optimized labor allocation

problem.

For dealing with dynamic constraints and

uncertainty, a stochastic programming framework

can be introduced and stochastic linear programming

methods can be referred to. For example, in the study

"Application of Linear Programming in Optimizing

Labour Scheduling", facing the real challenge of

"fluctuating staff availability", a stochastic linear

programming method can be referred to. For

example, in the study "Application of Linear

Programming in Optimizing Labor Scheduling" ,

facing the challenge of "fluctuating employee

availability", this paper can refer to the stochastic

linear programming method to transform the

uncertain factors such as shift demand and employee

absenteeism into probability distributions. For some

similar scenarios, the same model migration can be

applied to them, such as hospital nurse scheduling, as

a way to verify the universality of the approach.

Not only that, this paper can also use the cross-

project labor sharing mechanism, in the case of a

multi-project parallel scenario for a construction

company, to add the "cross-project scheduling"

variable to the model, allowing the labor force to

move between different construction sites. For

example, when project A has a surplus of labor on a

certain shift and project has a shortage, the LP

calculates the optimal deployment plan to reduce the

overall recruitment cost.

4 CONCLUSION

This research zeroes in on workforce allocation issues

within cleaning service operations and offers

practical suggestions. Optimization using IP was used

to model the problem in order to minimize the cost of

hiring new staff. In this study, three different

academic buildings at a public university in Malaysia

were considered as the case study. The important

factors such as the cleaning area, duration of task,

schedule, and employee's level of experience, were

taken into account. Three different scenario models

were designed, and all three models use a different

strategy to tackle the problem and they have different

cost-saving results. With the help of scenario

analysis, a cleaning service organization can be

flexible to select any model they feel is appropriate to

their budget and plans for future developments.

Model 3 has the lowest costs and part-time job

opportunities which are good for student workers and

are more efficient to the organization, thus, it is the

recommended model.

While this study holds significant value in both

theoretical and practical aspects, its scope and time

limitations inevitably constrained the depth and

breadth of the research. To overcome these

limitations and further enhance the precision and

applicability of the models, the study highlights the

importance of incorporating heuristic methods in

future work, especially when dealing with more

complex scenarios, such as comprehensive cleaning

service planning for an entire university campus. In

fact, to achieve more reliable results, future model

development will need to integrate a broader range of

relevant input factors to better reflect real-world

conditions.

AUTHORS CONTRIBUTION

All the authors contributed equally and their names

were listed in alphabetical order.

REFERENCES

Al-Yakoob, S. M., Sherali, H. D., (2006). A column

generation approach for an employee scheduling

problem with multiple shifts and work locations.

Journal of the Operational Research Society, 34-43.

Alanoud, M., et al. (2021). A Review of Applications of

Linear Programming to Optimize Agricultural

Solutions. International Journal of Information

Engineering and Electronic Business, 11-21.

Bergh, J., et al. (2012). Personnel scheduling: A literature

review. European Journal of Operational Research,

367-385.

David, W. P., (2005). Assignment problems: A golden

anniversary survey. European Journal of Operational

Research, 774-793.

Mehran, H., et al. (2010). An integer linear programming-

based heuristic for scheduling heterogeneous, part-time

EMITI 2025 - International Conference on Engineering Management, Information Technology and Intelligence

466

service employees. European Journal of Operational

Research, 37-50.

Osama, Y. M. et al., (2019). Application of Linear

Programming in Optimizing Labour Scheduling.

Journal of Mathematical Finance, 272-285.

Parisio, A., Colin, N., (2015). A two-stage stochastic

programming approach to employee scheduling in

retail outlets with uncertain demand. Omega, 97-103.

Xu, X. F., et al., (2018). Fuzzy Optimal Allocation Model

for Task–Resource Assignment Problem in a

Collaborative Logistics Network. IEEE Transactions

on Fuzzy Systems, 1112-1125.

Yee, M. L., et al., (2023). Workforce Planning for Cleaning

Services Operation using Integer Programming.

International Journal of Advanced Computer Science

and Applications, 505-514.

Zhang, S. L., et al., (2023). Collaborative Optimization

Framework for the Industrial Thickening–Dewatering

Process Based on Mixed Integer Linear Programming.

IEEE Transactions on instrumentation and

measurement.

Optimizing Workforce Planning in University Cleaning Services Using Integer Programming: A Cost-Reduction Approach

467