Managing Waiting Lines

Definition of Queue
⚫ Queue is a line of waiting customers or objects who
require service from one or more servers.

⚫ Queue can be a
⚫ Physical line
⚫ Virtual line
 Queuing
⚫ Queues are almost unavoidable in our lives
⚫ People spent approximately 5 years of their lives
waiting in lines according to a 1988 study by Priority
Management, USA
⚫ Wait at
⚫ Traffic lights, toll booths, railway stations, bus
terminus, banks, airline check-in counters, movie
theatre, medical centers, grocery stores, airline
security, educational institutions etc.
 Waiting Realities
 Waiting results from variations in arrival rates (random)
and service rates (varying demands of customers)

 Queues form because current demand exceeds the existing

capacity to serve
 In other words, unpredictable demand give rise to conflicts
over the use of available resources
 Unevenness in the flow of customers or service requests can
be predicted very crudely, if at all
 Simply, waiting lines happens when arrivals are
bunched or when service times are lengthy, and they
are very likely to occur when both are present
 Queuing Theory

⚫ The science of waiting line is called queuing

theory

⚫ Queuing theory is a powerful tool to analyze service

systems

⚫ It helps in determining the level of service that a

company should provide (improve)
 Queuing in Service
⚫ For service oriented business, the ability to
match the available supply with the current
demand can be a major determinant of
success.
Mechanisms to Reduce Waiting Time

 Two main mechanisms used

 To increase the capacity of the service-
providing system by increasing staffing
levels

 To shift demand from peak to off-peak

periods
 Implications of Waiting Lines

⚫ Cost to provide waiting space

⚫ Loss of business
⚫ Customers leaving
⚫ Customers refusing to wait
⚫ Loss of goodwill
⚫ Reduction in customer satisfaction
⚫ Congestion may disrupt other business
operations
 Ten Principles of Waiting

1. Unoccupied time feels longer than

occupied time
2. Pre-process waits feel longer than in-
process waits
3. Anxiety makes the wait seem longer
4. Uncertain waits are longer than known,
finite waits
5. Unexplained waits seem longer than
explained waits
 Ten Principles of Waiting
6. Unfair waits are longer than equitable waits
7. The more valuable the service, the longer the
customers wait
8. Solo waiting feels longer than group waiting
9. Uncomfortable waits feel longer than
comfortable waits
10. New or infrequent users feel they wait longer
than frequent users
 Basic Elements of a Waiting Line
 The calling population (source of customers)
 Infinite
 Finite
 Queue discipline
 First come first served
 Last in first out
 Priority
 Random
 Queue Length
 Infinite (any size)
 Finite (limited in size)
 Basic Elements of Waiting Line
 Arrival rate (signified by λ, lambda)
 Average number of customers that arrive per unit of time, or “arrival
rate” of the system
 Assumed to be independent of each other and vary randomly over
time
 Described by Poisson distribution
 Balking, reneging, jockeying

 Service Rate (signified by µ, mu)

 Average number of customers that can be served by a single channel

per unit of time (also called “service rate”)

▪ For a system with M service channels, Mµ

▪ Service time described by exponential distribution
▪ Average utilization of the facility = λ / Mμ
 Queuing System - Schematic

Renege

Queue
Arrival Departure
Discipline
Process Queue Service
Calling
Configuration Process
Population

Balk No future
need
 Waiting line Models
 Patient customers
 Customers enter the waiting line and remain until
served

 Balking
 Upon arriving, decide the line is too long and decide not
to enter the line

 Reneging
 Waiting customers grow impatient and leave the line

 Jockeying
 Customers may switch to another line
Queuing System
Configurations
Phases refer to no. of sequential
Servers each customer must go through
to complete service

Channels refers to the number

Of parallel servers
 Queue Configurations Examples
Multiple Queue multichanel Single queue, multichannel

Railway reservations counters,

Supermarket cash registers

Take a Number
Enter
3 4 2 Foreign banks,
8 6 10 Airline security check-ins
Airline check-in counters
12 7
banks 11 9 VIRTUAL QUEUE
5
Banks
A Simple Queuing System
 Trade-off Between Cost of
Providing Service and Customer
Satisfaction
 Little’s Law

Lq = Avg. # units in the waiting line

Ls = Avg. # of units in the system
Wq = Avg. time a unit spends in waiting
line
Ws = avg. time a unit spends in the system

By Little’s Law, L = λ × W
Lq = λ × Wq
 Little’s Law

⚫ Under steady state conditions, the average number of

customers (Ls) in a queuing system is equal to the
average arrival rate (λ) at which the customers arrive
multiplied by the average time a customer spends in
the system (Ws)

Ls = λ×Ws
 Total Cost for a Waiting Line
 Total Cost (TC) = Service Cost + Waiting Cost
TC per hour = Cs × C + Cw × Ls
= Cs×C + Cw×λ×Ws
Where C = # servers
Cs = Hourly cost per server
Cw= Hourly cost of waiting customer
Ls = Expected # in the system
(queue + service)
Ws = Average waiting time in system
Objective: Minimize total costs
 Trade-Off
In waiting line decisions, the primary trade-off is the cost of
having too many employees (cost of service or idle time) vs.
cost of not having enough employees (cost of waiting). If
the company hires too many employees, it will have to
unnecessarily pay them to be idle most of the time. On the
other hand, if the company hires too few workers, the
company will face lost sales due to customers leaving the
store or not coming in due to an excessively long waiting
line. It is much more difficult to estimate the cost of waiting
in comparison to cost of service.
 Assumptions of Basic Single Server
Model and Multiple Server Model

⚫ Poisson arrival rate

⚫ Exponential service times
⚫ First-come, first-served
⚫ Infinite queue length
⚫ Infinite calling population
 Arrival Distribution
An arrival (arrivals per unit of time) follows a
Poisson distribution
λx e -λ
P(x) = ------------ for x = 0,1,2
X!

Where, X = number of arrivals in a specific

period
λ = average or expected # of arrivals
for the specific period
e = 2.71828
 Service Time Distribution
Service time:
Exponential probability distribution

The probability of a service time will be less than or

equal to a time length “T” is given by
P (service time t <=T) = 1 – e -µT
Where µ = service rate
t = service time of the customer

T = target service time

e = 2.71828
 Kendall’s Notation
 A popular system classifies parallel-server queuing models using
following notations in which three features are identified

 Notation: A / B / C
Where, A = probability distr. of arrivals
B = probability distr. of service time
C = number of parallel servers

Descriptive symbols used for arrival and service distr. are M, D, G

M = Poisson distr. for arrivals and exponential distr. for service times
D = Constant service time. Service times can’t be assumed
exponentially distributed. Example: Robots, automated
equipment
G = Arrival and service times have other probability
distribution with mean and variance (e.g., normal, uniform or
other empirical distribution)
 Kendall’s Notation
 M/M/1 designates a single server queuing model
 First M: Distribution of inter-arrival times is exponential (or
Poisson arrival rate)

 Second M: Distribution of service times is also exponential

“1” : It is a single server

 If multiple servers (S), then M/M/S

 Customarily, λ denotes arrival rate and µ service rate

 Standard M/M/1 Model

 M/M/1 requires following set of assumptions

 Calling population: An infinite or very large population of
callers arriving
 Arrival process: Poisson distribution
 Queuing configuration: Single waiting line with no
restrictions on length and no balking or reneging
 Queuing discipline: FCFS
 Service process: One server with negative exponential
distribution of service times
 Multiple Server, M/M/S
⚫ M/M/S: A multiple server exists when two or
more servers are working independently to
provide service to customer arrivals.
Following are the assumptions:
⚫ Poisson arrival and exponential service times
⚫ Servers work at the same average rate
⚫ Customers form a single waiting line (FCFS)
 Other Models
 Basic single Server and multiple server
models
 Single Server models with
 Constant Service Time - when automated equipment or machinery
performs the service
 The constant service time model is a special case of a more general
variation of the single-server model in which service times cannot be
assumed to be exponentially distributed. As such service time are said to
be general, or undefined.

 Finite Calling Population - when the number of “customers” that can

arrive to a system is limited

 Finite queue length – occurs when there is a physical limitation to the

length of the waiting line
Single Server – Constant Service
Time

When automated equipment or machinery performs the service

 Finite Calling Population

Example: fleet of police cars, garbage trucks, student advisor with finite no of students
assigned
 Finite Queue Length

⚫ For some waiting line systems, the length of the queue

may be limited by the physical area in which queue
forms; space may permit only a limited number of
customers to enter the queue. Such a waiting line is
called as a finite queue
⚫ Another variation of single-channel queuing model
⚫ Example: Car wash, petrol pump, ATM
Finite Queue Length
Classification of Queuing Models
 Queuing Problems

⚫ Problem Solving
 Case

Renaissance Clinic (A)