Service

Operations
Analysis
SESSION 7

HTM*3120
 Quick Review

Facility Location

Use of Geographic Information Systems

Workshop – Part I
Weekly
Topics Huff Model of Retail Location

Workshop – Part II

Supply Chain Management

Managing Service Industries

Internationalization and Outsourcing

Process
Improvement
Foundations of Continuous
Improvement
W. Edwards Deming created the Foundations of
Continuous Improvement
◦ Customer Satisfaction
◦ This requires an attitude of putting the customer first and a belief
that the principle is the object of one’s work.
◦ Management by Facts
◦ This approach requires formal data gathering and statistical
analysis of the data by the quality improvement teams.
◦ Respect for People
◦ Employees are given support and their ideas are solicited in an
environment of mutual respect.

Plan-Do-Check-Act (PDCA) Cycle

7-4
PDCA
Cycle
PDCA Cycle
Plan
◦ Selection of problems and causes are brainstormed
and, using data, agreement is reached on the root
causes. An action plan that includes a workable
solution is developed, and measures of success and
targets are agreed upon.

Do
◦ Implement the solution, even if on a trial basis.

7-6
PDCA Cycle
Check
◦ Review and evaluate the results of the change. Check
that the solution is having the intended effect and
note any unforeseen consequences.

Act
◦ Reflect and act on learning from the experience. If
successful, the process changes are standardized and
communicated to all involved workers with training.

7-7
Steps to the PDCA Cycle
Plan
◦ Define the problem and its impact – Start with identifying
what’s wrong; what’s the current state? What’s the problem
and what should it all look like in an ideal world?
◦ Understand the facts – This is the inductive problem solving
part. Watch, observe and measure the problem with data,
first.
◦ Contain the problem – As there’s a problem happening right
now, this next step is to ensure the problem doesn’t get any
bigger. If it’s a process problem, the main aim is to ensure it
never reaches the customer.
Steps to the PDCA Cycle
Plan (cont’)
◦ Measure the data – It’s now time to delve deeper into the
inductive problem-solving phase.
◦ Find the root cause to the problem – You’re now in a position
of knowledge. You’ve observed and analyzed what’s
happening. You’ve also created a list of the many potential
root causes.
Do
◦ Implement the corrective action – Here’s where you apply the
heavy lifting. You’ve looked at the problem; analyzed the data,
contained the problem and identified a root cause.
Steps to the PDCA Cycle
Check
◦ Check the data after the correction, to ensure its worked
– This is pretty self-explanatory. It’s now time to observe some
more, and to see if the solution has worked or not.

Act
◦ Standardize the process to support what you’ve learned
– Now, you must extrapolate the lessons learned and lock in
the change. In most instances, this would entail a process
change of some sort.
◦ Recognize and provide positive feedback – Celebrate a job
well done; congratulate the team and let everyone know of
the achievement.
Steps to the PDCA Cycle
Quality Tools for Analysis and
Problem Solving
◦ Check Sheet
◦ Run Chart
◦ Histogram
◦ Pareto Chart
◦ Flowchart
◦ Cause-and-Effect Diagram
◦ Scatter Diagram
◦ Control Chart
◦ Benchmarking

7-12
Check Sheet
A historical record of observations, and it
represents the source of data to begin the analysis
and problem identification.
Check Sheet of Problems faced by
an Airline
Lost Departure
Month Mechanical Overbooked Other
Luggage Delay
January 1 2 3 3 1
February 3 3 0 1 0
March 2 5 3 2 3
April 5 4 4 0 2
May 4 7 2 3 0
June 3 8 1 1 1
July 6 6 3 0 2
August 7 9 0 3 0
September 4 7 3 0 2
October 3 11 2 3 0
November 2 10 1 0 0
December 4 12 2 0 1
Total 44 84 24 16 12

7-14
Run Chart
A chart that tracks the changes in a process over
time to detect trends, shifts, or cycles in
performance.
◦ They are easy to interpret and are useful in
predicting trends.
◦ They can be used to track changes before and after
changes to processes.
Run Chart

Run Chart of
Departure
Delays
Histogram
A histogram presents data collected over a period
of time as a frequency distribution in bar chart
form.

Unusual features become obvious , such as lack of

symmetry or skewness.
Histogram

Histogram of
Lost Luggage
Pareto Chart
A pareto chart orders problems by their relative
frequency in a descending bar graph to focus
efforts on the problem that offers the greatest
potential improvement.
Pareto Chart
Flowchart
Flow charts are visual representations of the
process that can be used to identify points where
problems might occur.
Diamonds represent decision points.

Rectangles represent activities.

Ovals represent beginning and ending

points.

All Symbols are connected with

arrows to represent the sequence of
activities.
Flowchart
Flowchart
Cause-and-Effect Diagram
Cause-and-effect analysis offers a structured
approach for a team to identify, explore, and
display graphically, in increasing detail, all of the
possible causes related to a problem in order to
discover the root cause.
The cause-and-effect diagram is also known as a
fishbone chart, owing its skeletal shape, or an
Ishikawa chart, named after its originator.
 Cause-and-Effect Diagram
Cause-and-Effect Chart for Flight Departure Delay (Fishbone Chart)
Scatter Diagram
Scatter diagrams visually show the relationship
between two variables.
Plotting possible-cause variables against the
problem can identify where a strong correlation
exists.
 Scatter Diagram

Scatter Diagram of
Departure Delay vs
Late Passengers
Control Chart
Control charts are used to monitor process. Control charts help you to
see when a process is out of control.

Control Chart of
Departure
Delays
Quality
Improvement
Programs
Quality Improvement
Programs
◦ Benchmarking
◦ Deming’s 14 Point Program
◦ ISO 9000
◦ Six-Sigma
◦ Lean Service
Benchmarking
Comparison of how your company is doing through
comparison to other companies that are classified as
‘best-in-class’.
Benchmarking involves five steps:
1. Select a critical process that needs improvement;
2. Identify an organization that excels in the process;
3. Contact the benchmarked firm, make a visit and study the
process;
4. Analyze the findings;
5. Improve your process accordingly.
Deming’s 14 Point Program
1. Create constancy of purpose for improvements of product and
service.
◦ Move beyond preoccupation with next quarter and build for the future.
2. Adopt the new philosophy.
◦ Refuse to allow commonly accepted poor levels of work, delays, and lax
service.
3. Cease dependence on mass inspection.
◦ Focus on improving the process itself, instead of focusing on inspection.
4. End the practice of awarding business on price tag alone.
◦ Purchase should be made on the basis of statistical evidence of quality,
not price.
Deming’s 14 Point Program
5. Constantly and forever improve the system of production and
service.
◦ Search continually for problems in the system, and seek ways for
improvement.
6. Institute modern methods of training on the job.
◦ Restructure training to define acceptable levels of work.
7. Institute modern methods of supervising.
◦ Focus supervision on helping workers to do a better job. Provide the
tools and techniques to promote pride in one’s work.
8. Drive out fear.
◦ Eliminate fear by encouraging the communication of problems and
expression of ideas.
Deming’s 14 Point Program
9. Break down barriers between departments.
◦ Encourage problem solving through teamwork and use of quality-control
circles.
10. Eliminate numerical goals for the workforce.
◦ Goals, slogans, and posters cajoling workers to increase productivity
should be eliminated. Such exhortations cause worker resentment,
because most of the necessary changes are outside their control.
11. Eliminate work standards and numerical quotas.
◦ Production quotas focus on quantity, and they guarantee poor quality in
their attainment. Use statistical methods for continuing improvement of
quality and productivity.
Deming’s 14 Point Program
12. Remove barriers that hinder hourly workers.
◦ Workers need feedback on the quality of their work. All barriers to pride
in one’s work must be removed.
13. Institute a vigorous program of education and training.
◦ Because of changes in technology and turnover of personnel, all
employees need continual training and retraining. All training must
include basic statistical techniques.
14. Create a structure in top management that will push every day on
the above 13 points.
◦ Clearly define management’s permanent commitment to continuous
improvements in both quality and productivity.
Eureka
Ranch!
USING DEMING'S
MODEL OF
STATISTICAL
APPROACH | EXAMPLE
 “say what you do, and do

ISO 9000
what you say”

◦ The ISO 9000 is a voluntary standard.

◦ ISO is a series of quality standards defined by the International
Organization for Standardization, a consortium of the world’s
industrialized nations.
◦ Certification to an ISO 9000 standard signals that the firm has a
quality system in place that ensures consistency of output quality.
◦ Several important characteristics:
1. It does not prescribe specific practices.
2. It does not say anything directly about the quality of the product
or service itself.
3. Certification is provided by a highly decentralized system of
auditors and accreditation bodies.
ISO 9000
Documentation of processes and consistent performance
are the key features of ISO standards. ISO 9000 seeks to
achieve this by requiring that businesses implement a
three-component cycle:
1. Planning: to ensure that goals, authority, and responsibility are
planned.

2. Control: to ensure that specified requirements at all levels are

met, problems are anticipated and averted, and corrective actions
are planned and carried out

3. Documentation: to ensure an understanding of quality objectives

and methods, smooth interaction within the organization,
feedback for the planning cycle, and to serve as objective
evidence of quality system performance.
Six-Sigma
The objective of Six-Sigma is to reduce or narrow the
variation in performance to such a degree that six
standard deviations (99.7%) can be squeezed within the
limits defined by the customer’s expectations.
These limits are defined as upper specification limits
(USL) and a lower specification limit (LSL).
Six Sigma is a rigorous and disciplined methodology that
uses data and statistical analysis to measure and improve
a companies operational performance by identifying and
eliminating defects to enhance customer satisfaction.
Six-Sigma
Six Sigma Process Steps
Step Definition
Define Define project objectives, internal and external
customers.
Measure Measure current level of performance.
Analyze Determine causes of current problems.
Improve Identify how the process can be improved to
eliminate the problems.
Control Develop mechanisms for controlling the improved
process.
Lean Service
Lean Service Philosophy

1. Satisfy the needs of the customer by performing only

those activities that add value in the eyes of the
customer.

2. Define the “value stream” by flowcharting the process

to identify both value-added and non-value-added
activities.

3. Eliminate waste. Waste in the value stream is any

activity for which the customer is not willing to pay.
Lean Service
7 Steps to Achieve Lean Service
1. Identify the key processes in your organization.
2. Select the most important processes and order by
importance.
3. Analyze how the process can be changed to move toward
perfection.
4. Ask what changes will be needed to sustain the “future
state” process.
5. Implement the necessary changes to create the “future
state” process.
6. Determine what you will do with excess people and assets.
7. Start the cycle again.
Facility Location –
New Lesson
Geographic Representation
Location on a Plane
Y

A Plane is any
Destination j
representation in
two-dimensional
space: a surface

Origin i

0 X
Euclidean Distance

Location on a Plane
Y

Destination j
Yj Euclidean distance
(as the Crow flies)

The shortest
Yi
distance between
Origin i
two points is a
straight line.

0 X
Xi Xj
Euclidian Distance
Utilizing the Pythagorean theorem to measure Euclidian distance, we
get:

Where
dij = distance between points i and j
xi, yi = coordinates of the ith point
xj, yj – coordinates of the jth point
Euclidean Distance.. As the crow flies
Example of Euclidean
Distance Calculation
Calculate the shortest distance between two points with
the following coordinates:
X Y
Origin 7 9
Destination 1 1

Answer = [(7-1)2 + (9-1) 2]½

= [(6)2 + (8) 2]½
= [36 + 64]½ = [100] ½ = 10
Metropolitan Distance

Location on a Plane
Y

Destination j
Yj

Yi Origin i

Metropolitan Distance

0 X
Xi Xj
Example of Metropolitan distance..

END

START

To travel from one point to another, you must proceed on a grid-like route.
Example of Metropolitan distance..

END

START

Regardless of the route, the distance is the same

so long as you maintain direction towards the destination.
Example of Metropolitan
Distance Calculation
Calculate the shortest distance between two points with the following
coordinates:

X Y
Origin 5 1
Destination 2 8

Metropolitan Distance = |5 – 2| + |1 - 8| = 10
Comparing Euclidean and
Metropolitan Distances
X Y
Origin 3 8
Destination 5 3

Euclidean = [(3-5)2 + (8-3)2]1/2

= (29)1/2 = 5.385

Metropolitan = |3-5| + |8-3|

= 7
Workshop
Q1
Other Geographic
Representations
Networks

Time to travel from point to point is often more important than

euclidean distance due to geographic obstacles or traffic
Other Geographic
Representations
Radians (vectors and angles)

Primarily Used by
aircraft, ships, radar
A vector is a distance
from an origin,
whose direction is
defined by an angle.
Huff Model of
Retail Location
Huff Model of Retail Location
A quantitative model that predicts retail performance from
geo-demographic data

Gravity analog: attraction (between planets) is directly

proportional to size and inversely proportional to distance

A very established retail model developed in 1966 and still

used today!

Ideally suited for spreadsheets

Huff Model Assumptions
A consumers attraction to a particular retail location is
a measure of preference.

1. Consumers prefer larger retail facilities to smaller

retail facilities. Size Matters!
◦ Attraction is directly proportional to size.

2. Consumers prefer closer retail facilities to facilities

that are located far away
◦ Attraction is inversely proportional to distance
 1. Attractiveness of a Location (j)
Gravity analogy is used to estimate
attractiveness of store of size S at
a location j for customers in area i.
Aij= Attraction to store j for customers in area i
Sj = Size of the store (e.g. square feet)
Tij= Travel time from area i to store j
λ = Parameter reflecting propensity to travel
 Example: The Data
The Jay store offers a CD exchange service
There is an existing “Jay” store in the West End, and a
proposed new “Jay” store (Stone Road):

Store J Size (Sj)

Stone Road (New) 1 3,000 square feet (3k)
West End (existing) 2 1,000 square feet (1k)
 Calculate the Probabilities
Areas where residents live (i)

Measurement Store j Sj Downtown Uptown East End West End

(i=1) (i=2) (i=3) (i=4)

Stone Road (j=1,New) 3

Travel Times
West End (j=2, Existing) 1
Stone Road (j=1,New)
Attractiveness West End (j=2, Existing)
Total
Stone Road (j=1,New)
Probability (Pij) West End (j=2, Existing)
Total
 The Data
There are four main areas where students live: downtown,
uptown, residence and west end
Travel times (minutes by car) from
each living area to each store (Tij)

Areas where students live (i)

Downtown Uptown Residence West

Store j Sj
(i=1) (i=2) (i=3) (i=4)
Stone Road (j=1,New) 3 8 6 1 7
West End (j=2, Existing) 1 5 7 7 1
 Calculate the Probabilities
Areas where residents live (i)

Measurement Store j Sj Downtown Uptown East End West End

(i=1) (i=2) (i=3) (i=4)

Stone Road (j=1,New) 3 8 6 1 7

Travel Times
West End (j=2, Existing) 1 5 7 7 1
Stone Road (j=1,New)
Attractiveness West End (j=2, Existing)
Total
Stone Road (j=1,New)
Probability (Pij) West End (j=2, Existing)
Total
 1. Attractiveness of a Location (j)
Gravity analogy is used to estimate
attractiveness of store of size S at a
location j for customers in area i.
Aij= Attraction to store j for customers in area i
Sj = Size of the store (e.g. square feet)
Tij= Travel time from area i to store j
λ = Parameter reflecting propensity to travel
Calculate Aij
(the Attractiveness of store j for a consumer living in area i)

For now, set λ = 1 (Tijλ = Tij when λ= 1)

Calculate for each cell ij, the value of A

Areas where students live (i)

Downtown Uptown Residence West

Store j Sj
(i=1) (i=2) (i=3) (i=4)
(j=1,New) 3 8 6 1 7

Attractiveness 3 0.3750 0.5000 3.0000 0.4286

(j=2, Existing) 1 5 7 7 1

Attractiveness 1 0.2000 0.1429 0.1429 1.0000

 Calculate the Probabilities
Areas where residents live (i)

Measurement Store j Sj Downtown Uptown East End West End

(i=1) (i=2) (i=3) (i=4)

Stone Road (j=1,New) 3 8 6 1 7

Travel Times
West End (j=2, Existing) 1 5 7 7 1
Stone Road (j=1,New) 0.3750 0.5000 3.0000 0.4286
Attractiveness West End (j=2, Existing) 0.2000 0.1429 0.1429 1.0000
Total
Stone Road (j=1,New)
Probability (Pij) West End (j=2, Existing)
Total
 Total attractiveness
of all stores to consumers living in
a particular area i

Add up the attractiveness values for all stores, j= 1, 2, … for

region i
Calculate Aij
(the Attractiveness of store j for a consumer
living in area i)

Calculate the sum Aij for each area i

Attractiveness Matrix Areas where students live (i)

Downtown Uptown Residence West

Store j Sj
(i=1) (i=2) (i=3) (i=4)
(j=1,New) 3 0.3750 0.5000 3.0000 0.4286
(j=2, Existing) 1 0.2000 0.1429 0.1429 1.0000
Total Attractiveness 0.5750 0.6429 3.1429 1.4286
 Calculate the Probabilities
Areas where residents live (i)

Measurement Store j Sj Downtown Uptown East End West End

(i=1) (i=2) (i=3) (i=4)

Stone Road (j=1,New) 3 8 6 1 7

Travel Times
West End (j=2, Existing) 1 5 7 7 1
Stone Road (j=1,New) 0.3750 0.5000 3.0000 0.4286
Attractiveness (Aij) West End (j=2, Existing) 0.2000 0.1429 0.1429 1.0000
Total 0.5750 0.6429 3.1429 1.4286
Stone Road (j=1,New)
Probability (Pij) West End (j=2, Existing)
Total
2. Probability that a customer
from area i will go to store j
If you live downtown (i=1), The total attractiveness of going
to a J store is 0.5750
The attractiveness of going to the west end (existing) store
is 0.200
Then we conclude, the probability of going to the west end
store is 0.20 ÷ 0.575 = 34.8%
The attractiveness of going to the new
store is 0.3750
Then we conclude, the probability of
going to the new Stone Road store is
0.375 ÷ 0.575 = 65.2%
Calculate Pij
(the Attractiveness of store j for a consumer
living in area i)

Calculate the Pij for Areas where students live (i)

each area i
Residenc
Downtown Uptown West
Store j Sj e
(i=1) (i=2) (i=4)
(i=3)
(j=1,New) 3 0.3750 0.5000 3.0000 0.4286

(j=2, Existing) 1 0.2000 0.1429 0.1429 1.0000

Total Attractiveness 0.5750 0.6429 3.1429 1.4286

Probabilty Pij
(j=1,New) 3 0.6522 0.7778 .9545 0.3000
(j=2, Existing) 1 0.3478 0.2222 0.0455 0.7000
1.0000 1.0000 1.0000 1.0000
Probability that customer
from area i will go to store j
Probability that customer from
area i will go to store j

Probability Matrix Areas where students live (i)

Downtown Uptown Residence West

Store j Sj
(i=1) (i=2) (i=3) (i=4)
(j=1,New) 3 0.6522 0.7778 .9545 0.3000

(j=2, Existing) 1 0.3478 0.2222 0.0455 0.7000

Probability 1.000 1.000 1.000 1.000

 Calculate the Probabilities
Areas where residents live (i)

Measurement Store j Sj Downtown Uptown East End West End

(i=1) (i=2) (i=3) (i=4)

Stone Road (j=1,New) 3 8 6 1 7

Travel Times
West End (j=2, Existing) 1 5 7 7 1
Stone Road (j=1,New) 0.3750 0.5000 3.0000 0.4286
Attractiveness (Aij) West End (j=2, Existing) 0.2000 0.1429 0.1429 1.0000
Total 0.5750 0.6429 3.1429 1.4286
Stone Road (j=1,New) 0.6522 0.7778 0.9545 0.3000
Probability (Pij) West End (j=2, Existing) 0.3478 0.2222 0.0455 0.7000
Total 1.0000 1.0000 1.0000 1.0000
A little GeoDemographic data….
◦ The number of students living in each area (Ci)
◦ The annual expenditure on CD exchange by area Bik (subscript k refers
to the product or service sold)
◦ Total Area Expenditure (Eik = CiBik) = area pop. x Avg. Annual Spending

Areas where students live (i)

Downtown Uptown Residence West

(i=1) (i=2) (i=3) (i=4)
Student Population (Ci) 2,000 4,000 9,000 2,000

Avg. Annual Expenditure (Bik) $175 $225 $160 $300

Total Area Spend on Eik $350,000 $900,000 $1,440,000 $600,000

 Total Demand for product k…
Just add up the spending for each area!
$3,290,000 Market Size

Areas where students live (i)

Downtow Uptown Residence West
n (i=1) (i=2) (i=3) (i=4)
Population (Ci) 2,000 4,000 9,000 2,000
Annual Spend (Bik) $175 $225 $160 $300
Area Spend on Eik $350,000 $900,000 $1,440,000 $600,000 $3,290,000
 Spending
Areas where residents live (i)
Total/
Measurement Store j Sj Downtown East End West End
Uptown (i=2) Average
(i=1) (i=3) (i=4)
Stone Road (j=1,New) 3 8 6 1 7
Travel Times
West End (j=2, Existing) 1 5 7 7 1
Stone Road (j=1,New) 0.3750 0.5000 3.0000 0.4286
Attractiveness
West End (j=2, Existing) 0.2000 0.1429 0.1429 1.0000
(Aij)
Total 0.5750 0.6429 3.1429 1.4286
Stone Road (j=1,New) 0.6522 0.7778 0.9545 0.3000
Probability (Pij) West End (j=2, Existing) 0.3478 0.2222 0.0455 0.7000
Total 1.0000 1.0000 1.0000 1.0000
Student Population (Ci) 2,000 4,000 9,000 2,000
GeoDemographi Avg. Annual Expenditure (Bik) $175 $225 $160 $300
c Information
Total Area Spend on Eik 350000 900000 1440000 600000
East End (j=1,New)
Sales (Ejk) West End (j=2, Existing)
Total
East End (j=1,New)
Market Share
West End (j=2, Existing)
(Mjk)
Total
Predicting the sales by store…
Annual Sales = the sum for all areas of the probability of a
customer going to the store X the annual expenditure X number
of customers.

Pij = Prob. of customer from area i travelling to store j

Ci = Number of customers in area i
Bik = Annual budget for product k for customers in area i
m = Number of customer areas in the market region (4)

Expected Sales for j =

 Predicting the sales by store…
Sales Predictions

Areas where students live (i)

Downtown Uptown Residence West
(i=1) (i=2) (i=3) (i=4)
Population (Ci) 2,000 4,000 9,000 2,000
Annual Spend (Bik) $175 $225 $160 $300
Area Spend on Eik $350,000 $900,000 $1,440,000 $600,000 $3,290,000
Probability for j=1 0.6522 0.7778 0.9545 0.3000
Sales for Store j=1 228,270 700,020 1,374,480 180,000 $2,482,770
Probability for j=2 0.3478 0.2222 0.0455 0.7000
Sales for Store j=2 121,730 199,980 65,520 420,000 $807,230
 Sales
Areas where residents live (i)
Total/
Measurement Store j Sj Downtown East End West End
Uptown (i=2) Average
(i=1) (i=3) (i=4)
Stone Road (j=1,New) 3 8 6 1 7
Travel Times
West End (j=2, Existing) 1 5 7 7 1
Stone Road (j=1,New) 0.3750 0.5000 3.0000 0.4286
Attractiveness
West End (j=2, Existing) 0.2000 0.1429 0.1429 1.0000
(Aij)
Total 0.5750 0.6429 3.1429 1.4286
Stone Road (j=1,New) 0.6522 0.7778 0.9545 0.3000
Probability (Pij) West End (j=2, Existing) 0.3478 0.2222 0.0455 0.7000
Total 1.0000 1.0000 1.0000 1.0000
Student Population (Ci) 2,000 4,000 9,000 2,000
GeoDemographi Avg. Annual Expenditure (Bik) $175 $225 $160 $300
c Information
Total Area Spend on Eik 350000 900000 1440000 600000
East End (j=1,New) 228261 700000 1374545 180000 2482806
Sales (Ejk) West End (j=2, Existing) 121739 200000 65455 420000 807194
Total 350000 900000 1440000 600000 3290000
East End (j=1,New)
Market Share
West End (j=2, Existing)
(Mjk)
Total
What are the Market Shares?
Market share = Store sales ÷ Total Sales

Area Spend on Eik $3,290,000 Market share

Sales for Store j=1 $2,482,770 75.46%
Sales for Store j=2 $807,230 24.54%
 Market Share
Areas where residents live (i)
Total/
Measurement Store j Sj Downtown East End West End
Uptown (i=2) Average
(i=1) (i=3) (i=4)
Stone Road (j=1,New) 3 8 6 1 7
Travel Times
West End (j=2, Existing) 1 5 7 7 1
Stone Road (j=1,New) 0.3750 0.5000 3.0000 0.4286
Attractiveness
West End (j=2, Existing) 0.2000 0.1429 0.1429 1.0000
(Aij)
Total 0.5750 0.6429 3.1429 1.4286
Stone Road (j=1,New) 0.6522 0.7778 0.9545 0.3000
Probability (Pij) West End (j=2, Existing) 0.3478 0.2222 0.0455 0.7000
Total 1.0000 1.0000 1.0000 1.0000
Student Population (Ci) 2,000 4,000 9,000 2,000
GeoDemographi Avg. Annual Expenditure (Bik) $175 $225 $160 $300
c Information
Total Area Spend on Eik 350000 900000 1440000 600000
East End (j=1,New) 228261 700000 1374545 180000 2482806
Sales (Ejk) West End (j=2, Existing) 121739 200000 65455 420000 807194
Total 350000 900000 1440000 600000 3290000
East End (j=1,New) 65.22% 77.78% 95.45% 30.00% 75.47%
Market Share
West End (j=2, Existing) 34.78% 22.22% 4.55% 70.00% 24.53%
(Mjk)
Total 100.00% 100.00% 100.00% 100.00% 100.00%
 What About λ (Lambda)?

Perception of time passing varies by the nature of the

activity being done

If a trip is enjoyable, the time passes quickly (i.e. going to a

“special” mall in Las Vegas…

If a trip is not enjoyable, the time passes very slowly (i.e.

running to the store to get nuts for the cookies you are
baking seems to take forever)
Lambda Values
Recall we set λ = 1; then Tij1 = Tij

Assume the real time for a trip is Tij = 4 hours.

1. For an enjoyable trip… lambda = 0.5 means the trip

feels like 2 hours! (4½ = 2)
2. For an unpleasant trip… lambda = 2.0 means the trip
feels like 16 hours! (42 = 16)
Workshop
Q2
Supply Chain
Management
 What is Supply Chain
Management
The integration of the activities that procure materials and services,
transform them into intermediate goods and final products, and deliver
them to customers.
1. Transportation vendors, 6. Warehousing and inventory,
2. Credit and cash transfers, 7. Order fulfillment,
3. Suppliers, 8. Sharing customers, forecasting, and
4. Distributors, producing information
5. Accounts payable and receivable,

Picture Source: https://twitter.com/NextCraftBeer/status/927606842250878976/photo/1

 How Supply Chain Decisions
Affect Strategy
Supply Chain Low-Cost Strategy Response Strategy Differentiation Strategy
Decision
Supplier’s goal Supply demand at lowest Respond quickly to changing Share market research; jointly
possible cost requirements and demand to develop products
minimize stockouts
Primary selection Select primarily for cost Select primarily for capacity, Select primarily for product
criteria speed, and flexibility development skills
Process Maintain high average Invest in excess capacity and Use modular processes that lend
characteristics utilization flexible processes themselves to mass
customization
Inventory Minimize inventory Develop responsive system, Minimize inventory in the chain
characteristics throughout the chain to with buffer stocks positioned to avoid obsolescence
hold down costs to ensure supply
Lead-time Shorten lead time as long Invest aggressively to reduce Invest aggressively to reduce
characteristics as it does not increase production lead time development lead time
costs
Product-design Maximize performance Use product designs that lead Use modular design to postpone
characteristics and minimization costs to low setup time and rapid product differentiation for as long
production ramp-up as possible
Managing Supply Chain Risk
To reduce risk, management must manage:

Process: raw material and component availability, quality, and logistics

Controls: management metrics and reliable secure communication for
financial transactions, product designs, and logistic scheduling
Environment: customs duties, tariffs, security screening, natural
disaster management, current fluctuations, terrorist attacks, and
political issues
Supply-Chain Strategy
◦ Many Suppliers
◦ Few Suppliers
◦ Vertical Integration
◦ Joint Ventures
◦ Keiretsu
◦ Virtual companies
Supply-Chain Strategy
◦ Many Suppliers
◦ Requests for quotes are sent to many suppliers, and the
company decides between all responses (considerations – cost,
quality, and delivery competencies)
◦ Suppliers all compete with one another

◦ Few Suppliers
◦ The forming of long-term relationships with a few suppliers.
 Supply-Chain Strategy
◦ Vertical Integration
◦ Backward integration
◦ A firm purchases its
suppliers
◦ Forward integration
◦ A firm purchases its
buyers

https://www.investopedia.com/ask/answers/051315/what-difference-between-horizontal-integration-and-
vertical-integration.asp
Supply-Chain Strategy
◦ Joint Ventures
◦ Working with others in your industry through formal
collaboration.
◦ Can help to reduce costs or secure supply.

◦ Keiretsu
◦ Part purchasing from few suppliers and part vertical integration.
◦ Manufacturers are often financial supporters through
ownership or loans.
◦ Creates long-term relationships.
Supply-Chain Strategy
◦ Virtual companies
◦ Use of flexible suppliers.
◦ Fluid, moving, organizational boundaries that allow them to
create a unique enterprise to meet changing market demands.
◦ Advantages include
◦ Specialized management expertise
◦ Low capital investment
◦ Flexibility
◦ Speed
◦ Overall, a more efficient system.
Managing the Supply Chain
◦ Factors for success:
◦ Mutual agreement on goals
◦ Trust
◦ Compatible organizational cultures
Managing
Inventories
Distribution System
Role of Inventory in Services
Decoupling inventories
Seasonal inventories
Speculative inventories
Cyclical inventories
In-transit inventories
Safety stocks

15-98
Role of Inventory in Services
Decoupling inventories
◦ Each stage in the distribution channel has their own stores of information,
and their own replenishment levels.
◦ The time between when an order is placed and when it is received is
referred to as the replenishment lead time

Seasonal inventories
◦ High stocking of inventories may take place at specific times of the year, in
advance of a high predicted demand.

15-99
Role of Inventory in Services
Speculative inventories
◦ When a service anticipates a significant increase in the price of a specific
good, they may stock up on that good at the lower price. The strategy of
managing speculative inventories is known as forward buying.

Cyclical inventories
◦ Refers to the normal variations in inventory levels (stock after inventory
received – stock just before inventory received)

15-100
Role of Inventory in Services
In-transit inventories
◦ Stock that has been ordered but is yet to be received.

Safety stocks
◦ Maintenance of inventory stock that is above expected demand.

15-101
Considerations in Inventory
Systems
◦ Independent versus dependent demand

◦ Type of customer demand

◦ Planning time horizon

◦ Replenishment lead time

◦ Constraints and relevant costs

15-102
Considerations in Inventory
Systems
Independent versus dependent demand
◦ Independent – demands of different items not related to
each other (e.g. bread)
◦ Dependent – demand of one item is related to demand for
another (e.g. ketchup)

Type of customer demand

◦ Trends, cycles, seasonality

Planning time horizon

◦ For how long will a specific item be needed (parkas vs.
shorts)

15-103
Considerations in Inventory
Systems
Replenishment lead time
◦ The time between when an order is placed and when an
order will be received.

Constraints and relevant costs

◦ Storage (space, refrigeration, etc.)
◦ Labour
◦ Insurance

15-104
Relevant Inventory Costs
Ordering costs
Receiving and inspection costs
Holding or carrying costs
Shortage or stock-out costs
Depreciation of value

15-105
Relevant Inventory Costs
Ordering costs
◦ Preparing specifications for items to be purchased
◦ Locating and identifying potential suppliers and soliciting bids
◦ Evaluating bids and selecting suppliers
◦ Negotiating prices
◦ Preparing purchase orders
◦ Issuing or transmitting purchase orders to outside suppliers
◦ Following up to ensure that purchase orders are received by suppliers

15-106
Relevant Inventory Costs
Receiving and inspection costs
◦ Transporting, shipping and pickup
◦ Preparing and handling records of receipts and other paperwork
◦ Examining packages for visible damage
◦ Unpacking items
◦ Counting or weighing items to ensure that the correct amount has been
delivered
◦ Inspecting or testing items to ensure that they conform to purchase
specification
◦ Transferring items into storage area

15-107
Relevant Inventory Costs
Holding or carrying costs
◦ Interest charges on money invested in inventories
◦ Opportunity cost of capital tied up in inventory
◦ Theft or pilferage
◦ Security systems
◦ Breakage, damage, and spoilage
◦ Depreciation
◦ Storage space and facilities
◦ Controlling environments
◦ Managing (e.g. supervising, taking inventory, verifying records, etc.)

15-108
Relevant Inventory Costs
Shortage or stock-out costs
◦ Lost sales and profits
◦ Customer dissatisfaction and ill-will; lost customers
◦ Penalties for late delivery or non-delivery
◦ Expediting orders to replenish exhausted stock
Depreciation of value

15-109
Inventory Management
Questions
What should be the order quantity (Q)?
When should an order be placed, called a reorder point (ROP)?
How much safety stock (SS) should be maintained?

15-110
Outsourcin
g
Outsourcing Process
Need Identification
- Problem definition
-Do-vs-Buy analysis
Performance
- Involve interested parties
Evaluation
-Specification Development
- Identify Evaluator
- Quality of Work Information Search
- Communication - References
Vendor Selection
- Meet Deadlines - Personal Contact
- Experience
- Flexibility -Recommendations
- Reputation
- Dependability
- References
- Cost
- Location
- Size
Benefits of Outsourcing
◦ Allows the firms to focus on core competencies
◦ Decreases costs by purchasing from an outside source
rather than performing in-house.
◦ Provides access to latest technology without
investment.
◦ Leverages benefits from a supplier who has economies
of scale.
Risks of Outsourcing
◦ Loss of direct control over quality
◦ Jeopardizes employee loyalty because of job-loss fears
◦ Exposure to data security and customer privacy issues
◦ Dependence on one supplier compromises future
negotiation leverage
◦ Additional coordination expense and delays
◦ Atrophy of in-house capability to perform outsourced
service
Question
s
Next Week
Session 8
◦ Managing Capacity and Demand

◦ Scheduling

◦ Overbooking
Service Operations
Analysis
HTM*3120
WEEK 8 WORKSHOP
- WORKSHOP QUESTIONS
Managing Capacity and
Demand
Managing Capacity and Demand
Variable and Unpredictable demand is characteristic of
most services

Customer induced variability

◦ Arrival rates are not uniform
◦ Capability of customers varies.. Some require more time to
make decisions, or understand
◦ Requests from customers vary and result in non-uniform cycle
times
◦ Customer variation in the amount of effort they are willing to
expend to be part of the service process
◦ Customer’s variation in preferences result in differences in
cycle time
Generic Strategies
1. Level Capacity (Shift demand)
◦ Operate a service at a constant rate
◦ Try to influence customers to shift their business from busy periods to slow periods
2. Chase Demand (Adjust Capacity)
◦ Adjust operations for changes in customer demand
◦ Labour scheduling, adjustable capacity or capacity sharing
Strategies for Matching
Capacity and Demand for
Services
MANAGING MANAGING
DEMAND CAPACITY

Segmenting Increasing
demand customer
Developing participation
Sharing
complementary
capacity
services
Offering
Scheduling
price
Reservation Cross- work shifts
incentives
systems and training
Overbooking employees
Promoting Creating
off-peak adjustable
Using
Customer- demand capacity
part-time
induced employees
variability
Yield
management

11-5
Level Capacity (Shift
demand)
Customer-Induced Variability
Arrival: customer arrivals are independent decisions not evenly spaced.
Capability: level of knowledge and skills vary resulting in some hand-holding.
Request: uneven service times result from unique demands.
Effort: level of commitment to coproduction or self-service varies.
Subjective Preference: personal preferences introduce unpredictability.

11-7
Strategies for Managing
Customer-induced Variability
Type of Accommodation Reduction
Variability
Arrival Provide generous staffing Require reservations

Capability Adapt to customer skill Target customers based on

levels capability
Request Cross-train employees Limit service breadth

Effort Do work for customers Reward increased effort

Subjective Diagnose expectations and Persuade customers to adjust

Preference adapt expectations

11-8
Level Capacity
Consideration for leveling Capacity:

Segmenting demand
◦ business vs. leisure travelers
◦ price-sensitive vs. time sensitive
◦ walk-ins vs. appointments
Level Capacity
What actions can be taken to level capacity
beyond segmenting?

◦ price discounts to build volume during slow periods

(early-bird discounts)
◦ Promote off-peak times (weekend stay-over for
business clients)
◦ Add complementary services to capture excess
demand (bar food menu, sell drinks in the waiting area)
 Daily Scheduling of
Telephone Operator Work
shifts

Topline profile

Scheduler program assigns

tours so that the number of
operators present each
half hour adds up to the
number required

Tour

12 2 4 6 8 10 12 2 4 6 8 10 12 12 2 4 6 8 10 12 2 4 6 8 10 12

11-11
Scheduling
Chase (react to) Demand
Daily work shift scheduling
◦ Forecasting demand
◦ Calculate corresponding labour hours
◦ Increase / decrease schedule with full or part time employees

Actions
◦ Use of computer-optimized scheduling (Linear programming)
◦ Labour Scheduling Histogram Heuristic
Labour Scheduling
Histogram Heuristic
Example:
MacDonald’s restaurant has the following requirements for front-line servers
based on a demand forecast for the afternoon shift (4:00 PM – Midnight).
Employees may only work 3 shifts per week:

DAY Servers Time Servers

Monday 2 Friday 5
Tuesday 2 Saturday 4
Wednesday 3 Sunday 2
Thursday 5
 1. Find the number of Employees
DAY Servers Time Servers
Monday 2 Friday 5
Tuesday 2 Saturday 4
Wednesday 3 Sunday 2
Thursday 5

There are a total of 2+2+3+5+5+4+2 = 23 shifts needed.

23 ÷ 3 shift maximum for part-timers = 7.67 ~ 8
8 employees are needed to cover the afternoon shifts.
The maximum number of employees needed on any shift is 5
2. Construct the Histogram
1. Rows = max number of employees
needed
Number of employees required
5 2. Columns: Put Days in order from
highest to lowest number of shifts
4
3
2
1
Thur. Fri. Sat. Wed. Mon. Sun. Tue.
(5) (5) (4) (3) (2) (2) (2)

4:00 PM – Midnight Shift Days

 3. Start filling in 8 employees, A-H
Begin with employee
Number of employees required 5 E A in the lowest left
4 D hand corner of the
histogram, then B,
3 C H etc. Repeat after H.
2 B G
1 A F
Thur. Fri. Sat. Wed. Mon. Sun. Tue.

Start here! 4:00 PM – Midnight Shift Days

 4. Continue filling in 8 employees, A-H
Continue with
Number of employees required 5 E B employee A in after
4 D A F H, going to the next
column in turn
3 C H E
2 B G D H
1 A F C G
Thur. Fri. Sat. Wed. Mon. Sun. Tue.

Start here! 4:00 PM – Midnight Shift Days

 5. Finish filling in 8 employees, A-H
Number of employees required 5 E B Stop when the last
cell in the last
4 D A F column has been
filled.
3 C H E A
2 B G D H C E G
1 A F C G B D F
Thur. Fri. Sat. Wed. Mon. Sun. Tue.

Start here! 4:00 PM – Midnight Shift Days

5 E B 6. Transfer the assignments
4
3
D
C
A
H
F
E A
to a Work Schedule
2 B G D H C E G
1 A F C G B D F
Thu Fri Sat Wed Mon Sun Tue

Employee MON TUE WED THU FRI SAT SUN Total

A X X X 3
B X X X 3
C X X X 3
D X X X 3
E X X X 3
F X X X 3
G X X X 3
H X X 2
Totals 2 2 3 5 5 4 2 23
 7. Assign Employees to shifts

Employee MON TUE WED THU FRI SAT SUN Total

Adam X X X 3

Barb X X X 3
Cathy X X X 3
Dave X X X 3
Elisabeth X X X 3
Fred X X X 3
Gord X X X 3
Helen X X 2
Totals 2 2 3 5 5 4 2 23
Workshop
Part One:
Submission
Scheduling Part-time Bank Tellers
Tom Bailey is the Branch Manager at the Old Building Savings and Loan.
Tom is trying to schedule for next weeks operation. He has two full time
staff members and a few part timers. Utilizing the forecast below,
determine how many part timers Tom will have to schedule, and compose
a schedule for Tom.

7 5 6
Tellers required
2 3 4 1
0

Mon. Tues. Wed. Thurs. Fri.

Overbooking
Reservations Systems
Reservations are an integral part of service industries with fixed
capacity
Examples are hotels, restaurants, car rentals, cruise ships, doctor’s
offices, stadiums, etc.
The problem is customers who fail to show up, with the result that a
seat that could have been sold is empty and managers lose money
To counter-act the loss, managers overbook!
Overbooking as an Inventory
Management Strategy

Walk(ed)

To relocate a guest with a confirmed reservation

at a hotel to an alternative property
Overbooking as an Inventory
Management Strategy
The Case for Overbooking
Overbooking can be caused by
Damaged rooms
Staff errors
Inventory availability errors
Guest overstays
Overbooking as an Inventory
Management Strategy
The Case for Overbooking
Experienced RMs know it is best not to walk:

• Members of their hotel or brand’s loyalty

(frequent guest) programs
• Group meeting or event attendees
• Contracted rooms such as airline crew rooms
• Couples celebrating special occasions
• Families arriving late at night
Overbooking as an Inventory
Management Strategy
The Case for Overbooking

Ethical issues

Legal issues

Property reputation (in the Social Networking age)

RevPAR generation
 The Overbooking / No-show problem
“LIAT airlines serving Jonestown, Guyana is considering
overbooking its flights to avoid flying with empty seats. The
ticket agent is thinking of taking 7 reservations for an airplane
that has only 6 seats. He plans to give $50 cash
compensation, plus a seat on the next flight to anyone whose
reservation is not honoured.” Is this the best overbooking
policy?

The operating costs associated with each flight are:

◦ Pilot: $150; Co-Pilot $ 100; Fuel $30; Landing Fee $20.
◦ A one-way ticket sells for $80.
Overbooking Example
LIAT Overbooking Problem
Recent Data: No-Shows 0 1 2 3 4
Percentage 30% 25% 20% 15% 10%

Maximum revenue = 6 seats X $80 = $480

Cost of the flight: 150 + 100 + 30 + 20= $300
Maximum Net Income = $180

All flight costs are fixed: If the plane flies, all costs are
incurred.
Method 1: The Payoff Matrix
No- Prob. Number of passengers Overbooked
Shows 0 1 2 3 4
0 30% 180
1 25% 180
2 20% 180
3 15% 180
4 10% 180
Expected Profit

Step 1: When # overbooks = # no-shows, we earn $180.

 LIAT Overbooking Problem
Assume we do not overbook:
(and there are some no shows)

If we have 1 no-show:
Maximum Net Income = $180
Revenue Loss due to no-show = 80
Net Income = $100

If we have 2 no-shows:
Maximum Net Income = $180
Revenue Loss due to no-shows = 160
Net Income = $ 20
Method 1: The Payoff Matrix
No- Prob. Number of passengers Overbooked
Shows 0 1 2 3 4
0 30% 180
1 25% 100 180
2 20% 20 180
3 15% -60 180
4 10% -140 180
Expected Profit

Step 2: For each no-show > overbook, we lose $80.

Method 1: The Payoff Matrix
No- Prob. Number of passengers Overbooked
Shows 0 1 2 3 4
0 30% 180
1 25% 100 180
2 20% 20 100 180
3 15% - 60 20 100 180
4 10% -140 -60 20 100 180
Expected Profit

Continue Step 2:
For each no-show > overbook, we lose another $80.
 LIAT Overbooking Problem
Assume everyone shows up
(including the overbooked passengers (OP))

If we have 1 passenger overbooked:

Maximum Net Income = $180
Cash Payment to OP = 50
Net Income = $130

If we have 2 passengers overbooked:

Maximum Net Income = $180
Cash Payment to OP = 100
Net Income = $ 80
Method 1: The Payoff Matrix
No- Prob. Number of passengers Overbooked
Shows 0 1 2 3 4
0 30% 180 130 80 30 -20
1 25% 100 180
2 20% 20 100 180 130
3 15% -60 20 100 180
4 10% -140 -60 20 100 180
Expected Profit

Step 3: For each overbook > no-show, we lose $50.

Method 1: The Payoff Matrix
No- Prob. Number of passengers Overbooked
Shows 0 1 2 3 4
0 30% 180 130 80 30 -20
1 25% 100 180 130 80 30
2 20% 20 100 180 130 80
3 15% -60 20 100 180 130
4 10% -140 -60 20 100 180
Expected Profit

Continue Step 3: For each overbook > no-show, we lose

$50.
Method 1: The Payoff Matrix
No- Prob. Number of passengers Overbooked
Shows 0 1 2 3 4
0 30% 180 130 80 30 -20
1 25% 100 180 130 80 30
2 20% 20 100 180 130 80
3 15% -60 20 100 180 130
4 10% -140 -60 20 100 180
Expected Profit 60.00 101.00 109.50 92.00 55.00

Step 4: Expected profit = Σ (probability X payoff)

i.e. 30% X 180 + 25% X 100 + 20% X 20 - 15% X 60 -10% X 140 = $60
Best Overbooking policy!
No- Prob. Number of passengers Overbooked
Shows 0 1 2 3 4
0 30% 180 130 80 30 -20
1 25% 100 180 130 80 30
2 20% 20 100 180 130 80
3 15% - 60 20 100 180 130
4 10% -140 -60 20 100 180
Expected Profit 60.00 101.00 109.50 92.00 55.00

Best Policy is to overbook by 2 passengers.

The ticket agent should book 8 people per flight.
Method 2: Critical Fractile
When does the problem of overbooking happen?
When demand (d) is greater than the capacity (x).

When is the service provider happy?

When the expected revenue is greater than the expected
loss!!

Let CU = Cost of underbooking 1 passenger

= $80

Let CO = Cost of overbooking 1 passenger

= $50
 Method 2: Critical Fractile
Expected revenue > Expected loss
E (revenue on the last sale) ≥ Expected (loss on the last sale)

Then:
CU X Prob (d≥x) ≥ CO X Prob (d<x)
CU X [1 - Prob (d<x)] ≥ CO X Prob (d<x)
Prob (d<x) ≤ CU
Cu = Cost of underbooking
CU + CO Co = Cost of overbooking
CU
The ratio CU + CO is known as the critical fractile.
Method 2: Critical Fractile
Let CU = Cost of losing 1 passenger
= $80 (lost fare due to underbooking) )

Let CO = Cost of overbooking 1 passenger

= $50 (cash payment)

CU = 80 = 0.6154 = 61.5%
CU + CO 80 + 50
Method 2: Critical Fractile
No- Prob. Cum
shows Prob.

0 30% 30%
1 25% 55%
2 20% 75% 61.5%

3 15% 90%
4 10% 100%
Expected Profit
1. Add a Cumulative probability column. (d<X)
2. Critical Fractile = 61.5% (between 75% and 55%)
3. Answer: Go to next higher cumulative probability (75%)
Method 2: Critical Fractile
No need to calculate the payoff table…
Go right to the solution! No Prob. Cum
Shows Prob.
Steps:
1. Prepare cumulative probability 0 30% 30%
distribution 1 25% 55%
2. Calculate critical fractile
2 20% 75%
3. Solution corresponds to next 3 15% 90%
Higher cumulative probability
4 10% 100%
Workshop Part II :
Critical Fractile
Hotel example!
A small country inn is considering adopting an overbooking policy
due to a history of no-shows in the busy summer season. A
vacant room represents a loss of $69. However, providing a
confirmed guest a room at a nearby hotel costs $119. The hotel
still receives their room rate but pays the partner hotel the full
$119.

No Shows 0 1 2 3
Frequency 4 3 2 1
What is the optimal overbooking policy?
- Calculate using Critical Fractal Method and Payoff Matrix Method
Workshop Part III :
Critical Fractile
Service Shop example!
Jennies Quickie Lube has been noticing several instances
when customers do not show up for their service
appointments. She is also turning away customers due to
completely booked schedules. If a no-show costs Jennie
$100, and an overbook costs her a fee oil change ($20),
what should Jennie do?
No Shows 0 1 2 3
Frequency 10 30 20 10
What is the optimal overbooking policy?
- Calculate using Critical Fractal Method and Payoff Matrix Method
Critical Fractile Method
1. Prepare a cumulative probability distribution
No Shows 0 1 2 3 Total
Frequency 10 30 20 10 70
Probability 14% 43% 29% 14% 100%
Cum. Prob 14% 57% 86% 100%

2. Calculate the critical fractile:

3. The solution is ?
Queuing
Systems
Essential Features of Queuing Systems

Renege

Arrival Queue
process Departure
Calling Queue discipline Service
population configuration process
Balk No future
need for
service

12-53
Essential Features of Queuing Systems

Renege

Arrival Queue
process Departure
Calling Queue discipline Service
population configuration process
Balk No future
need for
service

12-54
Arrival Process
Arrival
process

Static Dynamic

Random Random arrival Customer-

Facility- exercised
arrivals with rate varying
controlled control
constant rate with time

Accept/Reject Price Appointments Reneging Balking

12-55
Variation in Arrival Rates
Ambulance Calls Per Hour Patient Arrivals at Health Clinic

12-56
Poisson and Exponential Equivalence
Poisson distribution for number of arrivals per hour (top view)

One-hour
1 2 0 1 interval
Arrival Arrivals Arrivals Arrival

62 min.
40 min.
123 min.

Exponential distribution of time between arrivals in minutes (bottom view)

12-57
Essential Features of Queuing Systems

Renege

Arrival Queue
process Departure
Calling Queue discipline Service
population configuration process
Balk No future
need for
service

12-58
Queue Configurations

12-59
Queue Configurations
Single-queue
◦ Advantages:
◦ The arrangement guarantees fairness by
ensuring that a first-come, first-served rule
(FCFS) applies to all.
◦ There is a single queue; thus, no anxiety is
associated with waiting to see if one
selected the fastest line.
◦ With only one entrance at the rear of the
queue, the problem of cutting-in is resolved
and reneging made difficult.
◦ Privacy is enhanced because the
transaction is conducted with no one
standing immediately behind the person
being served.
◦ This arrangement is more efficient in terms
of reducing the average time that
customers spend waiting in line.

12-60
Queue Configurations

Multiple-queue
◦ Advantages:
◦ Service provided can be differentiated.
◦ Ex. Supermarkets and lane 1
◦ Division of labour is possible.
◦ Season pass vs first time
◦ The customer has the option of selecting a
particular server of preference.
◦ Balking behavior may be deterred.

12-61
Queue Configurations

Take a Number
◦ Advantages:
◦ No need for a formal line.
◦ Customers are free to wander, possibly
increasing sales.

12-62
Essential Features of Queuing Systems

Renege

Arrival Queue
process Departure
Calling Queue discipline Service
population configuration process
Balk No future
need for
service

12-63
Queue Discipline
A queue discipline is a policy established by management pertaining to how to select the next
customer from the queue.

© 2011 JOHN WILEY AND SONS, INC.. ALL RIGHTS RESERVED

 Queue Discipline
Queue
discipline

Static
Dynamic
(FCFS rule)

selection Selection based

based on status on individual
of queue customer
attributes

Number of Processing time

customers Round robin Priority Preemptive of customers
waiting (SPT or cµ rule)

12-65
Essential Features of Queuing Systems

Renege

Arrival Queue
process Departure
Calling Queue discipline Service
population configuration process
Balk No future
need for
service

12-66
Service Facility Arrangements
Service facility Server arrangement
Parking lot Self-serve

Cafeteria Servers in series

Toll booths Servers in parallel

Supermarket Self-serve, first stage; parallel servers, second stage

Hospital Many service centers in parallel and series, not all used by each
patient

12-67
Queuing Analysis
Queuing Formulas
1. Mean arrival rate:
2. Mean service rate:
3. Mean number in service:
4. Probability of exactly “n” customers in the system:
5. Probability of “k” or more customers in the system:
6. Mean number of customers in the system: Ls = Lq + p (M/M/c)
Ls = / ( - ) (M/M/1)
7. Mean number of customers in queue:
(M/M/1)

Lq = see values table using C and p (M/M/c)

8. Mean time in system: (M/M/1)

Ws = Ls (M/M/c)

9. Mean time in queue: (M/M/1) Wq = Lq (M/M/c)

 Standard M/M/1 Model
The standard M/M/1 model has five assumptions:
1. Calling population: an infinite or very large population of callers arriving. The
callers are independent of each other and not influenced by the queueing system
(e.g., an appointment is not required) = Single Server
2. Arrival process: Negative exponential distribution of interarrival times or Poisson
distribution of arrival rate.
3. Queue configuration: Single waiting line with no restrictions on length and no
balking or reneging.
4. Queue discipline: FCFS
5. Service process: One server with negative exponential distribution of service
times.
Standard M/M/1 Model
Lake Travis has one launching ramp near the dam for people who trailer
there small boats to the recreational site. A study of cars arriving with
boats in tow indicates a Poisson distribution with a mean rate of = 6
boats per hour during the morning launch. A test of the data collected
on launch times suggests that an exponential distribution with a mean
of 6 minutes per boat (equivalent service rate = 10 boats launched
per hour) is a good fit.
What is the average number of customers in service (the launch)?

What is the probability of having at least 1 customer in the system upon

arrival?

What is the probability of the ramp being available to pull right in?

What is the average number of boats in the system?

Standard M/M/1 Model
Lake Travis has one launching ramp near the dam for people who trailer
there small boats to the recreational site. A study of cars arriving with
boats in tow indicates a Poisson distribution with a mean rate of = 6
boats per hour during the morning launch. A test of the data collected
on launch times suggests that an exponential distribution with a mean
of 6 minutes per boat (equivalent service rate = 10 boats launched
per hour) is a good fit.
What is the average number of customers in service (the launch)?
= 6/10 = 0.6

What is the probability of having at least 1 customer in the system upon

arrival?

What is the probability of the ramp being available to pull right in?

What is the average number of boats in the system?

Standard M/M/1 Model
Lake Travis has one launching ramp near the dam for people who trailer
there small boats to the recreational site. A study of cars arriving with
boats in tow indicates a Poisson distribution with a mean rate of = 6
boats per hour during the morning launch. A test of the data collected
on launch times suggests that an exponential distribution with a mean
of 6 minutes per boat (equivalent service rate = 10 boats launched
per hour) is a good fit.
What is the average number of customers in service (the launch)?

= 6/10 = 0.6
What is the probability of having at least 1 customer in the system upon
arrival?
◦ = ( / )1 = (6/10)1 = 0.6

What is the probability of the ramp being available to pull right in?

What is the average number of boats in the system?

Standard M/M/1 Model
Lake Travis has one launching ramp near the dam for people who trailer
there small boats to the recreational site. A study of cars arriving with
boats in tow indicates a Poisson distribution with a mean rate of = 6
boats per hour during the morning launch. A test of the data collected
on launch times suggests that an exponential distribution with a mean
of 6 minutes per boat (equivalent service rate = 10 boats launched
per hour) is a good fit.
What is the average number of customers in service (the launch)?

= 6/10 = 0.6
What is the probability of having at least 1 customer in the system upon
arrival?
◦ = ( / )1 = (6/10)1 = 0.6

What is the probability of the ramp being available to pull right in?
P0 = pn (1-p) = 0.60(1 – 0.6) = 1(0.4) = 0.4

What is the average number of boats in the system?

Standard M/M/1 Model
Lake Travis has one launching ramp near the dam for people who trailer
there small boats to the recreational site. A study of cars arriving with
boats in tow indicates a Poisson distribution with a mean rate of = 6
boats per hour during the morning launch. A test of the data collected
on launch times suggests that an exponential distribution with a mean
of 6 minutes per boat (equivalent service rate = 10 boats launched
per hour) is a good fit.
What is the average number of customers in service (the launch)?
= 6/10 = 0.6
What is the probability of having at least 1 customer in the system upon
arrival?
◦ = ( / )1 = (6/10)1 = 0.6

What is the probability of the ramp being available to pull right in?
P0 = pn (1-p) = 0.60(1 – 0.6) = 1(0.4) = 0.4

What is the average number of boats in the system?

Ls = / ( - ) = 6 / (10-6) = 1.5 boats
 Standard M/M/c Model
The standard M/M/c model has five assumptions:
1. Calling population: an infinite or very large population of callers arriving. The
callers are independent of each other and not influenced by the queueing system
(e.g., an appointment is not required)
2. Arrival process: Negative exponential distribution of interarrival times or Poisson
distribution of arrival rate.
3. Queue configuration: Single waiting line with no restrictions on length and no
balking or reneging.
4. Queue discipline: FCFS
5. Service process: servers across all channels are independent and equal (all servers
are considered equal).
 Standard M/M/c Model
The Dean of CBE has assigned a secretary to each of her four departments.
Each secretary assists the faculty with class material and correspondence.
The Dean received complaints from the Economics faculty, about delays in
getting work accomplished. After collecting information on the requests of
each secretary, it was found that secretarial work arrives with a Poisson
distribution at an average rate of = 2 requests per hour for all
departments except economics. Economics has an average rate of = 3.
The average time to complete a piece of work is 15 minutes ( = 4 per
hour). The current system is essentially four M/M/1 operations.
What is the average turnaround for the Economics secretaries as compared
to the other three?
 Standard M/M/c Model
The Dean of CBE has assigned a secretary to each of her four departments.
Each secretary assists the faculty with class material and correspondence.
The Dean received complaints from the Economics faculty, about delays in
getting work accomplished. After collecting information on the requests of
each secretary, it was found that secretarial work arrives with a Poisson
distribution at an average rate of = 2 requests per hour for all
departments except economics. Economics has an average rate of = 3.
The average time to complete a piece of work is 15 minutes ( = 4 per
hour). The current system is essentially four M/M/1 operations.
What is the average turnaround for the Economics secretaries as compared
to the other three?
Economics = = 1 / (4-3) = 1 hour

Other Three = = 1 / (4-2) = 0.5 hour

 Standard M/M/c Model
The Dean of CBE has assigned a secretary to each of its four departments. Each
secretary assists the faculty with class material and correspondence. The Dean
received complaints from the Economics faculty, about delays in getting work
accomplished. After collecting information on the requests of each secretary, it
was found that secretarial work arrives with a Poisson distribution at an
average rate of = 2 requests per hour for all departments except economics.
Economics has an average rate of = 3. The average time to complete a piece
of work is 15 minutes ( = 4 per hour). The current system is essentially four
M/M/1 operations.
What would be the average number of customers in the system if the Dean was
to pool all four secretaries?
 Standard M/M/c Model
The Dean of CBE has assigned a secretary to each of its four departments. Each
secretary assists the faculty with class material and correspondence. The Dean
received complaints from the Economics faculty, about delays in getting work
accomplished. After collecting information on the requests of each secretary, it
was found that secretarial work arrives with a Poisson distribution at an
average rate of = 2 requests per hour for all departments except economics.
Economics has an average rate of = 3. The average time to complete a piece
of work is 15 minutes ( = 4 per hour). The current system is essentially four
M/M/1 operations.
What would be the average number of customers in the system if the Dean was
to pool all four secretaries?
Ls = Lq + p

=2+2+2+3= 9 = 9 / 4 = 2.25 c=4 Lq = ?

Ls = Lq + p
= ? + 2.25
Lq for
Various
Values of
C and
 Standard M/M/c Model
The Dean of CBE has assigned a secretary to each of its four departments. Each
secretary assists the faculty with class material and correspondence. The Dean
received complaints from the Economics faculty, about delays in getting work
accomplished. After collecting information on the requests of each secretary, it
was found that secretarial work arrives with a Poisson distribution at an
average rate of = 2 requests per hour for all departments except economics.
Economics has an average rate of = 3. The average time to complete a piece
of work is 15 minutes ( = 4 per hour). The current system is essentially four
M/M/1 operations.
What would be the average number of customers in the system if the Dean was
to pool all four secretaries?
Ls = Lq + p
= 0.31 + 2.25
= 2.56 per hour

What would be the average time in the system?

 Standard M/M/c Model
The Dean of CBE has assigned a secretary to each of its four departments. Each
secretary assists the faculty with class material and correspondence. The Dean
received complaints from the Economics faculty, about delays in getting work
accomplished. After collecting information on the requests of each secretary, it
was found that secretarial work arrives with a Poisson distribution at an
average rate of = 2 requests per hour for all departments except economics.
Economics has an average rate of = 3. The average time to complete a piece
of work is 15 minutes ( = 4 per hour). The current system is essentially four
M/M/1 operations.
What would be the average number of customers in the system if the Dean was
to pool all four secretaries?
Ls = 2.56 per hour

What would be the average time in the system?

Ws = Ls = 2.56 / 9 = 0.28 hour = 0.28 x 60 = 17 m
Workshop IV
Queue Features

Calling
Customer Counter Garage Car Wash Example
Population Q1. Thrifty Car Rental is looking at its products
Arrival and services. They have asked you to break
Process down the queue features for three of their
Queue
service offerings.
Configuration
Queue
Discipline
Service
Process
Foto Mat Example
Q2. The Foto-Mat is looking at their service processes. They have told you that on average 2 customers
arrive per hour at the Foto-Mat to process film. The one clerk in attendance spends an average of 15
minutes per customer. Given this information answer the following questions:
Q2.a. what is the arrival rate?
Q2.b. what is the service rate?
Q2.c. what is the average number of customers in service?
Q2.d. what is the probability of having at least 2 customers in the system upon arrival?
Q2.e. what is the average number of customers in the system?
Q2.f. what is the average number of customers in the queue?
Q2.g. what is the average waiting time in queue?
Airport Example
Q3. The Airport Runway is looking at its service times. The airports runway receives approximately 20
arrivals per hour, and each plane takes approximately 2 minutes to take off. Given this knowledge answer
the following questions.
Q3.a. Arrival rate =
Q3.b. Service rate =
Q3.c. What is the average number of customers in service?
Q3.d. What is the probability of having at least 3 planes in the queue upon arrival?
Q3.e. What is the average number of planes in the system and average number of planes in the queue?
Q3.f. What is the average waiting time the planes must spend in queue?
Q3.g. What is the average time that planes spend in the system?
Q3.h. What is the average time that planes spend in the system?
Questions
Service Operations
Analysis
HTM*3120
SESSION 9
Important Notes
1. Final Test opens March 19 and closes April 1, 2025.
1. Consists of:
1. 48 -52 Points
1. MC = 10 Points
2. Fill in the blank
Weekly Topics
Queuing Analysis Review

Measuring Forecasting Error

• Sum of the Absolute Deviations (SAD)

• Mean of the Absolute Deviations (MAD)
• Mean Absolute Percentage Error (MAPE)
• Mean Forecast Error (MFE)
• Mean Squared Error (MSE)

Forecasting Techniques

Exponential Smoothing
 Example
Q1. Thrifty Car Rental is looking at its products
Queue Customer Counter Garage Car Wash
Features

Calling
Population
People Cars (things) Cars (things)
and services. They have asked you to break down
Arrival Facility Controlled (Appointments) Facility Controlled (Appointments) Static (Random arrival the queue features for three of their service
offerings.
Process rate varying with time)

Queue Multiple Queues Single Queue Single Queue

Configuration

Queue Dynamic (Selection based on Dynamic (selection based on status of Static (FCFS)
Discipline individual customer Attributes – queue - # customers waiting)
priority/pre-booking)

Service Servers in Parallel 2 stage (1st – dropping off car, 2nd – car Self-Serve
Process serviced, 3rd – car picked up)
Queuing Formulas
1. Mean arrival rate:
2. Mean service rate:
3. Mean number in service:
4. Probability of exactly “n” customers in the system:
5. Probability of “k” or more customers in the system:
6. Mean number of customers in the system: Ls = Lq + p (M/M/c)
Ls = / ( - ) (M/M/1)
7. Mean number of customers in queue:
(M/M/1)

Lq = see values table using C and p (M/M/c)

8. Mean time in system: (M/M/1)

Ws = Ls (M/M/c)

9. Mean time in queue: (M/M/1) Wq = Lq (M/M/c)

Foto Mat Example
Q2. The Foto-Mat is looking at their service processes. They have told you that on average 2 customers
arrive per hour at the Foto-Mat to process film. The one clerk in attendance spends an average of 15
minutes per customer. Given this information answer the following questions:
Q2.a. what is the arrival rate?
Q2.b. what is the service rate?
Q2.c. what is the average number of customers in service?
Q2.d. what is the probability of having at least 2 customers in the system upon arrival?
Q2.e. what is the average number of customers in the system?
Q2.f. what is the average number of customers in the queue?
Q2.g. what is the average waiting time in queue?
Airport Example
Q3. The Airport Runway is looking at its service times. The airports runway receives approximately 20
arrivals per hour, and each plane takes approximately 2 minutes to take off. Given this knowledge answer
the following questions.
Q3.a. Arrival rate =
Q3.b. Service rate =
Q3.c. What is the average number of customers in service?
Q3.d. What is the probability of having at least 3 planes in the queue upon arrival?
Q3.e. What is the average number of planes in the system and average number of planes in the queue?
Q3.f. What is the average waiting time the planes must spend in queue?
Q3.g. What is the average time that planes spend in the system?
Demand
Forecasting
Importance of Demand
Forecasting

Demand (customers)
The number of potential buyers with the
interest and ability to purchase the
products sold by a business at the specific
price offered.

9
Importance of Demand
Forecasting

Accurate forecasts are

important for:

Scheduling Purchasing Managing

workers supplies cash flow

10
Importance of Demand
Forecasting
To forecast accurately, we use three types of data:

Type of Information Describes

Events that have already
Historical data
occurred
Events occurring now or in
Current data
the very near term
Events that will occur in the
Future data
future

11
Historical Data
Historical Data
Common historical data tracked by RMs include

Number of reservations/ room nights booked per

day

Number of reservations/ room nights denied per

day

13
Historical Data
Common historical data tracked by RMs include

Number of daily reservation cancellations

Total number of room nights canceled
Number of check-ins (arrivals)
Number of check-outs (departures)

14
Historical Data

Common historical data tracked by RMs include

No-shows

Walk-ins

ADR achieved

15
Historical Data
Common historical data tracked by RMs include

Occupancy % achieved
• By the property
• By room type
Average number of guests per room
Average length of guest stay
16
Forecasting
RULE #1:
EVERY FORECAST IS
WRONG!!!
(ESPECIALLY IF IT’S ABOUT THE FUTURE)
Techniques for
Measuring Forecasting
Errors
How Wrong Can You Be?
(measuring forecast error)
Absolute Deviation
Sum of the Absolute Deviations (SAD)
Mean of the Absolute Deviations (MAD)
Mean Absolute Percentage Error (MAPE)
Mean Forecast Error (MFE)
Mean Squared Error (MSE)
Absolute Deviations
1. Absolute Deviation = | At – Ft |
2. Sum of the Absolute Deviations
SAD = Σ | At – Ft |
◦ Used when the total amount of error in a model is important

3. Mean of the Absolute Deviations

MAD = Σ | At – Ft | ÷ n
◦ Used to compare models with different numbers of periods, or
when the average deviation is important
Absolute Deviations
4. Mean Absolute Percentage Error:

A t – Ft 100
MAPE = Σ At X n

◦ Used when the proportion of error is important

 Measuring forecast bias..
5. Mean Forecast Error
MFE = Σ (At – Ft ) = 0 n
◦ Forecasts should not be biased positively or negatively.
◦ The sum of the positive and negative errors should balance out to = 0 (or close to
zero…)
◦ If MFE is ≠ 0, the forecast is biased and on average the forecast will be greater
(or less than) the actual amount.
The effect of larger errors…
Mean Squared Error:

MSE = Σ (At – Ft )2
n
◦ squaring makes a large number proportionately larger than a
small number
◦ i.e. 22 = 4, 32 = 9, 42 = 16, 52 = 25, etc.
◦ MSE is a good measure if managers don’t sweat the small stuff,
but are concerned with the larger forecast errors
Example: Calculating SAD & MAD
Actual vs
Forecast
t At Ft Abs.
t At Ft Dev.
1 136 150 1 136 150 14
2 151 160 2 151 160 9
3 185 165 20
3 185 165
4 144 150 6
4 144 150
5 127 150 23
5 127 150 6 183 160 23
6 183 160 7 172 165 7
7 172 165 8 158 160 2 104
8
SAD 104
8 158 160
MAD 13
Example: Calculating MAPE
Actual vs
Forecast t At Ft Abs. Abs.
Dev. % error |151 -160 |
t At Ft
1 136 150 14 10.29 151
1 136 150
2 151 160 9 5.96
2 151 160
3 185 165 20 10.81
3 185 165
4 144 150 6 4.17
4 144 150
5 127 150 23 18.11
5 127 150
6 183 160 23 12.57
6 183 160
7 172 165 7 4.07
7 172 165
8 158 160 2 1.27 67.25
8 158 160
Σ Percentage error 67.25 8

MAPE 8.41
Example: Calculating MFE
Actual vs
t At Ft Dev.
Forecast (error)
t At Ft 1 136 150 -14
1 136 150 2 151 160 -9
2 151 160 3 185 165 +20
3 185 165 4 144 150 -6
4 144 150 5 127 150 -23
5 127 150 6 183 160 +23
6 183 160 7 172 165 +7
7 172 165 8 158 160 -2 -4.0
8 158 160 Σ Forecast errors -4.0 8

MFE -0.5
Example: Calculating MSE
Actual vs
Forecast
t At Ft Dev. Dev2
t At Ft (error) (error2)
1 136 150 1 136 150 -14 196
2 151 160 2 151 160 -9 81
3 185 165 3 185 165 +20 400
4 144 150 4 144 150 -6 36
5 127 150 5 127 150 -23 529
6 183 160 6 183 160 +23 529
7 172 165 7 172 165 +7 49
8 158 160 8 158 160 -2 4
Σ Squared error 1,824
MSE 228
Forecasting Techniques
Forecasting Techniques
1. Delphi Technique
◦ Expert panels (managers, staff and external advisors)
2. Associative Methods
◦ Regression
3. Time Series
◦ Moving average (MA)
◦ Exponential smoothing (S)
◦ Trend adjustment (T)
◦ Seasonal adjustment (I)
Delphi Techniques
◦ Expert panels (managers, staff and external advisors)
◦ The use of a series of well constructed questionnaires collected from a panel of experts in
their field.
◦ Objectives of the Delphi Technique/Method:
1. “To determine or develop a range of possible program alternatives;
2. To explore or expose underlying assumptions or information leading to different judgments;
3. To seek out information which may generate a consensus on the part of the respondent group;
4. To correlate informed judgments on a topic spanning a wide range of disciplines, and;
5. To educate the respondent group as to the diverse and interrelated aspects of the topic” (Delbecq,
Van de Ven, and Gustafson, 1975, p. 11).
Associative Methods
◦ Regression
◦ Multiple Regression
 Simple Linear Regression (SLR)
Regression is fitting a line to data points using statistical methods to get the
best fit.
1 = Δy / Δx
y = 0 + 1 x, where: = slope
0 = intercept
y
data point (x,y)
Δy
a0
Δx
0 a1
x
If we know the value of parameters 0 and 1, we
can calculate a forecasted value of y for any x.
Multiple Linear Regression (MLR)
Same as simple Linear regression, except the line is in n-dimensional
space (multiple independent variables)
y = 0 + 1x1 + 2x2 + 3x3 + …..
SLR Example:
A local restaurant suspects that sales are related to local employment, as measured in area
payroll data (in Billions of dollars).

Year
2005 2006 2007 2008 2009 2010

Sales (millions) 2.0 3.0 2.5 2.0 2.0 3.5

Payroll (billions) 1 3 4 2 1 7

a0
y = 0 + 1 x, where:
a1
SLR Example:
A local restaurant suspects that sales are related to local employment,
as measured in area payroll data (in Billions of dollars).

Year
2005 2006 2007 2008 2009 2010

Sales (millions) 2.0 3.0 2.5 2.0 2.0 3.5

Payroll (billions) 1 3 4 2 1 7
Using this information, we can ask excel to calculate for us the slope
and y intercept.
Slope = linset(y’s, x’s)
Intercept = intercept(y’s, x’s)
Forecast:
Chamber of Commerce Poll: Local industry estimates $6 billion in payroll for fiscal Year 2011.
What is the forecasted sales for 2011?
Parameters: 0 = 1.75; 1 = 0.25

y= 0 + 1 x
= 1.75 + (0.25 x 6)
= $3.25 Million
MLR Example
A local restaurant suspects that sales (millions) are related to local employment, as measured in
payroll data (in Billions of dollars) and mortgage interest rates

Year
2005 2006 2007 2008 2009 2010

Sales (millions) 2.0 3.0 2.5 2.0 2.0 3.5

Payroll (billions) 1 3 4 2 1 7
Interest Rate 2% 3% 7% 9% 8% 4%
MLR Example
Chamber of Commerce Poll for fiscal 2011:
◦ Local industry estimates $6 billion in area payroll (x1)
◦ Financiers set Mortgage Interest at 3%. (x2)

What is the forecasted sales for 2011?

Parameters: 0 = 2.13; 1 = 0.24; 2 = -6.2

y = 0 + 1 x1 + 2 x2
= 2.13 + (0.24 x 6) - (6.2 x .03)
= $3.38 Million
Problems with Regression Approaches
Great when causative factors are known with certainty…

But often based on other estimates.. For example, area payroll and mortgage interest rates,
weather, etc. are all forecasts that could be wrong.. Or simply unpredictable
Time Series Models
◦ Moving Averages
◦ Exponential Smoothing
Time Series Forecasting
Based on the idea that the past is our best predictor of the future

Data collected over a sequence of evenly spaced time periods has a lot of information:
◦ Yesterday’s performance

◦ Trends over longer periods

◦ Cyclical (weekly or seasonal variations)

A naive forecasting
rule….
Last month’s sales were pretty good.

Let’s go proceed on the basis that next month’s sales will be

the same as the last.
What happened last month is the best predictor of
what will happen this month….
Month Period Sales Forecast Dev.
t At Ft (error)
Jan 1 10 ~ ~
Feb
Mar
2
3
12
13
10
12
2
1
Ft = At-1
Apr 4 16 13 3
May 5 19 16 3
Jun 6 23 19 4
Jul 7 26 23 3
Aug 8 30 26 4 FFeb = AJan
Sep 9 28 30 2
Oct 10 18 28 10
Nov 11 16 18 2
Dec 12 14 16 2
SAD 36
MAD 3.27
What is going on…
We are using 100% of the previous period’s data to estimate the future

We are ignoring all prior periods…

Let’s try this….

Take an average of the last 3 periods and use it to forecast the next period..

This is called a 3-period moving average

Time Series Model –
Moving Average
(N-Period Moving Average)
Let : MAT = The N period moving average at the end of period T
AT = Actual observation for period T

Then: MAT = (AT + AT-1 + AT-2 + …..+ AT-N+1)/N

Next period forecast = F t+1 = MAt

Characteristics:
Needs N observations to make a forecast
Very inexpensive and easy to understand
Gives equal weight to all observations
Does not consider observations older than N periods
The average of the past 3 months is the best predictor of
what will happen next month….

Ft = MAt-1
Period Sales 3 Month Forecast Dev.
t At Average Ft (error)
1 10 ~ ~ ~
2 12 ~ ~ ~
MA3 = 10 + 12 + 13
3 13 11.67 ~ ~
3
4 16 13.67 11.67 4.33
5 19 16.00 13.67 5.33
6 23 19.33 16.00 7.00 MA4 = 12 + 13 + 16
7 26 22.67 19.33 6.67
3
8 30 26.33 22.67 7.33
9 28 28.00 26.33 1.67
10 18 25.33 28.00 10.00 MA10 = 30 + 28 + 18
11 16 20.67 25.33 9.33 3
12 14 16.00 20.67 6.67
SAD 58.33
MAD 6.48
What is going on…
We are using the 3 preceding period data to estimate the future
Every period has the same weight or importance

Let’s try this….

Weight the most recent period X 3
Weight the second period X 2
Weight the oldest period X 1
Total is divided by: 6
The weighted average of the past 3 months is the best
predictor of what will happen this month….
Period Sales 3 Month Forecast Dev.
t
1
At
10
WMAt
~
Ft
~
(error)
~
Ft = WMAt-1
2 12 ~ ~ ~
3 13 12.17 ~ ~ WMA3
4 16 14.33 12.17 3.83 = 10 + (2 X 12) + (3 X 13)
5 19 17.00 14.33 4.67 6
6 23 20.50 17.00 6.00
7 26 23.83 20.50 5.50
8 30 27.50 23.83 6.17 WMA4
9 28 28.33 27.50 0.50 = 12 + (2 X 13) + (3 X 16)
10 18 23.33 28.33 10.33 6
11 16 18.67 23.33 7.33
12 14 15.33 18.67 4.67
SAD 49.00
MAD 5.44
Historical Data
Trailing period
A data collection method characterized by the act of discarding the
oldest piece of data in a data set when the newest data are added,
thus updating the set’s information while keeping the set size
constant. Data contained in a trailing period are often used in
calculating a rolling average.

49
Moving averages
How long a period should I use?

What are benefits of different lengths?

Where does weighting fit in?

Moving averages
Longer time periods
◦ Smooth out extremes

Shorter time periods

◦ Reflect changing circumstances more effectively

There are other things we can do

◦ More than one – long and short
◦ Recognize trends
◦ Incorporate seasonality
Moving Averages
Day 1 87 Day 21 79 Day 60 86
Day 2 88 Day 22 89 Day 61 83 Average of
Day 3 89 Day 23 86 Day 62 81
Last 10 Days
Day 4 85 Day 24 84 Day 63 80
Day 5 86 Day 25 81 Day 64 83 83.9%
Day 6 87 Day 26 80 Day 65 84
Day 7 79 Day 27 82 Day 66 87
Day 8 89 Day 28 85 Day 67 88
Day 9 86 Day 29 86 Day 68 89 Average of
Day 10 84 Day 30 83 Day 69 85 Last 20 Days
Day 11 81 Day 31 81 Day 70 86 84.25%
Day 12 80 Day 32 80 Day 71 87
Day 13 82 Day 33 83 Day 72 79
Day 14 85 Day 34 84 Day 73 89
Day 15 86 Day 35 87 Day 74 86 Average of
Day 16 83 Day 36 88 Day 75 84
All Cases
Day 17 81 Day 37 89 Day 76 81
Day 18 80 Day 38 85 Day 77 80 84.25%
Day 19 83 Day 39 86 Day 78 82
Day 20 84 Day 40 87 Day 79 85
Calculate Moving Averages for:
Mondays Rooms Sold 5-day
1 82
3-day
2 82
3 82 3-day weighted (20%, 30%, 50%)
4 82
5 82
6 82
7 84
8 86
9 87
10 86
Calculate Moving Averages for:
Mondays Rooms Sold 5 Monday Moving Average
1 82
Week 6 = (86+87+86+84+82)/5 = 85.0
2 82
3 82 Week 5 = (87+86+84+82+82)/5 = 84.2
4 82
Week 4 = (86+84+82+82+82)/5 = 83.2
5 82
6 82 Week 3 = (84+82+82+82+82)/5 = 82.4
7 84 Week 2 = (82+82+82+82+82)/5 = 82.0
8 86
9 87 Week 1 = (82+82+82+82+82)/5 = 82.0
10 86
Calculate Moving Averages for:
Mondays Rooms Sold 3 Monday Moving Average
1 82
Week 8 = (86+87+86)/3 = 86.33
2 82
Week 7 = (87+86+84)/3 = 85.67
3 82
4 82 Week 6 = (86+84+82)/3 = 84.00

5 82 Week 5 = (84+82+82)/3 = 82.67

6 82 Week 4 = (82+82+82)/3 = 82.00
7 84 Week 3 = (82+82+82)/3 = 82.00
8 86 Week 2 = (82+82+82)/3 = 82.00
9 87 Week 1 = (82+82+82)/3 = 82.00
10 86
Calculate Moving Averages for:
Mondays Rooms Sold 3 Monday Weighted Moving Average
1 82 (50% - 30% - 20%)
2 82 Week 8 = 86x20%+87x30%+86x50% = 86.3
3 82
Week 7 = 87x20%+86x30%+84x50% = 85.2
4 82
Week 6 = 86x20%+84x30%+82x50% = 83.4
5 82
Week 5 = 84x20%+82x30%+82x50% = 82.4
6 82
7 84 Week 4 = 82x20%+82x30%+82x50% = 82.0

8 86 Week 3 = 82x20%+82x30%+82x50% = 82.0

9 87 Week 2 = 82x20%+82x30%+82x50% = 82.0
10 86 Week 1 = 82x20%+82x30%+82x50% = 82.0
Part II - Simple
Exponential Smoothing
Some Forecasting Terms
At = Actual value at time t
Ft = Forecasted value at time t
St = Smoothed value at time t

t=1 period 1
t=7 period 7
n = the number of periods in the model

If today is period 1 and tomorrow is period 2:

t = t + 1 tomorrow (next period)
t = t – 1 yesterday (previous period)
 Simple Exponential
Smoothing
◦ New Forecast
= Last period’s forecast
+  (Last Period’s Actual demand – Last Period’s forecast)

Ft = Ft-1 +  (At-1 – Ft-1),  is a parameter.

◦ (At-1 – Ft-1) is the deviation or error from the previous period

◦ We are adding back a proportion of the forecast error to “even out” or smooth the
forecast for the next period
◦ The proportion added back is , where 0 <  < 1
◦  is the exponential smoothing parameter
 Simple Exponential
Smoothing
To recognize the smoothed forecast calculation, S is introduced into the formula
instead of F:

This:
Ft = Ft-1 +  (At-1 – Ft-1)

Becomes:
St = St-1 +  (At – St-1)
and Ft+1 = St

(S becomes the forecast for the next period)

Here’s how it works….
Given: Ft = Ft-1 +  (At-1 – Ft-1)
Add 1 period: Ft+1 = Ft +  (At – Ft)

Recall: Ft+1 = St
and thus: Ft = St-1

Then: substituting for Ft+1 and Ft :

St = St-1 +  (At – St-1)
Simple Exponential Smoothing
For convenience,

This: St = St-1 +  (At - St-1)

Is algebraically the same as:

St =  At + (1- ) St-1
 An example calculation of St:
St =  At + (1- ) St-1
Period Sales
t At St
…. ..... ….
 = 0.7
2 12 11.40
3 13 12.52 Calculate S3, then S4
4 16 14.96

S3 =  A 3 + (1- ) S3-1 S4 =  A4 + (1- ) S4-1

= 0.7 X A3 + (1- 0.7) X S2 = 0.7 X A4 + (1- 0.7) X S3
= 0.7 X 16 + (0.3) X 12.52
= 0.7 X 13 + (0.3) X 11.40 = 11.20 + 3.76
= 9.1 + 3.42 = 14.96
= 12.52
Simple Exponential Smoothing with  = 0.7
The full Monty!
Period Sales Forecast Dev.
t At St Ft (error) We set St=1 = At=1
1 10 10** ~ ~ to start since
2 12 there is no prior
3 13 data available.
4 16
5 19
6 23

Ft = St-1
7 26
8 30
9 28
10 18
11 16
12 14
Here’s how it’s done!!
Period Sales Forecast Dev.
t At St Ft (error) We set S1 = A1
1 10 10** ~ ~ to start since
2 12 10.0 there is no prior
3 13 data available.
4 16
5 19

Set Ft = St-1
6 23
7 26
8 30 = 10
9
10
28
18
Then.. F2 = S1
11 16 = 10
12 14
SAD
MAD
Here’s how it’s done!!
Period Sales Forecast Dev.
t At St Ft (error) We calculate the
1 10 10** ~ ~ Value of S2 using the
2 12 11.40 10.0 formula:
3 13 11.40 St =  At + (1- ) St-1
4 16 S2 = 0.7 X 12
5 19 + (0.3 ) X 10
6 23
= 11.40
7 26
8 30
9
10
28
18
Ft = St-1
11 16 F3 = S2 = 11.40
12 14
SAD
MAD
That was fun.. Let’s do it again!!
Period Sales Forecast Dev.
t At St Ft (error) We calculate the
1 10 10** ~ ~ Value of S3 using the
2 12 11.40 10.0 formula:
3 13 12.52 11.40 St =  At + (1- ) St-1
4 16 12.52 S3 = 0.7 X 13
5 19 + (0.3 ) X 11.40
6 23
7 26 = 12.52
8 30
9 28
10 18
Ft = St-1
F4 = S3 = 12.52
11 16
12 14
SAD 45.32
MAD 4.12
Go ahead.. Knock yourself out!
Simple Exponential Smoothing with  = 0.7
Period Sales Forecast Dev.
t At St Ft (error) We set St=1 = At=1
1 10 10** ~ ~ to start since
2 12 11.40 10.00 2.00 there is no prior
3 13 12.52 11.40 1.60 data available.
4 16 14.96 12.52 3.48
5 19 17.79 14.96 4.04
6 23 21.44 17.79 5.21

Ft = St-1
7 26 24.63 21.44 4.56
8 30 28.39 24.63 5.37
9 28 28.12 28.39 0.39
10 18 21.04 28.12 10.12
11 16 17.51 21.04 5.04
12 14 15.05 17.51 3.51
SAD 45.32
MAD 4.12
How do you find the value for ?
◦ New Forecast
= Last period’s forecast
+  (Last Period’s Actual demand – Last Period’s forecast)

Ft = Ft-1 +  (At-1 – Ft-1),  is a parameter.

◦ (At-1 – Ft-1) is the deviation or error from the previous period

◦ We are adding back a proportion of the forecast error to “even out” or smooth the forecast
for the next period
◦ The proportion added back is , where 0 <  < 1
◦  is the exponential smoothing parameter
Problems with SES
SES, like any other moving average technique fails to recognize
trends in the data

The SES formula must be modified to incorporate an adjustment for

trends in the data

Trend is the average change in the data over a number of periods

Like a SES, we can use exponential smoothing to weight recent

changes more heavily than older ones
Workshop Part I
 t At Ft Application of
Mon 89 81 Knowledge
Tue 91 89 Trisha is looking at her average RevPAR for last
week and she has asked you to let her know
Wed 94 96 how well her forecasts predict her actual sales.
She has provided you the following information:
Thur 101 99 SAD
Fri 134 124 MAD
MAPE
Sat 129 130
MFE
Sun 94 98 MSE
Calculate Moving Averages for:
Mondays Rooms Sold 5-day
1 82
3-day
2 82
3 82 3-day weighted (20%, 30%, 50%)
4 82
5 82
6 82
7 84
8 86
9 87
10 86
 Using an  of
Period Sales Abs.
t At Forecast Dev.
St (error)

0.6, calculate the

Ft
1 100

SAD and MAD for

2 112
3 113

the following
4 116
5 119
6
7
123
126 sales.
8 130
9 128
10 118
SAD
MAD
Questions
◦ Project Management
◦ Gantt Charting
◦ Critical Path Management (CPM)
Service Operations
Analysis
HTM*3120
SESSION 8B
Session Topics
◦ Queuing Systems
◦ Queuing Analysis
 Essential Features of Queuing Systems

Renege

Arrival Queue
process Departure
Calling Queue discipline Service
population configuration process
Balk No future
need for
service

12-3
 Essential Features of Queuing Systems

Renege

Arrival Queue
process Departure
Calling Queue discipline Service
population configuration process
Balk No future
need for
service

12-4
 Arrival Process
Arrival
process

Static Dynamic

Random Random arrival Customer-

Facility- exercised
arrivals with rate varying
controlled control
constant rate with time

Accept/Reject Price Appointments Reneging Balking

12-5
Variation in Arrival Rates
Ambulance Calls Per Hour Patient Arrivals at Health Clinic

12-6
 Poisson and Exponential Equivalence
Poisson distribution for number of arrivals per hour (top view)

One-hour
1 2 0 1 interval
Arrival Arrivals Arrivals Arrival

62 min.
40 min.
123 min.

Exponential distribution of time between arrivals in minutes (bottom view)

12-7
 Essential Features of Queuing Systems

Renege

Arrival Queue
process Departure
Calling Queue discipline Service
population configuration process
Balk No future
need for
service

12-8
Queue Configurations

12-9
Queue Configurations
Single-queue
◦ Advantages:
◦ The arrangement guarantees fairness by
ensuring that a first-come, first-served rule
(FCFS) applies to all.
◦ There is a single queue; thus, no anxiety is
associated with waiting to see if one
selected the fastest line.
◦ With only one entrance at the rear of the
queue, the problem of cutting-in is resolved
and reneging made difficult.
◦ Privacy is enhanced because the
transaction is conducted with no one
standing immediately behind the person
being served.
◦ This arrangement is more efficient in terms
of reducing the average time that
customers spend waiting in line.

12-10
Queue Configurations

Multiple-queue
◦ Advantages:
◦ Service provided can be differentiated.
◦ Ex. Supermarkets and lane 1
◦ Division of labour is possible.
◦ Season pass vs first time
◦ The customer has the option of selecting a
particular server of preference.
◦ Balking behavior may be deterred.

12-11
Queue Configurations

Take a Number
◦ Advantages:
◦ No need for a formal line.
◦ Customers are free to wander, possibly
increasing sales.

12-12
 Essential Features of Queuing Systems

Renege

Arrival Queue
process Departure
Calling Queue discipline Service
population configuration process
Balk No future
need for
service

12-13
Queue Discipline
A queue discipline is a policy established by management pertaining to
how to select the next customer from the queue.

© 2011 JOHN WILEY AND SONS, INC.. ALL RIGHTS RESERVED

 Queue Discipline
Queue
discipline

Static
Dynamic
(FCFS rule)

selection Selection based

based on status on individual
of queue customer
attributes

Number of Processing time

customers Round robin Priority Preemptive of customers
waiting (SPT or cµ rule)

12-15
 Essential Features of Queuing Systems

Renege

Arrival Queue
process Departure
Calling Queue discipline Service
population configuration process
Balk No future
need for
service

12-16
Service Facility Arrangements
Service facility Server arrangement
Parking lot Self-serve

Cafeteria Servers in series

Toll booths Servers in parallel

Supermarket Self-serve, first stage; parallel servers, second stage

Hospital Many service centers in parallel and series, not all used by each
patient

12-17
Queuing Analysis
Queuing Formulas
1. Mean arrival rate:
2. Mean service rate:
3. Mean number in service:
4. Probability of exactly “n” customers in the system:
5. Probability of “k” or more customers in the system:
6. Mean number of customers in the system: Ls = Lq + p (M/M/c)
Ls = / ( - ) (M/M/1)
7. Mean number of customers in queue:
(M/M/1)

Lq = see values table using C and p (M/M/c)

8. Mean time in system: (M/M/1)

Ws = Ls (M/M/c)

9. Mean time in queue: (M/M/1) Wq = Lq (M/M/c)

 Standard M/M/1 Model
The standard M/M/1 model has five assumptions:
1. Calling population: an infinite or very large population of callers arriving. The
callers are independent of each other and not influenced by the queueing system
(e.g., an appointment is not required)
2. Arrival process: Negative exponential distribution of interarrival times or Poisson
distribution of arrival rate.
3. Queue configuration: Single waiting line with no restrictions on length and no
balking or reneging.
4. Queue discipline: FCFS
5. Service process: One server with negative exponential distribution of service
times.
Standard M/M/1 Model
Lake Travis has one launching ramp near the dam for people who trailer
there small boats to the recreational site. A study of cars arriving with
boats in tow indicates a Poisson distribution with a mean rate of = 6
boats per hour during the morning launch. A test of the data collected
on launch times suggests that an exponential distribution with a mean
of 6 minutes per boat (equivalent service rate = 10 boats launched
per hour) is a good fit.
What is the average number of customers in service (the launch)?

What is the probability of having at least 1 customer in the system upon

arrival?

What is the probability of the ramp being available to pull right in?

What is the average number of boats in the system?

Standard M/M/1 Model
Lake Travis has one launching ramp near the dam for people who trailer
there small boats to the recreational site. A study of cars arriving with
boats in tow indicates a Poisson distribution with a mean rate of = 6
boats per hour during the morning launch. A test of the data collected
on launch times suggests that an exponential distribution with a mean
of 6 minutes per boat (equivalent service rate = 10 boats launched
per hour) is a good fit.
What is the average number of customers in service (the launch)?
= 6/10 = 0.6

What is the probability of having at least 1 customer in the system upon

arrival?

What is the probability of the ramp being available to pull right in?

What is the average number of boats in the system?

Standard M/M/1 Model
Lake Travis has one launching ramp near the dam for people who trailer
there small boats to the recreational site. A study of cars arriving with
boats in tow indicates a Poisson distribution with a mean rate of = 6
boats per hour during the morning launch. A test of the data collected
on launch times suggests that an exponential distribution with a mean
of 6 minutes per boat (equivalent service rate = 10 boats launched
per hour) is a good fit.
What is the average number of customers in service (the launch)?

= 6/10 = 0.6
What is the probability of having at least 1 customer in the system upon
arrival?
◦ = ( / )1 = (6/10)1 = 0.6

What is the probability of the ramp being available to pull right in?

What is the average number of boats in the system?

Standard M/M/1 Model
Lake Travis has one launching ramp near the dam for people who trailer
there small boats to the recreational site. A study of cars arriving with
boats in tow indicates a Poisson distribution with a mean rate of = 6
boats per hour during the morning launch. A test of the data collected
on launch times suggests that an exponential distribution with a mean
of 6 minutes per boat (equivalent service rate = 10 boats launched
per hour) is a good fit.
What is the average number of customers in service (the launch)?

= 6/10 = 0.6
What is the probability of having at least 1 customer in the system upon
arrival?
◦ = ( / )1 = (6/10)1 = 0.6

What is the probability of the ramp being available to pull right in?
P0 = pn (1-p) = 0.60(1 – 0.6) = 1(0.4) = 0.4

What is the average number of boats in the system?

Standard M/M/1 Model
Lake Travis has one launching ramp near the dam for people who trailer
there small boats to the recreational site. A study of cars arriving with
boats in tow indicates a Poisson distribution with a mean rate of = 6
boats per hour during the morning launch. A test of the data collected
on launch times suggests that an exponential distribution with a mean
of 6 minutes per boat (equivalent service rate = 10 boats launched
per hour) is a good fit.
What is the average number of customers in service (the launch)?
= 6/10 = 0.6
What is the probability of having at least 1 customer in the system upon
arrival?
◦ = ( / )1 = (6/10)1 = 0.6

What is the probability of the ramp being available to pull right in?
P0 = pn (1-p) = 0.60(1 – 0.6) = 1(0.4) = 0.4

What is the average number of boats in the system?

Ls = / ( - ) = 6 / (10-6) = 1.5 boats
 Standard M/M/c Model
The standard M/M/c model has five assumptions:
1. Calling population: an infinite or very large population of callers arriving. The
callers are independent of each other and not influenced by the queueing system
(e.g., an appointment is not required)
2. Arrival process: Negative exponential distribution of interarrival times or Poisson
distribution of arrival rate.
3. Queue configuration: Single waiting line with no restrictions on length and no
balking or reneging.
4. Queue discipline: FCFS
5. Service process: servers across all channels are independent and equal (all servers
are considered equal).
 Standard M/M/c Model
The Dean of CBE has assigned a secretary to each of her four departments.
Each secretary assists the faculty with class material and correspondence.
The Dean received complaints from the Economics faculty, about delays in
getting work accomplished. After collecting information on the requests of
each secretary, it was found that secretarial work arrives with a Poisson
distribution at an average rate of = 2 requests per hour for all
departments except economics. Economics has an average rate of = 3.
The average time to complete a piece of work is 15 minutes ( = 4 per
hour). The current system is essentially four M/M/1 operations.
What is the average turnaround for the Economics secretaries as compared
to the other three?
 Standard M/M/c Model
The Dean of CBE has assigned a secretary to each of her four departments.
Each secretary assists the faculty with class material and correspondence.
The Dean received complaints from the Economics faculty, about delays in
getting work accomplished. After collecting information on the requests of
each secretary, it was found that secretarial work arrives with a Poisson
distribution at an average rate of = 2 requests per hour for all
departments except economics. Economics has an average rate of = 3.
The average time to complete a piece of work is 15 minutes ( = 4 per
hour). The current system is essentially four M/M/1 operations.
What is the average turnaround for the Economics secretaries as compared
to the other three?
Economics = = 1 / (4-3) = 1 hour

Other Three = = 1 / (4-2) = 0.5 hour

 Standard M/M/c Model
The Dean of CBE has assigned a secretary to each of its four departments. Each
secretary assists the faculty with class material and correspondence. The Dean
received complaints from the Economics faculty, about delays in getting work
accomplished. After collecting information on the requests of each secretary, it
was found that secretarial work arrives with a Poisson distribution at an
average rate of = 2 requests per hour for all departments except economics.
Economics has an average rate of = 3. The average time to complete a piece
of work is 15 minutes ( = 4 per hour). The current system is essentially four
M/M/1 operations.
What would be the average number of customers in the system if the Dean was
to pool all four secretaries?
 Standard M/M/c Model
The Dean of CBE has assigned a secretary to each of its four departments. Each
secretary assists the faculty with class material and correspondence. The Dean
received complaints from the Economics faculty, about delays in getting work
accomplished. After collecting information on the requests of each secretary, it
was found that secretarial work arrives with a Poisson distribution at an
average rate of = 2 requests per hour for all departments except economics.
Economics has an average rate of = 3. The average time to complete a piece
of work is 15 minutes ( = 4 per hour). The current system is essentially four
M/M/1 operations.
What would be the average number of customers in the system if the Dean was
to pool all four secretaries?
Ls = Lq + p
0.56, 0.004 = 0.564
=2+2+2+3= 9 = 9 / 4 = 2.25 c=4 Lq = ?

Ls = Lq + p
= ? + 2.25
Lq for
Various
Values of
C and
 Standard M/M/c Model
The Dean of CBE has assigned a secretary to each of its four departments. Each
secretary assists the faculty with class material and correspondence. The Dean
received complaints from the Economics faculty, about delays in getting work
accomplished. After collecting information on the requests of each secretary, it
was found that secretarial work arrives with a Poisson distribution at an
average rate of = 2 requests per hour for all departments except economics.
Economics has an average rate of = 3. The average time to complete a piece
of work is 15 minutes ( = 4 per hour). The current system is essentially four
M/M/1 operations.
What would be the average number of customers in the system if the Dean was
to pool all four secretaries?
Ls = Lq + p
= 0.31 + 2.25
= 2.56 per hour

What would be the average time in the system?

 Standard M/M/c Model
The Dean of CBE has assigned a secretary to each of its four departments. Each
secretary assists the faculty with class material and correspondence. The Dean
received complaints from the Economics faculty, about delays in getting work
accomplished. After collecting information on the requests of each secretary, it
was found that secretarial work arrives with a Poisson distribution at an
average rate of = 2 requests per hour for all departments except economics.
Economics has an average rate of = 3. The average time to complete a piece
of work is 15 minutes ( = 4 per hour). The current system is essentially four
M/M/1 operations.
What would be the average number of customers in the system if the Dean was
to pool all four secretaries?
Ls = 2.56 per hour

What would be the average time in the system?

Ws = Ls = 2.56 / 9 = 0.28 hour = 0.28 x 60 = 17 m
Service Operations
Analysis
HTM*3120
SESSION 10A
Important Notes
Final Test opens March 19, 2025, and closes
April 1, 2025, and consists of:

Padlet Sign Up – Final Test Review

 Project
Management
 The Nature of Project
Management
Projects involve a series of inter-related activities, often performed by individuals at different times.
Project Management Process planning
◦ work breakdown structure
◦ Scheduling
◦ controlling.

Principles of Effective Project Management

◦ direct people individually and as a team
◦ keep everyone informed
◦ manage healthy conflict
◦ Empower the team
Project Metrics: Cost, Time, Performance

15-4
Project Management
Questions
What activities are required to complete a project
and in what sequence?
When should each activity be scheduled to begin and
end?
Which activities are critical to completing the project
on time?
What is the probability of meeting the project
completion due date?
How should resources be allocated
to activities?

16-5
Sources of Unexpected Problems

16-6
Three Types of Project
Management Techniques
Gantt Charting
Critical Path Management (CPM)
Project Evaluation and Review Technique (PERT)
Gantt Charts
Developed by Henry Gantt in 1916
Determines the timing of individual tasks or activities within a project
Plots a time line for each activity across a time line for the project
A graphic representation of the work breakdown structure..
 Gantt Chart for Tennis
Tournament
ID Activity Days Day of Project Schedule
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
A Negotiate for 2
Location
B Contact Seeded 8
Players
C Plan Promotion 3

D Locate Officials 2

E Send RSVP 10
Invitations
F Sign Player 4
Contracts
G Purchase Balls 4
and Trophies
H Negotiate 1
Catering
I Prepare Location 3

J Tournament 2

Personnel Required 2 2 2 2 2 3 3 3 3 3 3 2 1 1 1 2 1 1 1 1
Problems with Gantt Charts
Do not show interdependence of activities

Do not show relative importance of activities regarding completion

times

Unwieldy and awkward for complex projects because the value lies in
seeing the whole project at once, and paper size limits that ability

Network methods (CPM and PERT) provide a better solution…

Critical Path and PERT
Project management techniques in broad usage for many years

“Microsoft Project” is software based on these methods

Gives completion time for complex projects

Identifies activities that constrain the timely completion of the project

Identifies the earliest and latest start and finish times for each activity

Gives the amount of slack for each activity

Problems with CPM….
Task completion times are seldom known with
complete accuracy..

Consequently, projects can get bogged down if critical

path activities get bogged down

Non-critical activities may become critical if delays are

substantial

We need a way to incorporate uncertainty into the

Project Management technique…
Project Evaluation and Review
Technique (PERT)
An extension of CPM
◦ Facilitates predictions of completion when task times are variable and
uncertain

Uses estimates of “Best”, “Most Likely” and “Worst” completion times

to construct a beta probability distribution

The Beta distribution behaves similar to the Normal distribution

A Simple
Project
Management
Example…
ORGANISING A TENNIS
TOURNAMENT!
Tasks Involved….
Sign Location contract
Contact players
Plan promotion
Locate Officials
Send Invites
Buy trophies
Player Contracts
Catering Contract
Set-up Location
Run tournament
CPM
First Step
Using the Work Breakdown Structure, draw a CPM Network Diagram.

2 Kinds of CPM Network Diagrams:

1. Arrows are activities, nodes show progression (AOA)

2. Nodes are activities, arrows show progression. (AON)

We will do the AON style

Work Breakdown Structure…
Activity Code Predecessor Time
Location contract A - 2 days
Contact players B - 8 days
Plan promotion C A 3 days
Locate Officials D C 2 days
Send Invites E C 10 days
Buy trophies F B&C 4 days
Player Contracts G D 4 days
Catering Contract H E&F 1 days
Set-up Location I E&G 3 days
Run tournament J H&I 2 days
Step 1: Draw the CPM Network

A C D G
2 3 2 4

Start E I J
10 3 2

B F H
8 4 1
Some CPM notation…
TS
ES = earliest start time
EF = earliest finish time
LS = latest start time
LF = latest finish time
ES EF
TS = total slack
= LF – EF
= LS – ES
LS LF
 CPM Scheduling Formulas
ES = EFpredecessor (max) (1)
EF = ES + t (completion time) (2)
LF = LSsuccessor (minimum) (3)
LS = LF - t (completion time) (4)
TS = LF - EF (5)
or TS = LS - ES (6)
The “Forward Pass”
Begin by setting the Earliest Start time for the Start = 0. Since the Start has
no completion time, the Earliest Finish for the start also = 0.

The earliest start time (ES) = 0 for the activities that begin the project, A
(Location Contract) and B (contact Players).

A (Location Contract) ES = 0
B (Contact Players) ES = 0

ES + completion time = Earliest Finish Time

The Earliest Finish time becomes the Earliest Start time for the next activity.

Continue….
 Step 2: The forward pass…
ES 0 2 EF = ES + completion time
= 0 + 2

A C D G
2 3 2 4
EF
0 0

Start E I J
10 3 2

TS B F H
8 4 1
ES EF
LS LF
 Step 2: The forward pass…
2+3=5 5+2=7 7+ 4 = 11

0 2 2 5 5 7 7 11

A C D G
2 3 2 4

0 0 5 15 5 + 10 = 15

Start E I J
10 3 2

TS
ES EF
B F H
8 4 1
LS LF
 Step 2: The forward pass…
0 2 2 5 5 7 7 11
What’s
going
A C D G on
2 3 2 4 here?

0 0 5 15 15 18

Start E I J
10 3 2

TS B F H
8 4 1
ES EF
LS LF
What’s Both E and G precede
going activity I.
on
here? Activity I cannot begin
until the activity with
5 7 7 11
the latest Early Finish
D G time is complete.
2 4

5 15 15 18 EFE = 15, EFG = 11,

We use the EFmax
E I J time (from activity E)
as the Earliest Start
10 3 2

TS
18 20 time for activity I.
ES EF H
LS LF 1
Forward Pass Rule:
Whenever an activity (z) has two (2) or more predecessor
activities (x, y), the Earliest Start time of z is the largest earliest
finish time of the predecessor (x, y) activities

Formula: ES = EFpredecessor x or y (max)

Predecessors z

y
 Step 2: The forward pass…
0 2 2 5 5 7 7 11

A C D G
2 3 2 4

0 0 5 15 15 18

Start E I J
10 3 2

0 8
8 12 18 20

B F H
8 4 1
TS
15 16
ES EF Don’t forget the activities in the lower
LS LF part of the network!!
The “Backward Pass”
After the forward pass is completed, we begin the Backward Pass.

1. The earliest finish time (EF) of the final step becomes the latest finish
time (LF) of the previous step.

2. Update LS = LF – completion time

3. Now, we can calculate TS = LF – EF or TS = LS – ES

LF of the predecessor activity = LS

 Step 3: The backward pass…
0 2 2 5 5 7 7 11
1. EF of J
A C D G becomes
2 3 2 4 LF of J

0 0 5 15 15 18

Start E I J
10 3 2

0 8 3. TS = LF - EF 0
8 12 18 20
18 20
B F H
8 4 1
TS
ES EF 2. LS = LF – completion time 15 16
LS LF = 20 – 2 = 18
 Step 3: The backward pass…
0 2 2 5 5 7 7 11
TS = LF - EF
A C D G = 18 -18 = 0
2 3 2 4
0
0 0 5 15 15 18
15 18

Start E I J
10 3 2

0 8 0
8 12 18 20
18 20
B F H
8 4 1
TS 2
ES EF TS = LF - EF 15 16
= 18 -16 = 2 17 18
LS LF
 Step 3: The backward pass… 4
0 2 2 5 5 7 7 11
11 15

A C D G
2 3 2 4
0 0
0 0 5 15 15 18
What’s 5 15 15 18
going
Start on E I J
here? 10 3 2

0 8 0
8 12 18 20
18 20
F H
4 1
TS 2
ES EF 15 16
LS LF 17 18
 Both I and H follow
0 0 activity E.
5 15 15 18 The LF of Activity E
5 15 15 18
cannot be earlier than
E I the earliest LS of I and
10 3
H.
LSI = 15, LSH = 17, We
H use the earliest start
1
2 time from activity I as
15 16 the Latest Finish time
TS
17 18
for activity E.
ES EF
LS LF
 Backward Pass Rule
Whenever an activity (z) has two (2) or more following
(successor) activities (x, y), the Latest Finish time of z is the
smallest Latest Start time of the successor (x, y) activities
LF = LSsuccessor x or y (minimum)

Z x
Successor
Activities
TS
ES EF y
LS LF
 Step 3: The backward pass…
0 0 4 4
0 2 2 5 5 7 7 11
0 2 2 5 9 11 11 15

A C D G
2 3 2 4
0 0 0
0 0 5 15 15 18
0 0 5 15 15 18

Start E I J
10 3 2
5
5 0
0 8
8 12 18 20
5 13
13 17 18 20
B F H
TS 8 4 1
2
ES EF
15 16
LS LF 17 18
 Step 3: The backward pass…
0 0 4 4
0 2 2 5 5 7 7 11
0 2 2 5 9 11 11 15

A C D G
2 3 2 4
0 0 0
0 0 5 15 15 18
0 0 5 15 15 18

Start E I J
10 3 2
5
5 0
0 8
8 12 18 20
5 13
13 17 18 20
B F H
TS 8 4 1
2
ES EF
15 16
LS LF 17 18
Identify the Critical Path…
The critical path is the sequence of activities with TS = 0.

TS
ES EF
LS LF

If the path is incomplete from start to finish.. you have made a mistake!
 Step 4: The Critical Path
0 0 4 4
0 2 2 5 5 7 7 11
0 2 2 5 9 11 11 15

A C D G
2 3 2 4
0 0 0
0 0 5 15 15 18
0 0 5 15 15 18

Start E I J
10 3 2
5
5 0
0 8
8 12 18 20
5 13
13 17 18 20
TS B F H
ES EF 8 4 1
2
LS LF 15 16
17 18
Critical Path: A→C→E→I→J
 Tennis Tournament
Activity Code Predecessor Time
Location contract A - 2 days 2
Contact players B - 8 days
Plan promotion C A 3 days 3
Locate Officials D C 2 days
Send Invites E C 10 days 10
Buy trophies F B&C 4 days
Player Contracts G D 4 days
Catering Contract H E&F 1 days
Set-up Location I E&G 3 days 3
Run tournament J H&I 2 days 2
Minimum completion time 20
The Tennis Tournament…
conclusions
1. It takes 20 days to organise and run.
2. We need to supervise A, C, E, I & J activities very carefully (no slack).
If any of these steps are delayed, the tournament will be delayed.
3. All other activities have varying amounts of slack… and are less
critical.
4. Earliest and latest start and finish times help managers keep the
project on track.
Workshop
Application of Knowledge
A local event company is planning their end of the year event. They have
provided you a work breakdown and asked you for the following
information:
Activity Time (days) Predecessor
1. a CPM network A 4 ~
B 3 ~
2. calculation of the earliest
C 4 ~
Start, Finish and Latest
Start, Finish and slack D 6 A
time for each activity. E 3 A

3. Identification of the F 5 C
Critical Path and G 4 D
statement of the H 6 A, B, F
completion time.
I 2 E, G, H
 Step 1: The Model

A D G
4 6 4

I
B E 2
Start 3 3

C F H
TS 4 5 6

ES EF
LS LF
Yield
Management
What is Yield Management?
Optimization of profit by jointly maximising capacity
utilization and average price

Given a downward sloping demand curve (more people

will buy if the price is reduced) a trade-off develops
between the desire to obtain the highest average price
and sell all available units of capacity

The manager s problem is to know which combination

of prices and volume for the same product will
optimize revenue.
 Fixed capacity (short – medium term)
◦ Difficult to adjust to fit current
demand
Perishable Products
Low Variable costs
What ◦ Ability to discount
businesses can High fixed costs
benefit ◦ Urgency to utilize existing capacity
most from Variable (but predictable) demand
Yield ◦ Advance selling (reservations)
Management? ◦ Cyclical demand
◦ Duration and Intensity
◦ Expenditure level
Plus
Ability to Manage Differential Pricing
Ability to Communicate Efforts
Yield Management
Yield Management Strategy:
Charge different prices for the same product or
service while simultaneously selling as much
capacity as possible.

The Yield Manager s problem:

1. Sell all units of capacity and
2. No customer who is willing to pay a higher price is
denied due to the prior sale of a room at a lower
price.
Yield Statistics
Used to identify whether price and capacity utilization strategies are
working (or not)

Consist of two parts:

Rate efficiency X Volume efficiency

Rate efficiency is often in the form of average rate, average check, average
price, etc.

Volume efficiency is often in the form of capacity utilization, occupancy,

etc.
Yield Management Tools

◦Pricing Tools

◦Capacity Tools
Pricing Tools

1. General Pricing

2. Demand-based Pricing

3. Nested Pricing
Pricing Tools - General Pricing
A single-price structure
Everyone pays the same price, regardless of time of
day, time of booking, etc.
Pricing Tools –
General Price-Setting Goals
Set the price such that all capacity is sold and revenue is
maximized.
◦ Unsold capacity is wasted
◦ Price is too low if demand exceeds capacity
◦ Price is too high if capacity is unsold

Revenue is optimized when the price is such that every

seat is sold and no additional demand is unfulfilled at
that price
◦ This is very hard to do..
◦ Variable demand (by time and market segment)
◦ Elasticity of demand is not constant
◦ How do you know if the price is perfectly matched to demand?
Pricing Tools – General Price
Small General Price Increase
◦ Generally accepted to keep up with inflation if modest

Large General Price Increase

◦ May result in volume losses, particularly among guests who are price
sensitive

◦ Invites direct comparisons with competitors

◦ The easiest way to make money is to get your customer to give it to you!
Pricing Tools –
Demand-based Pricing:
◦ Higher prices charged during busy periods
◦ Discounts during low volume periods
◦ Price-sensitive customers shift from busy periods to slow
periods
◦ Time-sensitive customers pay more to go when they
want to go
◦ Low industry margins (high variable costs) reduce the
ability to discount.. (restaurants vs. airlines)
Pricing Tools - Nested Pricing
◦ Charge different prices for the same product sold at the same time
(differential pricing)
◦ Airline seats: people sitting next to each other pay different prices!
◦ Increase the average price by selling as many seats as possible to
time-sensitive buyers at higher prices
◦ Increase capacity utilization by attracting price-sensitive buyers with
discounts to fill capacity
◦ Manage the average price by limiting the number of seats sold at
lower prices
Nested Pricing
Perception of unfairness among consumers:

◦ Early booking gets the lowest price is generally accepted as fair (a variant of
the first-come first-served protocol)

◦ Price according to ability to pay is considered unfair by many

Fairness Perceptions can be managed:

◦ Fences (restrictions on use such as transferability, changes, cancellation,
etc.)
◦ Full fares receive available upgrades, deep discount fares do not, etc.
◦ Coupons, loyalty programs
Managing Perceptions of
Fairness – Fences
Restrictions on fares
◦ Cancellations / refund or credit
◦ Change fees
◦ Upgrade to business class
◦ Transferability to another person
◦ Reward points / no points
◦ Fare blocking (limits on the number of discounted seats for sale)
Managing Perceptions of
Fairness – Coupons
Some customers receive discount coupons, others do not.

A way for restaurants to achieve nested pricing…

Managing Perceptions of
Fairness – Unfairness?
YM serves travelers who need a flight urgently extremely well:

◦ If a single, general price is set so that the aircraft is always full (capacity is
optimized), there are no seats available for “last minute” travelers.

◦ A few, “full fare” seats are only available because of the high prices charged
and, there is a possibility that these seats will be vacant (loss to the airline)

◦ Stand-by top-ups (the lowest fare class) are primarily used as perks for
airline employees, or students who are willing to wait around airports until
such a seat is available
Unfairness?
It serves travelers on a budget extremely well:

◦ If a single, general price is set so that the aircraft is always full (capacity is
optimized), budget travelers may not travel, travel less often, or not take
their children with them

◦ “Super-saver” seats can be used to utilize capacity that would otherwise be

unsold at a higher price and yield no revenue for the airline
How can you estimate the number
of full-fare seats that are “kept back” ?
1. Expected Marginal Seat Revenue:

The price of a seat on a plane is $1,000.

If the probability of selling 1 full-fare seat in the last hour is
25%, the “break-even” Full-fare price for the airline is:

$1,000 = .25 X Full-fare

Full-fare = 1,000 ÷ 0.25 = $4,000

The airline should charge at least $4,000 for this seat.

2. Critical Fractile Technique (Example previously and later..)

The Case For Nested Pricing
An aircraft has 100 standard seats and total flight costs are $100,000.
There is no business class seating.

Fares available:
Full-fare: $2,000 per seat
Economy: $1,200 per seat
Super-saver: $ 750 per seat

If all 100 seats were sold @ $2,000, Revenue = $200,0000

The flight would yield a contribution of $ $100,000.

If all 100 seats were sold @ $1,200, Revenue = $120,0000

The flight would yield a contribution of $ $20,000.

If all 100 seats were sold @ $750, Revenue = $75,0000

The Case For Nested Pricing
Assume yield managers have a contribution target
of $20,000 per flight:
◦ This requires an average price of $1,200 per seat, and
100% occupancy (100 seats X $1200 = $120,000 less
$100,000 expenses)
◦ What happens if only 80 seats are sold at this price? (80 X
$1,200 = $96,000 ~ $4,000 loss
◦ Suppose you can sell the remaining 20 seats @ $750 … 20 X $750 =
$15,000 less $4,000
= $11,000 contribution
The Case For Nested Pricing
1. If all seats are sold at the Super-saver rate, the
plane is full, but the airline goes out of
business.

2. If all seats were sold at the economy rate, they

meet the target. If they are 20 seats short of a
full plane, they lose money.

3. A mix of seats, using super-saver rates to “top

up” capacity, full-fare seats to increase the
average rate and economy seats illustrates how
nested pricing can be used to optimize yield…..
The Case For Nested Pricing
Nested Pricing Scenario:
80 seats @1,200 = $96,000 (Economy)
10 seats @ 750 = 7,500 (Super-saver)
10 seats @ 2,000 = 20,000 (Full fare)
Total Revenue = $123,500
Contribution = $ 23,500

 A mix of seats, using super-saver rates to “top up”

capacity and full-fare seats to increase the average rate
illustrates how nested pricing can be used to optimize
yield..
Nested Pricing
The manager s problem is to know how many seats not to sell at each
price so that no passenger willing to pay more for that seat is turned
away

Nested pricing requires capacity allocations for each level of price

This is accomplished using Capacity Tools

Capacity tools
Critical Fractile Technique
Blackjack Airlines
◦ Full-coach, 1-way fare = $69
◦ Discounted, 1-way fare = $49
◦ Capacity = 95 seats
◦ Demand for full coach tickets has an average of 60 pax per flight and a
standard deviation of 15 pax.

What is the minimum number of seats that should

be reserved for the full-coach rate?
Blackjack Airlines
Cu = lost revenue due to underestimating
demand for a full fare seat
= $69 - $49
= $20

Co = lost revenue due to overestimating

demand for a full fare seat
= $49

Critical Fractile = $20 / ($20 + $49)

= 0.2899
Blackjack Airlines
Critical Fractile = $20 / ($20 + $49)
Prob. (d<x) = 0.2899 ~ .29
= area under the curve (< 50%)

Look-up Z-.5 +.29 = Z -.21

= -.55
Prob. d<x
Min. # of full-fare seats
= 0.29
= μ + (zσ)
= 60 + (-.55 X 15)
-z 0 +z
= 51 seats

Do not sell more than 95-51 = 44

Discounted seats
Blackjack Airlines
The Blackjack example uses only 2 rates

It can be done for as many rates as you have demand data for…

Virtual nesting..
◦ Computationally intensive

◦ Each seat has its own price

◦ Entire inventory remaining is recalculated each time a seat is sold..

◦ Algorithm was invented in 1987 by Peter Belobaba, a Canadian PhD student at MIT.
Low -Tech Price and Capacity Tool
Before Belobaba s algorithm, a low-tech methodology was in common
use to solve the price and capacity decisions for capacity-constrained
industries

The so-called “Booking Curve Method” was easy to understand, simple

to use and delivered improved results.
Step 1: Plot the Booking
Curve
What is a Booking Curve?

No-shows
No of Bookings

Historical average # of bookings

200 0
Days before customer arrives for service
Step 1: Plot the Booking
Curve
What is a Booking Curve?

No-shows
No of Bookings

Historical average # of bookings

200 0
Days before customer arrives for service
Step 2: Put Upper and Lower
bounds on the booking curve
Historical average bookings

- 10%

+ 20%
No of Bookings

- 20%

+ 10%

200 0
Days before customer arrives for service
Step 2: Put Upper and Lower
bounds
on the booking curve
Historical average bookings

- 10%

+ 20%
No of Bookings

- 20%

+ 10%

200 0
Days before customer arrives for service
Step 3: Plot actual bookings
against historical booking data
In this case, we are 120 days before customers arrive, and bookings are
10% higher than last year

Actual # of bookings 120

days before arrival - 10%

+ 20%
No of Bookings

- 20%

+ 10%

200 120 0
Days before customer arrives for service
 Step 3: Compare actual bookings
against historical booking data
Boundaries signal rate changes:
◦ When actual bookings are +10%, we increase rates 1 step.

Actual # of bookings 120

days before arrival - 10%

+ 20%
No of Bookings

- 20%

+ 10%

200 120 0
Days before customer arrives for service
 Step 3: Compare actual bookings
against historical booking data
Boundaries signal rate changes:
◦ When actual bookings are the same, rate stays the same

Actual # of bookings 120

days before arrival - 10%

+ 20%
No of Bookings

- 20%

+ 10%

200 120 0
Days before customer arrives for service
 Step 3: Compare actual bookings
against historical booking data
Boundaries signal rate changes:
◦ When actual bookings are +20%, we increase rates 2 steps.

Actual # of bookings 120

days before arrival - 10%

+ 20%
No of Bookings

- 20%

+ 10%

200 120 0
 Step 3: Compare actual bookings
against historical booking data
Boundaries signal rate changes:
◦ When actual bookings are -10%, we decrease rates 1 step.

Actual # of bookings 120

days before arrival - 10%

+ 20%
No of Bookings

- 20%

+ 10%

200 120 0
Days before customer arrives for service
Booking Curve Method
Booking curves are plotted using Excel, based on historical data for the
average number of bookings on hand for each day prior to arrival.

Reservations staff keep track of the actual number of bookings for a

particular date and use the chart to quote rates.

Rates are set to correspond to boundaries: i.e. +10% might signal a

price 10% higher than the base, -20% might signal a rate 20% below
normal

Higher rates decrease the number of bookings, lower rates increase the
number of bookings
Yield Statistics
REVPAR (Revenue per Available Room)

REVPAR = Occupancy % X ADR

Which hotel is better off?

ADR Occ. % REVPAR

Hotel A $150 75% $112.50
Hotel B $250 50% $125.00
REVPAR Math…
If a hotel has 100 rooms available, and REVPAR = $225, what is the Total
Room Revenue?

Total Revenue
= REVPAR X # of Rooms available
= $225 X 100
= $22,500
REVPAR Math….
If a hotel has REVPAR = $225, and occupancy of 75%, what is the
Average Daily Rate?

ADR = REVPAR ÷ Occ. %

= $225 ÷ .75
= $300 per room
REVPAR is not perfect…
In cases where variable costs are small, REVPAR works very well

Variable costs have an impact on Profit, because rooms not sold are
rooms that do not contribute to variable costs

The following example, with variable costs of an occupied room = $14

per room night demonstrates this…
 Budget Hotel Example
100 room budget Hotel
Data: Situation 1 Situation 2 Situation 3
Total Revenue $6,000 $6,000 $6,300
Number of rooms Sold 100 75 100
Cost per occupied room 14 14 14
ADR
(Average Daily Rate)
Cap. Util.
(Occupancy %)
REVPAR
(ADR X Occ.%)
 Budget Hotel Example
100 room budget Hotel
Data: Situation 1 Situation 2 Situation 3
Total Revenue $6,000 6,000 $6,300
Number of rooms Sold 100 75 100
Cost per occupied room 14 14 14
ADR $60 $80 $63
(Average Daily Rate)
Cap. Util.
(Occupancy %)
REVPAR
(ADR X Occ.%)
 Budget Hotel Example
100 room budget Hotel
Data: Situation 1 Situation 2 Situation 3
Total Revenue $6,000 6,000 $6,300
Number of rooms Sold 100 75 100
Cost per occupied room 14 14 14
ADR $60 $80 $63
(Average Daily Rate)
Cap. Util. 100% 75% 100%
(Occupancy %)
REVPAR
(ADR X Occ.%)
 Budget Hotel Example
100 room budget Hotel
Data: Situation 1 Situation 2 Situation 3
Total Revenue $6,000 6,000 $6,300
Number of rooms Sold 100 75 100
Cost per occupied room 14 14 14
ADR $60 $80 $63
(Average Daily Rate)
Cap. Util. 100% 75% 100%
(Occupancy %)
REVPAR $60 $60 $63
(ADR X Occ.%)
 Budget Hotel Example
100 room budget Hotel
Data: Situation 1 Situation 2 Situation 3
Total Revenue $6,000 6,000 $6,300
Number of rooms Sold 100 75 100
Cost per occupied room 14 14 14
ADR $60 $80 $63
(Average Daily Rate)
Cap. Util. 100% 75% 100%
(Occupancy %)
REVPAR $60 $60 $63
(ADR X Occ.%)

REVPARmax = $63 indicates that Situation 3 is best.

 Budget Hotel Income Summary
Data: Situation 1 Situation 2 Situation 3
Total Revenue $6,000 6,000 $6,300
Number of rooms Sold 100 75 100

Expenses
Variable Costs ($14/room) 1,400 1,050 1,400
Gross Operating Income 4,600 4,950 4,900

Fixed Costs 4,000 4,000 4,000

Net Operating Income
REVPAR
Hotel Example
Data: Situation 1 Situation 2 Situation 3
Total Revenue $6,000 6,000 $6,300
Number of rooms Sold 100 75 100

Expenses
Variable Costs ($14/room) 1,400 1,050 1,400
Gross Operating Income 4,600 4,950 4,900

Fixed Costs 4,000 4,000 4,000

Net Operating Income 600 950 900
REVPAR $60 $60 $63
Hotel Example
Data: Situation 1 Situation 2 Situation 3
Total Revenue $6,000 6,000 $6,300
Number of rooms Sold 100 75 100

Expenses
Variable Costs ($14/room) 1,400 1,050 1,400
Gross Operating Income 4,600 4,950 4,900

Fixed Costs 4,000 4,000 4,000

Net Operating Income 600 950 900
REVPAR $60 $60 $63

Despite REVPARmax = $63, Situation 2 is best.

Which Scenario is more profitable?

1. ADR = $200, Occ. = 70%

2. ADR = $175, Occ. = 80%
Which Scenario is more profitable?

1. ADR = $200, Occ. = 70%

2. ADR = $175, Occ. = 80%

Both have the same REVPAR = $140

Restaurant Yield Statistics
Variable costs in Restaurants typically are 65% -
70% of Revenues

Yield Statistics based on Sales are misleading…

Yield Statistics based on Contribution Margin

(ConPAC) are more reliable.
YM Overview
The goal is to maximise the yield from an asset with
fixed capacity
Different market segments will pay different prices for
any product
Pricing and Capacity tools are used to simultaneously
increase average rate and optimise capacity utilization
Yield statistics (such as REVPAR or CONPAC) are a fast
and useful way to know if management actions
regarding price and capacity are improving yield, or
not.
Workshop
 The Mountain Motel
The 100 room Mountain
Motel is trying to determine
how many rooms they
should discount for next
Saturday. Their average
daily room rate is $90, and
their discount is typically
$15 less. Their typical
demand for their rooms at
ADR is 70% with a standard
deviation of 10%. How
many rooms should the
Mountain Motel reserve at
standard ADR for next
Saturday.?
Restaurant Sultra
Management estimates the maximum capacity of the
restaurant as 1,000 customers per day.

Variable costs (food and labour) are on average $7.00

per customer at the current price level

Fixed costs are $1,000 per day

Currently, average check is $10, but management is

trying a new price.. $10.99, and anticipates they will
lose some customers due to the higher price….
- Management estimates the maximum capacity of the restaurant as 1,000 customers per
day.
- Variable costs (food and labour) are on average $7.00 per customer at the current price
level
- Fixed costs are $1,000 per day
- Currently, average check is $10, but management is trying a new price.. $10.99, and
anticipates they will lose some customers due to the higher price….

Data: Situation 1 Situation 2 Situation 3

Number of covers sold 600 546 630
Cap. Util. 60% 54.6% 63%
Avg. Check $10.00 $10.99 10.00
Total Revenue
Expenses
Variable costs ($7 / cover)
Contribution
Fixed Costs
Net Income
REVPAC
CONPAC
Next Session – 11
◦ Test 2 Review
◦ Final Test
HFTM*3120
Service Operations
Management
WEEK 11
Weekly Topics
◦ Introduction to Revenue Management
◦ Core Concepts for Revenue Management
Introduction to
Revenue Management
The role of demand?
Price and Quantity interact:
As price is reduced, demand increases
Quantity

200
75

$150 $300 Price

The role of supply?
Price and Quantity interact:
As price is reduced, fewer rooms are supplied
Quantity

200
75

$150 $300 Price

The role of equilibrium?
Price and Quantity offered are perfectly matched

Quantity

150

$225 Price
The role of Revenue Management?
Revenue Managers operate within the area where prices are
neither too high or too low, and rooms supply matches demand
in that range

Rooms not
wanted
200 Zone of Maximum
75 Revenue Potential
Rooms not
offered

Price is unprofitable $150 $300 Price is too high

What Is Revenue?
Unit Price X Quantity of goods or services sold

Price:Premium Price
Regular Price Average Price
Discounted (ADR)

Quantity: Inventory
-Type of Item Utilization
-Quantity of item (Occupancy)
How do managers increase profits?

Revenue
- Variable costs
- Fixed Costs
= Net Income
 “Revenue management (RM) is the application of
information systems and pricing strategies to allocate the
What is right capacity to the right customer at the right place at
the right time” (Kimes and Wirtz, 2015, p. 2).
revenue
management The act of skillfully, carefully, and tactfully managing,
? controlling, and directing capacity and sources of income,
given the constraints of supply and demand.
•Tranter, Stuart-Hill and Parker
Revenue Management is a dynamic
process
◦ Situations change
◦ Markets change
◦ Opportunities change
◦ You may have been wrong and you might learn something new

Common misperception is that it is a computer program

Effective management through
the use of revenue management

1 2 3
Understand the Understand the many Become better at
importance of complex factors that making revenue
revenue management influence revenue management
management strategy decisions than their
and tactics competitors
Revenue
Management
Cycle
  Forecasting
 Differential Pricing
 Strategic Pricing
 Negotiating Skills
 Inventory and Price
Management

 Displacement Analysis
 Distribution Channels
 Inventory and Price Management
 Performance metrics
 Competitive fair market
share and market penetration
 Utilizing performance metrics
 Calculating internal and external
benchmarking
 Understanding the challenges of
revenue performance  Maximizing Revenue Optimization
measurement  Strategic Pricing & Displacement Analysis
 Restaurant Revenue Management
 Revenue Management outside of hotels and restaurants
The Best conditions for
Revenue Management
Fixed capacity (short – medium term)
◦ Difficult to adjust to fit current demand

Perishable Products

Low Variable costs

◦ Ability to discount

High fixed costs

◦ Urgency to utilize existing capacity

Variable (but predictable) demand

◦ Advance selling (reservations)
◦ Cyclical demand
◦ Duration and Intensity
◦ Expenditure level
Plus
Ability to Manage Differential Pricing
Ability to Communicate Efforts
Hospitality Industry Characteristics
Revenue Management Characteristics in the Hospitality Industry

Fitness
Issue Hotel Restaurant
Centre
SUPPLY CONSTRAINT Hard Soft Soft
ABILITY TO INCREASE
None Limited Limited
SUPPLY
ABILITY TO DECREASE
None Limited Limited
SUPPLY
Hospitality Industry Characteristics
Revenue Management Characteristics in the Hospitality Industry

Issue Hotel Restaurant Fitness Centre

ABILITY TO CARRY OVER Service Service capacity: Service capacity:
EXCESS SUPPLY TO THE capacity: No No No
NEXT DAY Product: No Product: Yes Product: No
PRACTICE PRICE
MODIFICATION IN
RESPONSE TO Almost always Almost never Sometimes
VARIATION IN SHORT-
TERM DEMAND
What makes ◦ Identifiable customer segments

revenue ◦ Different value propositions

◦ What attributes add value?

management ◦ How much value do they add?

work?
◦ How big are the segments?
What makes revenue
management work?
◦ Effectively Managing Segments
◦ Prevent those who would pay more from paying less
◦ The ability to distinguish
◦ The ability to separate
Remember the
◦ The perception of fairness Critical Fractile

◦ Long-term effects
◦ Customers learn the system
◦ Perceptions of fairness leading to dissatisfaction
The role of segmenting demand
Are the people willing to pay $300 the same as the
people who pay $150?
Quantity

200
75

$150 $300 Price

 Managing Segments: Paint
1. What are three major market segments?
2. How does each derive value from Paint?
3. How would you optimise revenue as a paint seller?

Segment 1 Segment 2 Segment 3

Value Proposition
Product /Price structure
Ability to separate segments
Other
 Segment 1 Segment 2 Segment 3
Value Proposition Value Brand Premium Builder
Managing
Segments: Paint
Product /Price $24 per gal $49 per gal $16 per gal
structure
Ability to separate - Price Sensitive - Time Only available to
segments Consumer Sensitive builders
(Lowest Cost) Consumer
1. What are three major market
- Larger segments?
Variety 2. How does each derive value
Consumer from Paint?
Other • Variety of • Can match • Sold only in 5
Colours any colour gal
3. How would you optimise
• Multiple • One coat • Only 10 basic revenue as a paint seller?
Applications application colours
Required available
• One coat
application
(if on primed
dry wall)
What can we control?
Hospitality Industry Areas of RM Responsibility
Segment
Full-service resorts Guest rooms
Meeting space
Food and beverage operations
Resort activities

Full-service hotels Guest rooms

Meeting space
Food and beverage operations

Limited-service hotels Guest rooms

What can we control?
Hospitality Industry Areas of RM Responsibility
Segment
Theme /amusement/ Admission tickets
water parks Food and beverage operations
Rides/activities
Spas Spa services
Salon services
Intermediaries/service Services designed to
organizations inform/advise RMs
(ex Smith Travel Research)
Consider some scenarios
Beach front hotel
◦ Occupancy is 76% - up 5% from last month
◦ ADR is $144 – up $4 from last month

◦ Is this a good sign?

◦ What are some levers they could consider for managing

revenue?
Another Scenario…
Beach front hotel
◦ Historically low profitability
◦ New GM was efficiency expert
◦ Cut costs
◦ Cut staff
◦ Profits declined with costs
◦ What are they missing?
◦ What should they think about doing differently?
 1 Focus on Price rather than costs when balancing supply
and demand
2 Replace cost-based pricing with market based pricing
Robert Cross’ 3 Sell to segmented micromarkets instead of mass
Core Concepts markets

for Revenue 4 Save your products for your most valuable customers

Management
5 Make decision based on knowledge, not supposition
6 Exploit each product’s value cycle
7 Continually evaluate your revenue opportunities
#1. Focus on price when
balancing supply and demand
Equilibrium points
◦ Exist for each market segment and product
◦ Identify where supply and demand produces optimal returns
#1. Focus on price when balancing
supply and demand
Conventional Wisdom
◦ In times of high demand, increase capacity
◦ In times of low demand, decrease capacity

Entropic Event:
◦ Demand fluctuations grow in frequency and intensity

RM Tactic:
◦ Address short term fluctuations with price, then capacity
#2. Replace cost-based pricing with
market based pricing
Customers do not care about an operations costs.
Customers care about:
◦ Availability of alternatives from competitors
◦ Amount of disposable income
◦ Urgency or need (real or perceived) for the product.
#2. Replace cost-based pricing with
market based pricing
Conventional Wisdom
◦ Set prices to cover costs and provide an acceptable profit margin.

Entropic Event:
◦ Nonconformist consumers determine the price they are willing to pay; market provides numerous
choices.

RM Tactic:
◦ Forget cost-based pricing. Set prices consumers will accept in a price-flexible environment; reduce
costs, if necessary, to meet margin requirements.
 #3. Sell to segmented
micromarkets instead of mass
markets
“Market Segmentation is the key to market-based pricing and revenue
maximization ” Robert G Cross
#3. Sell to segmented micromarkets
instead of mass markets

Think of McDonalds, who are their markets?

#3. Sell to segmented micromarkets
instead of mass markets
Conventional Wisdom
◦ Price should be set to sell the greatest number of units at the highest possible
price in the mass market.

Entropic Event:
◦ Consumer individualism shatters the mass market.

RM Tactic:
◦ Different segments demand different prices. To maximize revenue and stay
competitive, prices must vary to meet the price sensitivity of each market
segment.
#4. Save your products for your most
valuable customers

“First come first served”…?

“…you should find a way to predict which segments are willing

to pay the most and save those products for them.” Robert G
Cross
#4. Save your products for your most
valuable customers
Conventional Wisdom
◦ Sell products on a “first come, first-served” basis.

Entropic Event:
◦ Traditional business practices don t satisfy investor demands for aggressive
revenue growth.

RM Tactic:
◦ Understand demand at the micro-market level as accurately as possible, and
save products for the most valuable customers to achieve optimum revenue.
#5. Make decision based on knowledge,
not supposition
Humans Computers
◦ Intuitive ◦ Logical
◦ Creative ◦ Mindless
◦ Biased ◦ Indifferent
◦ Need to sleep ◦ Tireless

Robert Cross conducted “tests that pitted human “experts” against a

computer programmed to calculate optimal revenue from product value
cycles. The computer consistently won in every instance because it could
precisely locate the optimal point in each revenue curve. In fact, the
computer produced revenue increases of 5%-10%”.
#5. Make decision based on knowledge,
not supposition
Conventional Wisdom
◦ General assumptions can be made about future consumer behavior based
on intuition and personal observation.

Entropic Event:
◦ Nonconformist consumers are continually fragmenting the market and
changing buying behavior.

RM Tactic:
◦ Forecast demand at the micro-market level to gain knowledge of subtle
changes in consumer behavior patterns.
#6. Exploit each product s value cycle
“Generate maximum
revenue by
understanding the
value cycle and
optimally timing the
availability and price
of the product to each
micro-market
segment.”
Robert G. Cross
#6. Exploit each product s value cycle
To understand the value cycle, businesses need to:
1. Break their business down into logical stages (sub-markets)
2. Project a revenue trajectory for each stage
3. Determine the optimal time to launch the subsequent stage

How does this apply to:

◦ Restaurants
◦ Hotels
◦ Airlines
#6. Exploit each product s value cycle
Conventional Wisdom
◦ Decisions on product availability and pricing are made on experience, gut
feel, tradition, or rule of thumb.

Entropic Event:
◦ Rapidly changing market conditions defy traditional approaches.

RM Tactic:
◦ Generate maximum revenue by understanding the value cycle and optimally
timing the availability and price of the product to each micro-market
segment.
#7. Continually evaluate your revenue
opportunities
◦ Utilize acquired knowledge to continually evaluate your market and make changes accordingly.

◦ Gather data from all sources (customers, forecasts, employees, etc.)

◦ Shorten the timeframe in which data is gathered, synthesized and used to make changes to the
operation.

◦ Get data into the hands of employees at all levels of the organization.
#7. Continually evaluate your revenue
opportunities
Conventional Wisdom
◦ Information in the hands of a top manager ensures that the best decisions
will be made for the company.

Entropic Event:
◦ Breakdown of the mass market occurs, and the need for fast responses at
the micro-market level increases.

RM Tactic:
◦ Give decisions-support tools to the workers in the trenches to make dynamic
decisions at the micro-markets level.
 Perishable products and opportunities

Seasonal and other demand perks

Seven The product s value in different market segments

Uncertainties in Product wastage

Service
Industries Competition between individual and bulk purchases

Discounting to meet competition

Rapidly changing circumstances

Perishable products and opportunities
Types
◦ Literal
◦ Figurative

Revenue Management Tactics

◦ Use price balancing
◦ Maximize revenue through optimal timing
Seasonal and other demand perks
Issues
◦ Changes in seasonal, weekly and daily demand

Revenue Management Tactics

◦ Use price balancing
◦ Forecast at the micro-market level
The product s value in different market
segments
Issues
◦ Different groups of customers may place greater value on some products over
others.

Revenue Management Tactics

◦ Use market based pricing
◦ Use segment pricing
◦ Forecast at the micro-market level
 Product wastage
Issues
◦ Differs from perishability.
◦ Wastage is when a product is sold but then is unused.

Revenue Management Tactics

◦ Forecast at the micro-market level (including overbooking)
◦ Maximize revenue through optimal timing
Competition between individual and
bulk purchases
How do individual and bulk purchasing apply to the
hospitality, restaurant and airline industries?

What Revenue Management strategies should a company use

to maximize revenue?
Competition between individual and
bulk purchases
Revenue Management Tactics
◦ Use segment pricing
◦ Favor the most valuable customer
 Discounting to meet competition
Issue
◦ How to compete when targeting the same market segments?

Revenue Management Tactics

◦ Use market-based pricing
◦ Use segment pricing
Rapidly changing circumstances
Issue
◦ The market can change rapidly.

When Wayne Gretzky was asked what made him such an

outstanding player, he responded:
“All the other guys go where the puck is; I head
to where the puck is going to be.”
Rapidly changing circumstances

Revenue Management Tactics

◦ Forecast at the micro-market level
◦ Maximize revenue through optimal timing
◦ Arm the soldiers fighting the micro-market wars
HFTM 3120 –Service
Operations Analysis
Final Test Review Workshop
 Travel Time
X Y
Theresa Thompson is looking at
Origin 2 12 calculating the amount of time it will
take her to get from her pizza shop to
Destination 13 1 make a delivery at a house. She has
provided you the following information,
and asked you to let her know if she
should use Euclidean distance or
Metropolitan distance to get to the
delivery house. Also, she would like to
know how far it would be to either
option.
Tim Hortons
Tim Hortons is looking at opening a new store in Innisfil Ontario. They currently have one
location located in the West end that is 2,000 sq.ft., and the new location they are considering
opening would be located in the East end and it would be 1,200 sq.ft. Tim Hortons has provided
you with the following travel time and GeoDemographic information for Innisfil residents to get
to each location. Should they open the new store and why?
Areas where residents live (i)
Measurement Store j Sj Total/Average
Downtown (i=1) Uptown (i=2) East End (i=3) West End (i=4)
East End (j=1,New) 4 3 1 8
Travel Times
West End (j=2, Existing) 3 4 8 1
East End (j=1,New)
Attractiveness West End (j=2, Existing)
Total
East End (j=1,New)
Probability
West End (j=2, Existing)
(Pij)
Total
Innisfil Population (Ci) 2,000 1,200 900 2,800
GeoDemograp
hic Avg. Annual Coffee Expenditure (Bik) $520.00 $490.00 $420.00 $480.00
Information Total Area Spend on Coffee Eik
East End (j=1,New)
Sales West End (j=2, Existing)
Total
East End (j=1,New)
Market Share West End (j=2, Existing)
Total
Coffee Shop
Sally Simpson owns a small coffee shop in Alora and she is looking to
optimize her scheduling. She currently has three full time employees
that work every day but Monday and Tuesday and four part-time
employees. Using this information, prepare a schedule for Sally and let
her know if she needs to hire any additional employees.
Baristas Required
11
10
10
9
9
8
8
7
6 6
6
5
5
4
4
3
2
1
0
Monday Tuesday Wednesday Thursday Friday Saturday Sunday
West Jet
West Jet s flight from Nova Scotia to Toronto Island is considering overbooking its flights to
avoid flying with empty seats. The ticket agent is thinking of taking 11 reservations for an
airplane that has only 10 seats. He plans to provide a seat upgrade ($120) on the next flight to
anyone whose reservation is not honoured. What is the best overbooking policy, if the average
seat goes for $490?
Recent Data:

No-Shows 0 1 2 3 4
Frequency 32 21 16 7 4
SLR Example:
A local restaurant would like to know their sales if their wages increase to
$350,000. They have calculated a SLR of 2.24 and the intercept to be $180,000.
 Complete the
Table
Mondays Rooms Sold Forecast (Ft) Actual Dev Abs Dev

1.00 42.00
2.00 34.00
3.00 39.00 The 90 room Trail Top Inn is
4.00 44.00
looking at their rooms sales for the
last eight Mondays.
5.00 30.00
6.00 22.00 Using the information they have
provided, calculate a forecast using
7.00 44.00
a two-day moving average.
8.00 49.00
SAD
Also calculate the Trail Top Inns
SAD, MAD, MFE, and MSE.
MAD
MFE
MSE
 Toms Groceries
Toms groceries checkout counter is busy on the weekends. Tom has asked you to have a look at
his wait times and provide him your insights. He has told you that on average 6 customers arrive
per hour and that it takes an average of 8 minutes to serve one customer. He has provided a few
questions for you below:

Arrival rate =
Service rate =

What is the average number of customers in service?

What is the probability of having at least 2 customers in the system upon arrival?

What is the average number of customers in queue and average number of customers in the system?

What is the average waiting time in queue?

Next Class :
None
THANK YOU FOR AN
EXCELLENT
SEMESTER.