Beruflich Dokumente
Kultur Dokumente
Submitted To
Group Members
Acknowledgement
This term paper is prepared within the course, EMIS-517 (Management Science) under
the MBA (Evening) Program at the University of Dhaka. The research is accomplished
based on the knowledge developed through learning one of the course contents which is
Queuing Theory. We, the group members would like to place gratitude towards our
Course Teacher, Professor Dr. Md. Abdul Hannan Mia for generating the capacity
among us to implement Queuing Theory in real time customer service system in a
Commercial Bank.
Abstract:
Queuing is the process of moving customers in a specific sequence to a specific service
according to the customer need. This paper contains the analysis of Queuing systems for
the empirical data of a Government Commercial Bank Teller scheduling system for
optimizing its customer service. One of the expected gains from studying queuing
systems is to review the efficiency of the models in terms of utilization and waiting length.
We may use queuing simulation to obtain a sample performance result and we are more
interested in obtaining estimated solutions for multiple queuing models. Queues are
common sight of many banks in Bangladesh. The obvious implication of customers
waiting in long and winding queues could result to prolonged discomfort and economic
cost to them; however increasing the service rate will require additional number of tellers
which implies extra cost to management. Customer satisfaction is a concern to service
industries as customers expect to be served promptly when they arrive. Demand for
service is highly variable: depends one hour of the day, or day of the week, or even dates
of the month. For a service industry like a Bank, there is a need for enceinte Bank Teller
scheduling system which takes into account varying customer service demand levels.
The model takes into account real time system behavior including changing customer
arrival rates throughout the day and customer balking. It provides scheduling rules and
the corresponding service levels when demand varies with cost minimization as goals.
Table of Contents
Acknowledgement ........................................................................................................... 1
Abstract: .......................................................................................................................... 2
1.
Introduction: ............................................................................................................. 5
2.
3.
6.
5.1
5.2
References ............................................................................................................. 25
1. Introduction:
Today banks are one of the most important business units for the public. Since the
foundational work of banks, many researchers try to get full advantage of any new
technology to increase customer satisfaction. Therefore an active research has focused
on analyzing the queues to optimize their operations and to reduce waiting time for
customers. In service industries, demand for service is variable and often depends on the
day of the month, day of the week, or even on the time of the day. However, service need
to be delivered promptly when it is demanded. Bad staff customer scheduling results in
long customer waiting times, long queues, and consequently, waiting cost. Bad
scheduling can also result in loss of productively of Tellers due to idle times. On the other
hand, good scheduling results in low waiting cost, good Teller utilization, customer
satisfaction, and more profit. Operations managers are faced with the problem of
recognizing the trade-off that must be taken between providing good service and the cost
of waiting for such service. As service speeds up, time spent waiting on queue decreases.
Service cost however in-creases as the level of service increase. The goal of managers
is to schedule as few employees as possible while maintaining a minimum customer
service level. Managers want queues short enough so that customers do not become
dissatisfied and either leaves without transacting their business or transact and never
return in the future. Some waiting can be allowed if the waiting cost is balanced with
significant saving in service cost. A typical Teller-Customer scheduling system has two
goals:
It is therefore important to recognize peak and non-peak periods to decide the staffing
needs. Identifying the changes in the demand level from past or real- time data is not a
straightforward matter due to high variability. Using queuing theory, one can model and
analyze a real time queuing situation in a Bank, compare scheduling alternatives and the
corresponding service levels, and provide the best scheduling rule based on the desired
service level.
Application of Queuing Theory in Government Commercial Bank | 5
2. Literature Review
Several approaches have been used by researchers for Bank Teller scheduling. These
approaches include the use of Operations Research based scheduling tools such as
Statistics, Work measurement, Queuing theory, and Integer programming [2]. The use of
these operations research based scheduling tools for the staff scheduling problem do not
solve the problem adequately. Mathematical models are very cumbersome and seldom
provide complete answers to real time problems rather; they provide partial solutions [3].
For instance, real time situations violate the assumptions of classical analytical waiting
line models. Queuing theory approach may help Banks schedule services using a FIFO
scheduling policy. However, this approach does not capture and take customer
information or service information generated continuously with time into account. Queuing
theory rely mainly on probability predictions which may be unreliable. The existing
approaches forecast arrival pattern of customers based on historical data and use that to
schedule Tellers. Such approach does not perform real time scheduling using actual
arrival of customers. It does not handle dynamic resource availability. In such a system,
a sudden explosion in the arrival rate may cause a serious deviation in the forecasted
schedule based on historical data. This model presents a scheme for monitoring the
system at all times to identify immediately when changes have occurred in the system
and then generate corrective action to take.
It is the practice of Banks to often staff their Tellers as if all Teller staff had the same
service capabilities. Most applications of queuing models for recommending staffing
levels assume that service times are exponentially distributed and that each staff member
has the same service time distribution [4]. Previously, there has not been strong reason
to abandon these assumptions given the relative scarcity of Teller performance data. This
model in this paper shows how to tract Teller transaction activities to obtain the Teller
performance data thus, allowing service time data to be routinely generated and
integrated with staffing schedules. In most Banks, customer and service information are
identified generally based on manual observation and personal judgment (this is obtained
through information from visits to some local Banks). This gives inaccurate results and
wastes time. It also requires continuous observation by management personnel and thus
Application of Queuing Theory in Government Commercial Bank | 6
results in additional cost. There is no correct procedure for tracking customers sojourn in
the system therefore shunting and jockeying are typical occurrences. These lead to some
customers being dissatisfied as a customers who came first may be served last. The
present work presents a scheme for continuously monitoring the system changes,
generating the information as the changes occur [5].
Automated Teller Machine (ATM) etc. by banks in attempt to minimize waiting line
problem at over-the counter services have not yielded the much desired result due to
frequent breakdown of such computerization and fraudulent activities. Hence, long queue
persisted in almost all Bangladeshi banks. The danger of keeping customers waiting can
cause prolonged discomfort and economic cost to them. The time wasted on the queue
would have been judiciously utilized elsewhere (opportunity cost of time spent in
queuing). In a queuing system, managers must decide what level of service to offer. A
low level of service may be inexpensive, at least in the short run, but may incur high costs
of customer dissatisfaction, such as loss of future business. A high level of service will
cost more to provide and will result in lower dissatisfaction costs. When considering
improvements in services, the management of the bank weighs the cost of providing a
given level of service against the potential costs from having customers waiting. [6]The
goal of queuing analysis is therefore to strike a balance between minimize the economic
cost to the system and provision of satisfactory and reasonable quick service.
3. Queuing Systems
A queuing system consists of one or more servers that provide service to arriving
customers. Figure 1 shows the characteristics of queuing system. The population of
customers may be finite (closed systems) or infinite (open systems). The arrival process
describes how customers enter the system. The customers arrive to the service center in
a random fashion. Queue represents a certain number of customers waiting for service.
The capacity of a queue is either limited or unlimited. Bank is an example of unlimited
queue length. The service is an activity requested by a customer, where each service
takes a specific time. The scheduling algorithm is used to order the customers and to
choose the next customer from the queue.
Population of the
Customers
Arrival
Departure
Queue
Service
Unfortunately, life and quantitative analysis are complicated by the fact that people have
been known to balk or renege. Balking refers to customers who refuse to join the waiting
lines because it is to suit their needs or interests. [7] Reneging customers are those who
enter the queue but then become impatient and leave the need for queuing theory and
waiting line analysis. How many times have you seen a shopper with a basket full of
groceries, including perishables such as milk, frozen food, or meats, simply abandon the
shopping cart before checking out because the line was too long? This expensive
occurrence for the store makes managers acutely aware of the importance of service
level decisions.
Waiting Line characteristics
Queue
The waiting line itself is the second component of a queuing system. The length of a line
can be either limited or unlimited. A queue is limited when it cannot, by law of physical
restrictions, increase to an infinite length. Analytic queuing models are treated in this
article under an assumption of unlimited queue length. A queue is unlimited when its size
is unrestricted, as in the case of the tollbooth serving arriving automobiles.
Queue discipline
A second waiting line characteristic deals with queue discipline. This refers to the rule by
which passengers in the line are to receive service. Most systems use a queue discipline
known as the first in, first out rule (FIFO). This is obviously not appropriate in all service
system, especially those dealing with emergencies.
In most large companies, when computer-produced pay checks are due out on a specific
date, the payroll program has highest priority over other runs.
Service Facility Characteristics
The third part of any queuing system is the service facility. It is important to examine two
basic properties: (1) the configuration of the service system and (2) the pattern of service
times.
4. System Modeling
The problem is how managers can schedule the available tellers so that customers do
not wait too long before receiving the desired service irrespective of the time of their
arrival, and that tellers are not kept idle typically because there are no customers to be
served irrespective of the periods. How to get these optimum service conditions is the
objective of this work. The optimum point is reached when a balance is made between
the minimum service cost and the minimum waiting cost. During busy hours, such as
lunch hours, and during busy days, such as pay days, customers must wait in a queue
before being served by a Teller. Increasing the number of Tellers will actually reduce the
waiting time for the customers, but it does so at the expense of the cost of provision of
extra Teller stations and Tellers. Thus this solution has only a limited applicability. Even
when there are a large number of Tellers, they must be utilized efficiently. Waiting for
customers who are not there at Teller stations does not constitute such an efficient
utilization. Maximum efficiency together with maximum customer service is best
accomplished by scheduling Teller staff based on accurate prediction of levels of traffic
at the Bank during hours of the day. This can only be achieved accurately if the system
is monitored continuously.
Goals of this research:
1. To find a trade-off between the cost of waiting and the cost of providing service in Bank.
This helps managers to eliminate long queues and reduce labor cost by providing efficient
Teller scheduling.
2. To maintain a minimum wait time and service cost by knowing when to increase the
number of Tellers or withdraw Tellers. This maximizes the level of customer satisfaction
with the service provided thereby optimizing service.
4.1 Approach to solving the problem
An algorithm for a superimposed control system which can continuously sample the
throughput system waiting line is developed. This algorithm is used to monitor the system
at all times and to communicate changes in the system input and output to other distant
Application of Queuing Theory in Government Commercial Bank | 12
systems which performs some action (example, a system that raises alarm when the
number in the queue exceeds some threshold). This complementary system is
incorporated to cross-check the queuing system continuously to actually see what is
happening. This is necessary because of inaccuracy of queuing model which is based on
probability theory.
Collection of data samples:
Three branches from Sonali Bank of Bangladesh were been selected for our analysis.
The number of customers arriving in a given interval of time from Sunday to Thursday is
collected continuously, starting from first working day of the month. Service time duration
per-customer and Wait time duration per customer are collected. The average time a
customer spends while waiting for service from the time the customer enters the system
is collected. Customers monthly income and transaction volume which would be used for
cost analysis are also collected. The problem is formulated as a series of relationships in
a mathematical model. A mathematical queuing model is used to estimate the probability
of waiting time exceeding a given threshold when the arrival rate and service rates are
known. A model of the Banks customer-Teller scheduling system is produced such that
the customers waiting time does not exceed the threshold which is determined by
operations managers taking into account changing arrival rates and service time
durations while balancing the service cost with waiting cost.
4.2 Mathematical queuing models
Parameter definitions:
= the arrival rate (average number of arrivals/time period)
= the service rate (average number served/time period)
Customers must be served faster than they arrive (< ) or an infinitely large
queue will build up.
Probability that no customers are in the queuing system:
P0 1
P 1
Pn
0
Lq
W 1 L
Average time customer spends waiting in the queue:
Wq
I 1 U 1
Service time varies from one customer to the next and is independent of
another, but their average rate is known.
Parameter definitions:
= the service rate (average number served per time period) per server
(channel)
c = number of servers
c = mean effective service rate for the system (must exceed arrival rate)
P0
Pn
nc1 1
c
1
n! c! c
n0
n
n
1 P for n c
c!cnc 0
n
1 P for n c probability of n customers in system
Pn n
0
(c 1)!(c )2
Lq L
average number of customers in the queue
1 Lq average time customer is in the queue
Wq W
Pw 1
c! c 0
1. Queuing theory provides models that are capable of determining arrival pattern of
customers or most appropriate number of service stations.
2. Queuing models are helpful in creating balance between the two opportunity costs for
optimization of waiting costs and service costs.
3. Queuing theory provides better understanding of waiting lines so as to develop
adequate service with tolerable waiting.
5.2
1. Most of the queuing models are very complex and cannot be easily understood. The
element of uncertainty is there in almost all queuing situations. Uncertainty arises due to:
I.
We may not know the form of theoretical probability distribution which applies.
II.
We might not know the parameters of the process even when the particular
distribution is known.
III.
We would simply be knowing only the probability distribution of out-comes and not
the distribution of actual outcomes even when (i) and (ii) are known.
2. In addition to the above complications, queue discipline may also impose certain
limitations. If the assumption of First come first served is not a true one (and this
happens in many cases) queuing analysis becomes more complex.
3. In many cases, the observed distributions of service times and time between arrivals
cannot be fitted in the mathematical distributions of usually assumed in queuing models.
For example, the Poisson distribution which is generally supposed to apply may not fit
many business situations.
4. In multi-channel queuing, the departure from one queue often forms the arrival of
another. This makes the analysis more difficult.
6.
Collected Data:
= arrival rate
c = number of servers
50
4
15person/hr
60
Analyzed Data:
Average Customers in System, L =6.6219 Customers
Average Customers in queue, Lq = 3.2886 Customers
Average Time Spent in System , W = 0.1324 Hours
Application of Queuing Theory in Government Commercial Bank | 18
Collected Data:
= arrival rate
c = number of servers
36
4
15person/hr
60
Analyzed Data:
Average Customers in System, L =2.83 Customers
Average Customers in queue, Lq = 0.43 Customers
Average Time Spent in System , W = 0.078 Hours
Average Time Waiting in Line, Wq = 0.012 Hours
Server Utilization , p = 0.6
For Cash Transaction Counter] [2.00 P.M-4.00P.M] [Sun-Thurs]
Collected Data:
= arrival rate
c = number of servers
59
4
15person/hr
60
Analyzed Data:
Average Customers in System, L =30.752 Customers
Average Customers in queue, Lq = 26.8854 Customers
Average Time Spent in System , W = 0.5302 Hours
Average Time Waiting in Line, Wq = 0.4635 Hours
Server Utilization , p = 0.9667
Application of Queuing Theory in Government Commercial Bank | 19
49
= the service rate
6[Assumption] c = mean effective service rate
15person/hr
60
Analyzed Data:
If Server Count , C=4
L =6.05 Customers
L =3.85 Customers
L =3.43 Customers
Lq =2.79 Customers
Lq =0.58 Customers
Lq =0.16 Customers
W = 0.1236 Hours
W = 0.08 Hours
W = 0.07 Hours
Wq = 0.06 Hours
Wq = 0.0118 Hours
Wq = 0 Hours
p = 0.82
p = 0.6533
p = 0.54
Analyzed Data:
Analyzed Data:
Analyzed Data:
L =3.42 Customers
Lq = 1.92 Customers
W = 0.2286 Hours
Wq = 0.1286 Hours
p = 0.75
L =1.87 Customers
Lq = 0.67 Customers
W = 0.16 Hours
Wq = 0.06 Hours
p = 0.6
L =19.49 Customers
Lq = 17.59 Customers
W = 1.03 Hours
Wq = 0.96 Hours
p = 0.95
Assumptions:
[ = arrival rate]=16
[Average Arrival Rate]
[ = the service rate]=10/hr
[Server,C]=3
[c = mean effective
service rate]=20
[ = arrival rate]=16
[Average Arrival Rate]
[ = the service rate]=10/hr
[Server,C]=4
[c = mean effective
service rate]=20
[ = arrival rate]=16
[Average Arrival Rate]
[ = the service rate]=12/hr
[Server,C]=3
[c = mean effective
service rate]=20
Analyzed Data:
Analyzed Data:
Analyzed Data:
L =1.91 Customers
Lq = 0.31 Customers
W = 0.12 Hours
Wq = 0.02 Hours
p = 0.53
L =1.66 Customers
Lq = 0.06 Customers
W = 0.1 Hours
Wq = 0.00 Hours
p = 0.4
L =1.48 Customers
Lq = 0.14 Customers
W = 0.09 Hours
Wq = 0.01 Hours
p = 0.44
So, We Suggest Number of Cash Counter should be 3 and Service Rate Should be 12
person/hour.
Jatrabari Branch:
Data Collection and Analysis for CASH Counters[Sunday-Thursday] :
10.00 A.M-12.00P.M
Collected Data:
[ = arrival rate]=95
[ = the service rate]=20/hr
[Server,C]=5
[c = mean effective
service rate]=100
[ = arrival rate]=75
[ = the service rate]=20/hr
[Server,C]=5
[c = mean effective
service rate]=100
[ = arrival rate]=87
[ = the service rate]=20/hr
[Server,C]=5
[c = mean effective
service rate]=100
Analyzed Data:
Analyzed Data:
Analyzed Data:
L =21.43 Customers
Lq = 16.68 Customers
W = 0.23 Hours
Wq = 0.18 Hours
p = 0.95
L =5.14 Customers
Lq = 1.39 Customers
W = 0.07 Hours
Wq = 0.02 Hours
p = 0.75
L =9.01 Customers
Lq = 4.66 Customers
W = 0.1 Hours
Wq = 0.05 Hours
p = 0.87
Assumptions:
[ = arrival rate]=86
[Average Arrival Rate]
[ = the service rate]=10/hr
[Server,C]=6
[c = mean effective
service rate]=60
[ = arrival rate]=86
[Average Arrival Rate]
[ = the service rate]=10/hr
[Server,C]=7
[c = mean effective
service rate]=70
[ = arrival rate]=86
[Average Arrival Rate]
[ = the service rate]=12/hr
[Server,C]=7
[c = mean effective
service rate]=84
Analyzed Data:
Analyzed Data:
Analyzed Data:
L =5.22 Customers
Lq = 0.92 Customers
W = 0.06 Hours
Wq = 0.01 Hours
p = 0.72
L =4.59 Customers
Lq = 0.29 Customers
W = 0.05 Hours
Wq = 0.00 Hours
p = 0.61
L =5.94 Customers
Lq = 0.88 Customers
W = 0.07 Hours
Wq = 0.01 Hours
p = 0.72
So, We Suggest Number of Cash Counter should be 7 and Service Rate Should be 20
person/hour. If service rate decreases to 17person/hour then no negative impact.
Jatrabari Branch:
Data Collection and Analysis for SERVICE DESK [Sunday-Thursday] :
10.00 A.M-12.00P.M
Collected Data:
[ = arrival rate]=35
[ = the service rate]=10/hr
[Server,C]=4
[c = mean effective
service rate]=40
[ = arrival rate]=32
[ = the service rate]=10/hr
[Server,C]=4
[c = mean effective
service rate]=40
[ = arrival rate]=39
[ = the service rate]=10/hr
[Server,C]=4
[c = mean effective
service rate]=40
Analyzed Data:
Analyzed Data:
Analyzed Data:
L =8.67 Customers
Lq = 5.17 Customers
W = 0.25 Hours
Wq = 0.15 Hours
p = 0.88
L =5.59 Customers
Lq = 2.39 Customers
W = 0.17 Hours
Wq = 0.07 Hours
p = 0.8
L =40.76 Customers
Lq = 36.86 Customers
W = 1.05 Hours
Wq = 0.95 Hours
p = 0.98
Assumptions:
[ = arrival rate]36
[Average Arrival Rate]
[ = the service rate]=10/hr
[Server,C]=5
[c = mean effective
service rate]=50
[ = arrival rate]=36
[Average Arrival Rate]
[ = the service rate]=10/hr
[Server,C]=6
[c = mean effective
service rate]=20
[ = arrival rate]=36
[Average Arrival Rate]
[ = the service rate]=12/hr
[Server,C]=7
[c = mean effective
service rate]=17
Analyzed Data:
Analyzed Data:
Analyzed Data:
L =4.66 Customers
Lq = 1.06 Customers
W = 0.13 Hours
Wq = 0.03 Hours
p = 0.72
L =3.89 Customers
Lq = 0.29 Customers
W = 0.11 Hours
Wq = 0.01 Hours
p = 0.6
L =3.69 Customers
Lq = 0.09 Customers
W = 0.1 Hours
Wq = 0.00 Hours
p = 0.51
So, We Suggest Number of Cash Counter should be 6 and Service Rate Should be
10person/hour at least. If service rate decreases then long time in queue will be
occurred.
7. Conclusion:
In a queue system, the balance between dealing with all customers fairly and the
performance of the system is very important. Sometimes the performance of the system
is more important than dealing with the customers fairly. Queuing models have found
widespread use in the analysis of service facilities, production and many other situations
where congestion or competition for scarce resources may occur. A scheme for
scheduling Tellers in Banks has been developed as discussed. This scheme optimizes
Customer-Teller service. Due to high variable arrival rates, it is not sufficient to forecast
the system conditions using queuing theory and simulation results only. In addition, the
system should also be monitored to see that deviations in the forecasted values are
compensated immediately. This avoids unnecessary costs. These deficiencies in hitherto
presented models are what this model has solved. It seeks to reduce the waiting Time
and hence the queue length by optimizing the number of Tellers at a reduced cost. This
will build up the customer loyalty, who will return for similar business in the future. The
analysis of the result can be used to improve throw of customers through the queuing
system. This would speed up services and provide competitive edge.
8.
References
[1]. Thompson G. Accounting for the multi period impact of science when determining
employee requirement for labour scheduling. Journal of Operations Management, vol. 11,
1993, pp. 269 287.
[2]. Martha A.C., Ronald G., Richard L., and Abdulla M.I. A Simulation ILP Based Tool for
Scheduling ER Staff. In Proceeding of the winter Simulation conference, 2003, p. 1930.
[3]. Hammond D. A Simulation and Analysis of Bank Teller Manning. In Proceeding of
Winter Simulation Conference, 1995, pp. 1077 1080. www.patentstorm.us/patents/ .,
1992.
[4]. Taylor, Bernard W.,Introduction to Management ScienceEd. 9th, ISBN-13: 978-013-275191-9; Pearson Education, Inc.
[5]. Latouche G., and Ramaswami V. Introduction to Matrix Analytic Methods in Stochastic
Modeling, Quassi - Birth - and Death process. ASA SIAM, First Edition, 1999.
[6]. Ronald A.N., and James P.W. Queuing Theory and Customer Satisfaction, A review
of Terminology, Trends and Application to Pharmacy Practice. Hospital Pharmacy
Journal, vol. 36, 2001, pp. 275 279.
[7]. A. Willig, A short introduction to queuing theory, Technical University Berlin,
Telecommunication Networks Group Sekr. FT 5-2, Berlin, July 21, 1999.