MS&E 260

INTRODUCTION TO O P E R AT I O N S M A N A G E M E N T

Richard Kim
Stanford University
Management Science and Engineering
Introduction/Course Overview
Class Agenda
• Course Overview
• Team
• Administration
• Background
• Objectives and Materials
• Preview to inventory control

3
Teaching Team
Instructor
• Richard Kim, PhD
• Contact: richhkim@stanford.edu, (310) 804-2625 mobile
• Office: Huang Engineering Center 212S
• Office Hours:
• Mondays 1:00pm – 2:00pm (Huang 212A)
• Tuesdays 10:00am – 11:00am (Huang 212A)
Course Assistants
• Sara Pak, spak7@stanford.edu
• Office Hours:
• Tuesdays 12:00pm – 1:00pm (Huang 203)
• Thursdays 12:00pm – 1:00pm (Huang 203)
• Lynn Zeng, lynnzeng@stanford.edu
• Office Hours:
• Wednesdays 3:00pm – 4:00pm (Huang 203)
• SCPD-only sessions TBD

4
Teaching Team: Richard Kim
• Educational Background:
• B.S. Mathematics, UCLA, 2004
• M.S. Systems Engineering, Naval Postgraduate School, 2014
• Ph.D. MS&E, Stanford University, 2018
• Professional Background:
• Chief Engineer and Technical Fellow at SAIC
• Systems engineer for the U.S. Air Force on space-related programs
• Space Based Space Surveillance (SBSS) System
• Joint Space Operations Center (JSpOC) Mission System (JMS)
• Professional Interests:
• Space systems engineering and integration; battle management, command,
and control (BMC2) systems; space policy

5
Teaching Team: Sara Pak
• Educational Background:
• B.S. Computer Science, Duke University, 2016
• M.S. MS&E, Stanford University, 2018
• Professional Background:
• Incoming Software Engineer (SWE) at Facebook
• 3 SWE Internships at Facebook
• Product Manager & SWE at LinkedIn
• Lead UX & SW Developer at Vigor (Chronic Respiratory Disease
Management Startup)
• Professional Interests:
• Software engineering (esp. iOS development), product management,
mobile health innovations

6
Teaching Team: Lynn Zeng
• Educational Background:
• B.A. Mathematics and Economics, Northwestern University, 2017
• M.S. MS&E, Stanford University, 2018
• Past Research Experience:
• Kellogg School of Management, Northwestern University 2015
• Area: Supply Chain
• Graduate School of Business, Stanford University, 2018
• Area: Auction, Game Theory, Optimization.
• Research Interests:
• Mechanism Design, Game Theory

7
Course Administration
• Prerequisites
• Probability at the level of MS&E 120 (Probabilistic Analysis)
• You are allowed to take the class without meeting the prerequisites, but at
your own risk
• The teaching team will provide some probability and optimization tutorial
help, but ultimately it is the student’s responsibility to understand these
foundational concepts
• admoniti estis! ☺
• Textbook (recommended)
• S. Nahmias. Production and Operations Analysis. 6th Edition. McGraw-
Hill/Irwin Series
• This book is great. We recommend to purchase if you are very
interested although it will not be used in class (it is in the library)
• Cost: >$200

8
Course Administration, continued
• Course website: Canvas and Piazza
• Grading:
• Homework (5 assignments) 30%
• Midterm 30%
• Final Exam 30%
• Participation 5%
• Final Thoughts Paper 5%
• Problem Sessions:
• Fridays 10:30AM – 11:20AM
• Exams (logistics still TBD)
• Midterm
• Friday, July 20, 2018, 10:30am – 11:50am (in class)
• One sheet of notes allowed
• Final
• Saturday, August 18, 2018, Location TBD

9
Course Outline
Week Topics
1 Course Overview, Inventory Management with Deterministic Demand
(EOQ Model)
2 Inventory Management with Uncertain Demand [Newsvendor Model]
(note: lecture given by Lynn)
3 Inventory Management with Uncertain Demand [(Q,R) Systems]
Service Level in (Q,R) Systems
Capacity and Waiting Times
4 Capacity and Waiting Times, Midterm Review, Midterm
5 Supply Chain Management; Guest Speaker: Colin Kessinger
Supply Chain Contracts
6 Revenue Management
7 Decision Analysis
Lean Techniques; Guest Speaker: Alejandro Martinez
8 Auction Design, Marketplace Design, and Kidney Exchange, Guest
Speaker: Professor Itai Ashlagi
Case Study: Managing Space Surveillance Systems 10
Course Objectives
1. Be familiar with a range of common problems in operations management
2. Understand how to model these problems mathematically
3. Understand the implication of these operations decisions
4. Understand Basic Laws of the ‘Physics’ of Business Operations: role of
uncertainty, Little’s Law, behavior of queues, etc.
5. Recognize and Understand Fundamental Tradeoffs: capacity-inventory-
service level, inventory-management, etc.
6. Principles of Revenue Management and Auctions: yield management,
pricing, auction design, etc.

In addition: learn to develop and apply mathematical and analytical models to

solve these problems

11
Five Basic Functions of an Organization

HUMAN RESOURCES
selection and management of labor
PRODUCTION &
RESEARCH & OPERATIONS SALES / MARKETING
DEVELOPMENT selection, organization,
sales of products or
develop new products and control of resources services to a customer
and services to produce a good or
service
FINANCE
accumulation and maintenance of resources to create and sustain the
organization

12
Competitive Priorities

COST
• Low cost operations

SPEED QUALITY
• Fast delivery time • High performance
• On-time delivery design
• Time-to-market • Consistent quality
FLEXIBILITY
• Customization
• Volume flexibility

13
U.S. Employment by Sector (1850-2000)

14
What is Operations Management (OM)?
• OM is the field concerned with the effective planning, scheduling, and
control of manufacturing and service entities
• OM is concerned with managing the production process that converts inputs
(i.e., material, manpower, capital, information, and energy) into outputs (i.e.,
goods and services)

15
Nominal Production Process

Value-added
Inputs transformation/ Outputs
activities

• Physical – manufacturing
• Material
• Locational – transportation
• Capital
• Exchange – retailing • Goods
• Manpower
• Storage – warehousing • Services
• Energy
• Physiological – health care
• Information
• Informational - telecommunications

16
Process Performance Measures

Value-added
Inputs transformation/ Outputs
activities

1. Efficiency: output/input
2. Quality: of output
3. Capacity: maximum throughput = output production rate
4. Cycle Time: time elapsed from input to output
5. Flexibility: volume and product

17
Process Physics: Little’s Law

Average number of
flow units in process 𝐿 Process throughput λ

Average process
cycle time 𝑊

Conservation of Flow (at equilibrium):

𝐿 =λ×𝑊

18
Classic Examples of Process Performance
• Automobile Industry
• Inputs: steel, parts, cash, manpower, land,
energy
• Process: assembly, storage, distribution
• Outputs: automobiles
• Airline Industry
• Inputs: capital, aircraft, facilities, fuel, labor
• Process: reservation, departure, flying,
arrival
• Outputs: transportation (people, freight)
• High Tech Industry
• Inputs: capital, facilities, components,
labor
• Process: manufacture, packaging,
distribution
• Outputs: hardware/software products
19
Lean Manufacturing of the Boeing 737

20
Other Examples of Process Performance
• Platforms (Uber, O-desk/Upwork, Airbnb)
• Inputs: labor, users, home owners
• Process: matching, assignments
• Outputs: renting, transportation
• Organ allocation (Kidney exchange, cadaver allocation)
• Inputs: patients, donors, cadaver organs
• Process: matching, allocation, congestion
• Outputs: transplants

21
 What is Supply Chain Management (SCM)?
• SCM is a set of approaches utilized to efficiently integrate suppliers,
manufacturers, distributors, and retailers so that merchandise is produced
and distributed at the right quantities, to the right locations, and at the right
time, to minimize system-wide costs while satisfying service level
requirements.

Source: D. Simchi-Levi, P. Kaminsky and

E. Simchi-Levi (2000) “Designing and 22
Managing the Supply Chain,” McGraw Hill
 Why Study SCM/OM?
• “The topic of supply chain talent management is becoming ever more
urgent, and increasing numbers of organizations are waking up to the
challenges of recruiting, rewarding and retaining the high-quality executives
they need to run, ramp up and refine their supply chains.”

23
SDCExecutive, April/May 2008
SCM/OM Examples
Amazon customers had to wait for
months to receive their Kindle: “We
can’t go into details about exact wait
times. … We’ve had to ramp up
manufacturing pretty significantly,
and ramping up a manufacturing
takes a little time.”
USA Today, 02/2008
Supply < Demand
Shortage
Lost Sales
Lost Profits
Nortel posted $19.2 billion loss
(write-downs of $15.2 billion) for the
second quarter of 2001. The
company laid off 20,000 employees.
Nortel CEO John Roth: “It was only
in October that customers stopped
beating me up for not shipping fast
enough, and now they say, ‘Ship
what? I don’t need it.’”
The Industry Standard, 06/2001
Supply > Demand
Excess Inventory
Increase in CGS
Lower Profits
24
Goal: Match Supply and Demand

Match
Toyota vs. GM
Supply Demand
HP vs. Dell
Southwest vs. AA

Tools
Inventory Management
Scheduling MS&E 260
Capacity Planning
Revenue Management
Decision Analysis
25
Models and Their Applications
• Mathematical Models
• Abstraction of the real thing (business problem or process)
• Some elements must be omitted
• Contains only relevant characteristics
• Involves a decision variable and represents a system in mathematical terms
• Used as a guide for management decision making
• Cogency vs. verisimilitude
• Applications of a Model
• Improve decision making
• Increase understanding of a system
• Make decision tradeoffs explicit

26
Example: Production System Design

Job Shop Production Line

(flexible) (efficient)

vs.

• How might management decide between the two systems?

• Use judgment (intuition) to make a decision
• Political considerations
• Just change over to new system
• Our approach: Model the system and experiment

27
Solution Methods
• HEURISTIC METHODS
✓ Quick
✓ Easy to use and understand
✓ Computationally cheap
✓ Sometimes the only way to
solve a problem
❖ May produce a good solution
but may not necessarily be
optimal
• OPTIMAL METHODS
✓ Guaranteed to obtain the best
solution
✓ Better outcomes
✓ Solution elegance
✓ Further analysis
✓ You know you have the best
solution
❖ May be cost and/or resource
prohibitive

Which is better?
Tradeoff between cost of solution and value of solution

28
Example of Solution Approaches
• Classic Knapsack Problem:
• Suppose a hiker is scaling a hill: What items should he take in his
knapsack?

Item Value Weight

A 30 25
B 70 40
C 50 20
D 55 35

• Objective: Maximize the value of the contents in the knapsack

• Constraint: Hiker cannot carry more than 80lbs in weight

29
Example of Solution Approaches, continued
• Classic Knapsack Problem:
• Suppose a hiker is scaling a hill: What items should he take in his
knapsack?

Item Value Weight V/W

A 30 25 1.20
B 70 40 1.75
C 50 20 2.50
D 55 35 1.57

• Objective: Maximize the value of the contents in the knapsack

• Constraint: Hiker cannot carry more than 80lbs in weight

30
Heuristic Solution Method
• Value: Item Value Weight V/W
• B → D → C: value = 125, weight = 75 A 30 25 1.20
• Weight: B 70 40 1.75
• C + A + D: value = 135, weight = 80 C 50 20 2.50
• V/W Ratio: D 55 35 1.57
• C → B → D: value = 120, weight = 60

• Lesson: different heuristics yield different solutions

• Question: Which heuristic is better, and why?

31
Optimal Solution Method
• Exhaustive search methods Item Value Weight V/W
A 30 25 1.20
• Dynamic programming methods
B 70 40 1.75
• Integer programming methods
C 50 20 2.50
• Linear programming methods D 55 35 1.57
• Convex optimization methods

max ෍ 𝑉𝑘 𝑋𝑘
𝑋𝑘
𝑘

s. t. ෍ 𝑊𝑘 𝑋𝑘 ≤ 80
𝑘

𝑋𝑘 = 0,1

32
 Linear Program Example
• Imagine that you manage a factory that produces four different types of wood
paneling. Each type of paneling is made by gluing and pressing together a
different mixture of pine and oak chips. The following table summarizes the
required amount of gluing, pressing, and mixture of wood chips required to
produce a pallet of 50 units of each type of paneling:
Resources Required per Pallet of Paneling Type

Tahoe Pacific Savannah Aspen

Glue (quarts) 50 50 100 50

Pressing (hours) 5 15 10 5

Pine chips (pounds) 500 400 300 200

Oak chips (pounds) 500 750 250 500

• In the next production cycle, you have 5,800 quarts of glue; 730 hours of
pressing capacity; 29,200 pounds of pine chips; and 60,500 pounds of oak
chips available. Further assume that each pallet of Tahoe, Pacific, Savannah,
and Aspen panels can be sold for profits of $450, $1,150, $800, and $400,
respectively.
• What is the optimal mix of paneling type to produce?

33
https://www.solver.com/product-mix-example
Linear Program Example, Problem Formulation
• Let:
• 𝑥1 = number of Tahoe pallets produced
• 𝑥2 = number of Pacific pallets produced
• 𝑥3 = number of Savannah pallets produced
• 𝑥4 = number of Aspen pallets produced

• Problem formulation:

max 450𝑥1 + 1150𝑥2 + 800𝑥3 + 400𝑥4

s. t. 50𝑥1 + 50𝑥2 + 100𝑥3 + 50𝑥4 ≤ 5800
5𝑥1 + 15𝑥2 + 10𝑥3 + 5𝑥4 ≤ 5800
500𝑥1 + 400𝑥2 + 300𝑥3 + 200𝑥4 ≤ 29200
500𝑥1 + 750𝑥2 + 250𝑥3 + 500𝑥4 ≤ 60500
𝑥1 , 𝑥2 , 𝑥3 , 𝑥4 ≥ 0

34
Convex Optimization Example
• Kwajalein Atoll is a research station in the South Pacific (located 2,100
nautical miles southwest of Honolulu, Hawaii) which is focused on
atmospheric and seismic information gathering. Logistics Solutions, Inc. is
scheduled to resupply Kwajalein Atoll in two weeks to deliver fuel,
equipment, supplies, and a new team of scientists to the station. However,
the integrity of the station’s runway is now in question, after an earthquake.
• Logistics Solutions, Inc. is planning to fly a reconnaissance mission over
Kwajalein Atoll using a Boeing 737 transport aircraft. The aircraft is
instrumented with special ground penetrating radars to examine the runway
and provide information to the mission planners about the ability of the
damaged runway to allow for the landing of the cargo plane.
• Logistics Solutions, Inc. considers 3 failure modes for the reconnaissance
mission; not enough fuel, bad weather, and radar malfunction. The round
trip is at the current range limit (in terms of fuel capacity) of the Boeing 737,
so the mission may fail by running out of fuel. Also, the Boeing 737 could
crash due to bad weather. Finally, the radar may malfunction, meaning that
even if the Boeing 737 makes it all the way back and lands safely, the
radars may not collect valid data.
35
Convex Optimization Example
• The engineering team can make last minute adjustments to the Boeing 737
to increase the probability that it returns with the data needed before the
resupply mission. The engineers have three options for increasing the
probability of mission success:
• (1) Add fuel capacity to the Boeing 737 – Extra tanks can be fabricated in
any reasonable size, and the Boeing 737 can carry the extra fuel load.
• (2) Make aircraft improvements to improve reliability in bad weather – The
engineering team can make adjustments and additions to the aircraft to
enhance its ability to fly in extreme cold and high icing conditions.
• (3) Perform tests to improve the radar configuration – Short test flights can
provide sample data and allow engineers to calibrate the radar.
• There are only 360 work hours available to spend on all improvements.
Every hour spent on any of the three improvements to the Boeing 737
lowers the probability of failure for the reconnaissance mission. Further, the
engineers must spend a minimum of 40 hours calibrating the radar
regardless of how they allocate the rest of their time. Similarly, they must
spend a minimum of 10 hours working on the fuel system. Beyond these 50
combined hours, improvement times are additive in their contribution to
P(Mission Failure).

36
Convex Optimization Example
• The engineers assess the following:
• (1) The probability of the Boeing 737 making it to Kwajalein Atoll and back to
base,𝑷𝑭𝑪, depends on the number of hours spent improving the fuel
capacity,𝒕𝑭𝑪, and is given by: 𝑷𝑭𝑪 𝒕𝑭𝑪 = 𝟎. 𝟓𝟎𝒆−𝟎.𝟎𝟑𝒕𝑭𝑪
• (2) If the Boeing 737 makes it to Kwajalein Atoll and performs the
measurements, by the time it returns to base, the base may experience
extreme weather conditions. The probability that the Boeing 737 crashes due to
the bad weather while landing at the base, given that it has sufficient fuel to
make the trip, depends on the number of hours,𝒕𝑾 , and is given by: 𝑷𝑾 𝒕𝑾 =
𝟎. 𝟎𝟓𝒆−𝟎.𝟎𝟒𝒕𝑾
• (3) If the Boeing 737 makes it to Kwajalein Atoll, performs the measurements,
makes it back to base, and lands safely, the probability that the radar
malfunctions and its data are bad depends on the number of hours spent doing
radar calibration,𝒕𝑹 , and is given by: 𝑷𝑹 𝒕𝑹 = 𝟎. 𝟎𝟗𝒆−𝟎.𝟎𝟏𝒕𝑹
• The engineers assume that these 3 failures are the only three failure modes,
and that the rare event approximation holds (to keep the calculations simple).
Therefore, the probability that the Boeing 737 mission fails is given by:
𝑷 𝑴𝒊𝒔𝒔𝒊𝒐𝒏 𝑭𝒂𝒊𝒍𝒖𝒓𝒆 = 𝑷𝑭𝑪 𝒕𝑭𝑪 + 𝑷𝑾 𝒕𝑾 + 𝑷𝑹 𝒕𝑹
• Your job is to help the planners allocate the engineering team’s time optimally to
ensure the greatest probability of success for the reconnaissance mission.

37
Convex Optimization Example, Problem Formulation
• Problem formulation:

max 𝑃𝑠 = 1 − 0.5𝑒 −0.03𝑡𝐹𝐶 − 0.05𝑒 −0.04𝑡𝑊 − 0.09𝑒 −0.01𝑡𝑅

s. t. 𝑡𝐹𝐶 + 𝑡𝑊 + 𝑡𝑅 ≤ 360
𝑡𝐹𝐶 ≥ 10
𝑡𝑅 ≥ 40
𝑡𝑊 ≥ 0

• Lagrangian:

𝐿
= 0.5𝑒 −0.03𝑡𝐹𝐶 + 0.05𝑒 −0.04𝑡𝑊 + 0.09𝑒 −0.01𝑡𝑅 + 𝜆 𝑡𝐹𝐶 + 𝑡𝑊 + 𝑡𝑅 − 360
+ 𝜆𝐹𝐶 (10 − 𝑡𝐹𝐶 ) + 𝜆𝑊 (−𝑡𝑊 ) + 𝜆𝑅 (40 − 𝑡𝑅 )

38
Convex Optimization Example, Problem Formulation
• Gradient (primal) conditions
𝜕𝐿
= −0.015𝑒 −0.03𝑡𝐹𝐶 + 𝜆 − 𝜆𝐹𝐶 = 0
𝜕𝑡𝐹𝐶
𝜕𝐿
= −0.002𝑒 −0.04𝑡𝑊 + 𝜆 − 𝜆𝑊 = 0
𝜕𝑡𝑊
𝜕𝐿
= −0.0009𝑒 −0.01𝑡𝑅 + 𝜆 − 𝜆𝑅 = 0
𝜕𝑡𝑅
• Feasibility (dual) conditions
𝑡𝐹𝐶 + 𝑡𝑊 + 𝑡𝑅 ≤ 360
𝑡𝐹𝐶 ≥ 10
𝑡𝑅 ≥ 40
𝑡𝑊 ≥ 0
• Orthogonality (complementarity slackness) conditions:
𝜆 𝑡𝐹𝐶 + 𝑡𝑊 + 𝑡𝑅 − 360 = 0
𝜆𝐹𝐶 (𝑡𝐹𝐶 − 10) = 0
𝜆𝑅 (𝑡𝑅 − 40) = 0
𝜆𝑊 𝑡𝑊 = 0
• Non-negativity conditions:
𝜆≥0
𝜆𝐹 ≥ 0
𝜆𝑅 ≥ 0
𝜆𝑊 ≥ 0
• Solution: 𝑡𝐹𝐶 = 145.22, 𝑡𝑊 = 58.84, 𝑡𝑅 = 155.09

39