Professional Documents
Culture Documents
t
e
t
o
s
e
e
M
D
t
h
a
t
d
o
n
t
b
e
d
o
n
e
b
y
M
D
C
P
a
t
i
e
h
a
v
e
A
c
t
i
v
i
t
i
e
s
h
a
v
e
t
o
b
Total paid time Time in practice Time booked
For appointments
Time with
patients
True value
add time
Prof. Christian Terwiesch
Source: Marcus, Terwiesch, Werner
Productivity
Li b l i / i Line balancing / capacity
sizing
Prof. Christian Terwiesch
Staffing / Capacity Sizing
So far: we started the process analysis with the process flow diagram / capacities
Often, demand can change over time
At Subway: More customers at noon than at 3pm
Typical situation in practice Given are:
Demand (forecasts)
Activities that need to be completed
Decision situation: how to build a staffing plan?
Two strategies:
Production smoothing (pre-produce)
Staff to demand
Prof. Christian Terwiesch
Line Balancing and Staffing to Demand
45
45
Time
46
Time
45
30
Takt
45
37
37
1 2 3
Operator
1 2 3
Operator
Labor content: 120seconds / unit
3,600 sec/hour
Takt: 3,600sec / 80 units=45 sec/unit
Target manpower=
120 sec/unit
Labor content: 120 seconds / unit
Demand: 80 units per hour
Target manpower=
=2.67 =>round up
St ff t d d t t ith th t kt ti dd i th f th
45 sec/unit
Prof. Christian Terwiesch
=>Staff to demand: start with the takt time and design the process from there
What Do You Do When Demand Doubles?
Ideal Case Scenario
Time
22.5
Takt
1 2 3
Operator
3,600 sec/hour
T kt 3 600 / 160 it 22 5 / it
4 5 6
Labor content: 120 seconds / unit
Demand: 160 units per hour
Takt: 3,600sec / 160 units=22.5 sec/unit
Target manpower=
=5 33=>roundup
120 sec/unit
22.5 sec/unit
Prof. Christian Terwiesch
=5.33 =>round up
Balancing the Line
Determine Takt time
Assign tasks to resource so that total processing times < Takt time
Make sure that all tasks are assigned g
Minimize the number of people needed (maximize labor utilization)
What happens to labor utilization as demand goes up?
Difference between static and dynamic line balancing
Prof. Christian Terwiesch
Line Balancing and Staffing to Demand
Actual Demand
Volume
Time
60
30
Takt time 2 minutes
Step
1
Step
2
Step
3
Step
4
Step
5
Step
6
Leveled Demand
Volume
60 60
Takt time 1 minute
S S S S S S
Takt time*
Takt
30
Step
1
Step
2
Step
3
Step
4
Step
5
Step
6
Takt
1
1
2
Volume flexibility
Ability to adjust to changingdemands
Resource planning
Man
power
6 6
Ability to adjust to changing demands
Often implemented with temporary workers
Keeps average labor utilization high
Prof. Christian Terwiesch
3
Productivity
Q il l i / Quartile analysis /
Standardization
Prof. Christian Terwiesch
Call Center Example
Two calls to the call center of a big retail bank
Both have the same objective (to make a deposit)
Different operators Different operators
Take out a stop watch
Time what is going on in the calls.
Prof. Christian Terwiesch
Beyond Labor Utilization: Quartile Analysis
Bi t d ti it diff f k l d i t t k
Prof. Christian Terwiesch
Biggest productivity differences for knowledge intense tasks
Source: Immaneni and Terwiesch
Example: Emergency Department
Analyzed data for over 100k patients in three hospitals
80 doctors and 109 nurses
Up to 260% difference between the 10
th
%-tile and the 90
th
%-tile
=> Dramatic productivity effects
Prof. Christian Terwiesch
Source: McCarthy, Ding, Terwiesch, Sattarian, Hilton, Lee, Zeger
Productivity
Productivity Ratios
Prof. Christian Terwiesch
Basic definitions of productivity
Productivity = Output units produced / Input used
Problems:
Output is hard to measure=> often times, use revenue instead
Multiple input factors (Labor, Material, Capital) => use one cost category
Example:
Labor productivity at US Airways
1995: Revenue: $6.98B Labor costs: $2.87B
2011: Revenue: $13.34B Labor costs: $2.41B
Labor productivity at SouthWest
1995: Revenue: $2.87B Labor costs: $0.93B
2011: Revenue: $13.65B Labor costs: $4.18B
Prof. Christian Terwiesch
Basic definitions of productivity
But WHY is one firm more productive than the other?
The ratio alone does not tell! Use the following trick:
Airline example:
Revenue / labor costs =Revenue/RPM * RPM/ASM * ASM / Employee * Employees/Labor costs
Revenue/Cost= Revenue/Output * Output/Capacity * Capacity/Cost
Operational yield Transformation
efficiency
1/unit cost of
capacity efficiency capacity
Prof. Christian Terwiesch
Labor Productivity Comparison between Southwest
and US Airways
Prof. Christian Terwiesch
Do Calculations in Excel
Productivity
Review Session
Prof. Christian Terwiesch
Tom and Jerry
Tom and J erry run an ice cream business out of their condo in Solana Beach, CA. They have purchased a
fully automated ice cream making machine from Italy (at a $30k price tag) that they put in their basement.
T i lli i dJ t th i k Oft ti h th t f i Tom is selling ice cream and J erry operates the ice cream maker. Often times, however, they run out of ice
cream and so J erry suggested purchasing a second ice cream maker.
Tom, however, wants to first look at the usage of the current ice cream maker and suggests an Overall
Equipment Effectiveness (OEE) analysis. Preliminary data suggests that: q p ( ) y y gg
J erry is not particularly skilled at programming the machine, which needs to be done when a new
batch of ice cream gets made. Instead of spending a negligible time per set-up, he presently spends
20 minutes. A batch of ice cream takes 1h in the machine, once the machine is set-up.
A new batch is only started if there exists sufficient time to complete the batch the same day before
7pm(includingthe 20 minute set upand the 1h production) 7pm (including the 20 minute set-up and the 1h production)
Since J erry started dating a woman from the WWF, he is fascinated by energy efficiency. So he turns
the machine off when he goes home at 7pm. As a result of this, the next morning, the machine has to
be cooled down to its desired operating temperature, which takes from 7am to 8am.
J erry is also not particularly diligent at following the recipe that Toms aunt in Italy had sent them. So
roughly one quarter of the produced ice cream has to be thrown away.
Every other Friday, J erry prefers to go surfing rather than showing up for work. On those days, the
business has to stay closed.
TJ 1: Howmanygoodbatches of ice creamare producedeach day J errycomes to work? TJ 1: How many good batches of ice cream are produced each day J erry comes to work?
TJ 2: What is the OEE of the ice cream maker? (use 12h per day as the available time)
Prof. Christian Terwiesch
Preliminary data suggests that:
J erry is not particularly skilled at programming the machine, which needs to be done when a new
batch of ice cream gets made. Instead of spending a negligible time per set-up, he presently spends
20 i t A b t h f i t k 1h i th hi th hi i t 20 minutes. A batch of ice cream takes 1h in the machine, once the machine is set-up.
A new batch is only started if there exists sufficient time to complete the batch the same day before
7pm (including the 20 minute set-up and the 1h production)
Since J erry started dating a woman from the WWF, he is fascinated by energy efficiency. So he turns
the machine off when he goes home at 7pm. As a result of this, the next morning, the machine has to g p g
be cooled down to its desired operating temperature, which takes from 7am to 8am.
J erry is also not particularly diligent at following the recipe that Toms aunt in Italy had sent them. So
roughly one quarter of the produced ice cream has to be thrown away.
Every other Friday, J erry prefers to go surfing rather than showing up for work. On those days, the
business has to stay closed business has to stay closed.
TJ 1: How many good batches of ice cream are produced each day J erry comes to work?
TJ 2: What is the OEE of the ice cream maker? (use 12h per day as the available time)
Prof. Christian Terwiesch
Penne Pesto
Penne Pesto is a small restaurant in the financial district of San Francisco. Customers order from a variety
of pasta dishes. The restaurant has 50 seats and is always full during the four hours in the evening. It is not
possible to make reservations at Penne; most guests show up spontaneously on their way home from work. p ; g p p y y
If there is no available seat, guests simply move on to another place. On average, a guest spends 50
minutes in the restaurant, which includes 5 minutes until the guest is seated and the waiter has taken the
order, an additional 10 minutes until the food is served, 30 minutes to eat, and 5 minutes to handle the
check-out (including waiting for the check, paying, and leaving). It takes the restaurant another 10 minutes to
cleanthe table andhave it be readyfor the next guests (of which there are always plenty) The average clean the table and have it be ready for the next guests (of which there are always plenty). The average
guest leaves $20 at Penne, including food, drink, and tip (all tips are collected by the restaurant, employees
get a fixed salary).
The restaurant has 10 waiters and 10 kitchen employees, each earning $90 per evening (including any
preparation, the 4 hours the restaurant is open, and clean-up). The average order costs $5.50 in materials,
including $4.50 for the food and $1 for the average drink. In addition to labor costs, fixed costs for the
restaurant include $500 per day of rent and $500 per day for other overhead costs.
The restaurant is open365days in the year andis full to the last seat evenon weekends andholidays The restaurant is open 365 days in the year and is full to the last seat even on weekends and holidays.
There is about $200,000 of capital tied up in the restaurant, largely consisting of furniture, decoration, and
equipment.
Define the return on invested capital as the ratio of the profits (PER YEAR) and the invested capital. You can
O C S O C f draw an ROIC tree in the same way that we drew a KPI tree in class. Simply have the ROIC as the root of
the tree instead of profits. Then answer the following questions.
a. How many guests will the restaurant serve in one evening?
b. What is the Returnon InvestedCapital (ROIC) for the owner of the restaurant?
Prof. Christian Terwiesch
b. What is the Return on Invested Capital (ROIC) for the owner of the restaurant?
c. Assume that you could improve the productivity of the kitchen employees and free up one person who
would be helping to clean up the table. This would reduce the clean-up to 5 minutes instead of 10 minutes.
What would be the new ROIC?
Assign Tasks to Workers
Consider the following six tasks that must be assigned to four workers on a conveyor-paced assembly line
(i.e., a machine-paced line flow). Each worker must perform at least one task.
Time to Complete Task (seconds / unit)
Task 1 30
Task 2 25
Task 3 35
Task 4 40
Task 5 15
Task 6 30
The current conveyor-pacedassemblyline configurationassigns the workers in the followingway: The current conveyor paced assembly line configuration assigns the workers in the following way:
Worker 1: Task 1
Worker 2: Task 2
Worker 3: Tasks 3, 4
Worker 4: Tasks 5, 6
a. What is the capacity of the current line?
b. Now assume that tasks are allocated to maximize capacity of the line, subject to the conditions that (1) a
worker can only perform two adjacent operations and (2) all tasks need to be done in their numerical order.
What is the capacity of this line now? p y
c. Now assume that tasks are allocated to maximize capacity of the line and that tasks can be performed in
any order. What is the maximum capacity that can be achieved?
d. After focusing on capacity in questions a-c, you now want to factor in demand in questions d-e. Demand is
50 units per hour. What is the takt time?
e What is the target manpower?
Prof. Christian Terwiesch
e. What is the target manpower?
f. How many workers will you need?
Prof. Christian Terwiesch
Customer Choice
Introduction
Prof. Christian Terwiesch
Customer Choice for HP DeskJ et Printers
How many HP printers are there on Amazon?
Why are there so many?
HP Deskjet printer (a look at Amazon)
1000 line
2000 line
3000 line
3050 Printer Series
3050 All-in-One
3050A Wireless All-in-One
4000 line
5000 line
6000 line
=> A printer for every day of the year
Prof. Christian Terwiesch
Customer Choice in Henry Fords Days
Henry Ford: You can have any color of a car, as long as it is black
Why did Ford not offer color?
Actually, he did
Production reasons to keep the cars (a) in one color (b) black
In this module, we discuss different types of product variety; we discuss the benefits, but
also explore the costs associated with variety
End of intro lecture
Prof. Christian Terwiesch
Forms of Variety - Fit Based Variety
Customers differ in shirt sizes
Each customer has a unique utility maximizing shirt size
The further you go away (in either direction) from that point,
the lower the utility
Hotellings linear city
Example: sizes, locations, arrival times
Source: Ulrich
Prof. Christian Terwiesch
Forms of Variety - Performance Based Variety
Each customer prefers the high end model
Customers differ in their valuation of quality (performance) and/or
their ability to pay
Vertical differentiation
Example: screen resolutions, mpg, processor speeds, weight
Source: Ulrich
Prof. Christian Terwiesch
Forms of Variety - Taste Based Variety
Customers differ in their preferences for taste
Often times, these preferences vary over time
Rugged landscape
Example: taste for food, music, artists
Prof. Christian Terwiesch
Economic Motives for Variety
Heterogeneous preferences of customers
Price discrimination
Variety seeking by consumers
Avoiding price competition in channel
Channel self space
Niche saturation and deterrence to market entry
Source: Ulrich
Prof. Christian Terwiesch
Customer Choice
Impact on process capacity
Prof. Christian Terwiesch
Ordering Custom Shirts
Custom shirts ordered online
Large variety of styles
Basically infinitely many sizes
Four weeks lead time
Minimum order: 5 shirts
Prof. Christian Terwiesch
Cutting Department
The pattern is programmed into a machine and/or a cutting template is created. This takes a
certain amount of set-up time IRRESPECTIVE of how many shirts will be produced
afterwards.
Sewing Department
Sewing Section Cut pieces of fabric are sewn together and inspected
Assembly Section - Responsible for assembling shirts and measuring the size.
Finishing Department
Responsible for ironing shirts before folding, packaging and delivery to customers.
Custom Tailored Shirts: Production Process
Source: http://hosting.thailand.com/MWT00255/process1.htm
Prof. Christian Terwiesch
Example: Cutting Machine for shirts
20 minute set-up time (irrespective of the number of shirts)
4 minute/unit cutting time
15 Shirts in a batch
Capacity calculation for the resource with set-up changes:
Batch Size
Set-up time + Batch-size*Time per unit
Capacity given Batch Size=
Process Analysis with Batching
Prof. Christian Terwiesch
Example Calculations
Cutting Section 1 Section 2 Finishing
Set-up time: 20 minutes - - -
Processing time: 4 min/unit 40 min/unit 30 min/unit 3 min/unit
Resources: 1 machine 8 workers 5 workers 1 worker
What is the capacity of the cutting machine with a batch size of 15?
Prof. Christian Terwiesch
Capacity
1/p
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
1
0
5
0
9
0
1
3
0
1
7
0
2
1
0
2
5
0
2
9
0
3
3
0
3
7
0
4
1
0
4
5
0
4
9
0
5
3
0
5
7
0
6
1
0
6
5
0
Batch Size
Large Batches are a Form of Scale Economies
Prof. Christian Terwiesch
Customer Choice
Choosing a good batch size
Prof. Christian Terwiesch
Production with large batches
Production with small batches
Cycle
Inventory
End of
Month
Beginning of
Month
Cycle
Inventory
End of
Month
Beginning of
Month
Produce Sedan
Produce Station wagon
Production with large batches
Production with small batches
Cycle
Inventory
End of
Month
Beginning of
Month
Cycle
Inventory
End of
Month
Beginning of
Month
Produce Sedan
Produce Station wagon
Production with large batches
Production with small batches
Cycle
Inventory
End of
Month
Beginning of
Month
Cycle
Inventory
End of
Month
Beginning of
Month
Produce Sedan
Produce Station wagon
Production with large batches
Production with small batches
Cycle
Inventory
End of
Month
Beginning of
Month
Cycle
Inventory
End of
Month
Beginning of
Month
Produce Sedan
Produce Station wagon
Large batch sizes lead to more inventory in the process
This needs to be balanced with the need for capacity
Implication: look at where in the process the set-up occurs
If set-up occurs at non-bottleneck => decrease the batch size
If set-up occurs at the bottleneck => increase the batch size
The Downside of Large Batches
Prof. Christian Terwiesch
Example Calculations
Cutting Section 1 Section 2 Finishing
Set-up time: 20 minutes - - -
Processing time: 4 min/unit 40 min/unit 30 min/unit 3 min/unit
Resources: 1 machine 8 workers 5 workers 1 worker
Prof. Christian Terwiesch
- one cart every 10 seconds
- 2 sec boarding time per passenger
- 2 sec exit time per passenger
- 2 minutes to go down the elevator
Batch Size
120sec + Batch-size*4sec
Capacity given Batch Size=
Batch Size
120sec + Batch-size*4sec
1/10 [units/sec] =
Batch Size = 20 units
How to Set the Batch Size An Intuitive Example
Prof. Christian Terwiesch
Batching is common in low volume / high variety operations
Capacity calculation changes:
This reflects economies of scale (similar to fix cost and variable cost)
You improve the process by:
Setting the batch size:
(a) If set-up occurs at the bottleneck => Increase the batch size
(b) If set-up occurs at a non-bottleneck => Reduce the batch size
(c) Find the right batch size by solving equation
Process Analysis with Batching: Summary
Batch Size
Set-up time + Batch-size*Time per unit
Capacity given Batch Size=
Prof. Christian Terwiesch
Customer Choice
Understanding the Diseconomies of Scale
Extra inventory
Prof. Christian Terwiesch
Production with large batches
Production with small batches
Cycle
Inventory
End of
Month
Beginning of
Month
Cycle
Inventory
End of
Month
Beginning of
Month
Produce Sedan
Produce Station wagon
Production with large batches
Production with small batches
Cycle
Inventory
End of
Month
Beginning of
Month
Cycle
Inventory
End of
Month
Beginning of
Month
Produce Sedan
Produce Station wagon
Production with large batches
Production with small batches
Cycle
Inventory
End of
Month
Beginning of
Month
Cycle
Inventory
End of
Month
Beginning of
Month
Produce Sedan
Produce Station wagon
Production with large batches
Production with small batches
Cycle
Inventory
End of
Month
Beginning of
Month
Cycle
Inventory
End of
Month
Beginning of
Month
Produce Sedan
Produce Station wagon
Large batch sizes lead to more inventory in the process
This needs to be balanced with the need for capacity
The Downside of Large Batches
Prof. Christian Terwiesch
General Definition of a Batch
Product A: Demand is 100 units per hour
Product B: Demand is 75 units per hour
The production line can produce 300 units per hour of either product
It takes 30 minutes to switch the production line from A to B (and from B to A)
How would you set the batch size?
Prof. Christian Terwiesch
Introducing a Third Product into the Product Line
Consider a company that has two products, product A and product B.
Product A: Demand is 100 units per hour
Product B: Demand is 75 units per hour
The production line can produce 300 units per hour of either product (takt time: 12 sec/unit)
It takes 30 minutes to switch the production line from A to B (and from B to A)
How would you set the batch size?
Batch Size
Set-up time + Batch-size*Time per unit
Required Flow Rate =
Batch Size
1 hour+ Batch-size/300 hour
175 units per hour =
Batch size = 420
Batch size for A= 420 * 100 / (100+75) = 240
Batch size for B=420-240=180
Prof. Christian Terwiesch
Introducing a Third Product into the Product Line
Now, the Marketing folks of the company add a third product. Total demand stays the same.
Product A1: Demand is 50 units per hour
Product A2: Demand is 50 units per hour
Product B: Demand is 75 units per hour
How would you set the batch size?
Prof. Christian Terwiesch
Introducing a Third Product into the Product Line
Now, the Marketing folks of the company add a third product. Total demand stays the same (maybe they do
this because they can raise prices). Say they offer product A in two colors.
Product A1: Demand is 50 units per hour
Product A2: Demand is 50 units per hour
Product B: Demand is 75 units per hour
How would you set the batch size?
Batch Size
Set-up time + Batch-size*Time per unit
Required Flow Rate =
Batch Size
1.5 hour+ Batch-size/300 hour
175 units per hour =
Batch size = 630
Batch size for A1= 630* 50 / (50+50+75) = 180
Batch size for A2= 630* 50 / (50+50+75) = 180
Batch size for B= 630* 75 / (50+50+75) = 270
Prof. Christian Terwiesch
Customer Choice
Pooling Effects / Demand
Fragmentation
Prof. Christian Terwiesch
Demand Fragmentation
You have 3 products (different shirt sizes)
Demand for each product could be 1, 2, or 3 with equal (1/3) probability
How good is your forecast FOR YOUR OVERALL SALES?
Prof. Christian Terwiesch
Customer Choice
Building Flexibility: SMED / Heijunka
Prof. Christian Terwiesch
The 6-stage SMED approach
Source: McKinsey Ops Training Material
Before/after
shutdown During shutdown Stage
Measure total
changeover time
1
Determine internal and
external activities
2
Move external activities to
before or after the shutdown
3
Improve the internal
activities
4
Improve the external
activities
5
External
Internal
Standardize procedures
6
Reduce set-up so that you can change models as often as needed
=> Mixed model production (Heijunka)
Prof. Christian Terwiesch Source: J ordan and Graves
Full Flexibility
Prof. Christian Terwiesch Source: J ordan and Graves
Flexibility vs Chaining
Prof. Christian Terwiesch Source: Moreno and Terwiesch
Pooling vs Chaining
Fords manufacturing network Nissans manufacturing network
Chaining is a form of partial flexibility (pooling light)
Does not require full flexibility, but relies on a clever product-to-plant assignment
Prof. Christian Terwiesch
Customer Choice
Strategies to deal with variety /
Investing in flexibility
Prof. Christian Terwiesch
Design for supply chain performance
Source: Ulrich
Prof. Christian Terwiesch
Design for supply chain performance
Source: Ulrich
Prof. Christian Terwiesch
Isolate the variable elements of the product
vs.
Source: Ulrich
Prof. Christian Terwiesch
Customer Choice
Limits to customization
Prof. Christian Terwiesch
Introduction
Design Variables Performance Specifications User Needs/ utility function
User Utility
processor
display
memory
package
XGA / SXGA / UXGA
video card
hard drive
portability
gaming
performance
Information
on screen
View from distance
MS-office performance
affordability
Resolution
price
Physical dimensions
Frames per second
HD capacity
Viewable area
RAM
Instructions per second (MIPS)
Data storage potential
Integrated devices
Design Variables Performance Specifications User Needs/ utility function
User Utility
processor
display
memory
package
XGA / SXGA / UXGA
video card
hard drive
portability
gaming
performance
Information
on screen
View from distance
MS-office performance
affordability
Resolution
price
Physical dimensions
Frames per second
HD capacity
Viewable area
RAM
Instructions per second (MIPS)
Data storage potential
Integrated devices
Source: Randall, Terwiesch, Ulrich
Prof. Christian Terwiesch
Introduction
Prof. Christian Terwiesch
Customer Choice
Review Session
Prof. Christian Terwiesch
Window Boxes
Metal window boxes are manufactured in two process steps: stamping and assembly. Each window box is made
up of three pieces: a base (one part A) and two sides (two part Bs).
The parts are fabricated by a single stamping machine that requires a setup time of 120 minutes whenever
switching between the two part types. Once the machine is set up, the processing time for each part A is one
minute while the processing time for each part B is only 30 seconds.
Currently, the stamping machine rotates its production between one batch of 360 for part A and one batch of 720
for part B. Completed parts move from the stamping machine to the assembly only after the entire batch is
complete.
At assembly, parts are assembled manually to form the finished product. One base (part A) and two sides (two
part Bs), as well as a number of small purchased components, are required for each unit of final product. Each
product requires 27 minutes of labor time to assemble. There are currently 12 workers in assembly. There is
sufficient demand to sell every box the system can make.
a. What is the capacity of the stamping machine?
b. What is the capacity of the overall process?
c. What batch size would you recommend for the process?
Prof. Christian Terwiesch
(Gelato) Bruno Fruscalzo decided to set up a small production facility in Sydney to sell to local restaurants that
want to offer gelato on their dessert menu. To start simple, he would offer only three flavors of gelato: fragola
(strawberry), chocolato (chocolate), and bacio (chocolate with hazelnut). Demand is 10kg/hour for Fragola, 15 for
chocolate, and 5 for Bacio.
Bruno first produces a batch of fragola, then a batch of chocolato, then a batch of bacio and then he repeats that
sequence. After producing bacio and before producing fragola, he needs 45 minutes to set up the ice cream
machine, he needs 30 minutes to change to Chocolato and 10 minutes to change to Bacio.
When running, his ice cream machine produces at the rate of 50 kg per hour no matter which flavor it is producing
(and, of course, it can produce only one flavor at a time).
a. Suppose Bruno wants to minimize the amount of each flavor produced at one time while still satisfying the
demand for each of the flavors. (He can choose a different quantity for each flavor.) If we define a batch to be
the quantity produced in a single run of each flavor, how many kilograms should he produce in each batch?
b. Given your answer in part (a), how many kilograms of fragola should he make with each batch?
Prof. Christian Terwiesch
SmartPhone
Apfel is a German company selling smart-phones. Presently, the company is only selling a 64GB
model. The marketing department recently proposed to add a 128 GB model. Preliminary data
suggests that (a) the margins for the product will increase (b) the total sales will remain the same
and will be split 50:50 between the two models (c) there exists a mild positive correlation in the
demand between the two models.
Consider the following statements:
1. The coefficient of variation of the 64GB phone will go down
2. The coefficient of variation of the 64GB phone will stay constant
3. The coefficient of variation of the 64GB phone will go up
4. It would be nice for the production and distribution process if the memory component, which is
the only difference between the two models, would be inserted early in the process.
5. It would be nice for the production and distribution process if the memory component, which is
the only difference between the two models, would be inserted late in the process.
Which of the above statements is true?
1+4
1+5
2+4
2+5
3+4
3+5
Response Time
Introduction
Prof. Christian Terwiesch
Example
Physician office
- Patients arrive, on average, every five minutes
- It takes ten minutes to serve a patient
- Patients are willing to wait
What is the implied utilization of the barber shop?
How long will patients have to wait?
Prof. Christian Terwiesch
Example
Physician office
- Patients arrive, on average, every five minutes
- It takes four minutes to serve a patient
- Patients are willing to wait
What is the utilization of the barber shop?
How long will patients have to wait?
Prof. Christian Terwiesch
A Somewhat Odd Service Process
Patient
Arrival
Time
Service
Time
1 0 4
2
3
4
5
10
15
4
4
4 4
5
6
15
20
25
4
4
4
7
8
9
30
35
40
4
4
4
10
11
12
45
50
55
4
4
4
Prof. Christian Terwiesch
7:00 7:10 7:20 7:30 7:40 7:50 8:00
12 55 4
A More Realistic Service Process
Patient
Arrival
Time
Service
Time
1 0 5
Patient 1 Patient 3 Patient 5 Patient 7 Patient 9 Patient 11
Patient 2 Patient 4 Patient 6 Patient 8 Patient 10 Patient 12
1
2
3
0
7
9
5
6
7
Time
7:10 7:20 7:30 7:40 7:50 8:00 7:00
4
5
6
12
18
22
6
5
2
3
7
8
9
25
30
36
4
3
4
2
c
a
s
e
s
9
10
11
36
45
51
4
2
2
0
1
N
u
m
b
e
r
o
f
Prof. Christian Terwiesch
12 55 3
2 min. 3 min. 4 min. 5 min. 6 min. 7 min.
Service times
Patient
Arrival
Time
Service
Time
Variability Leads to Waiting Time
Service time
Patient
1
2
3
4
0
7
9
12
5
6
7
6
Wait time
5
6
7
8
18
22
25
30
5
2
4
3
7:00 7:10 7:20 7:30 7:40 7:50 8:00
8
9
10
11
12
30
36
45
51
55
3
4
2
2
3
7:00 7:10 7:20 7:30 7:40 7:50
5
4
8:00
12 55 3
Inventory
3
2
1
Prof. Christian Terwiesch
y
(Patients at
lab)
0
7:00 7:10 7:20 7:30 7:40 7:50 8:00
The Curse of Variability - Summary
Variability hurts flow
With buffers: we see waiting times even though there exists excess capacity
Variability is BAD and it does not average itself out y g
New models are needed to understand these effects
Prof. Christian Terwiesch
W i i i d l Th
Response Time
Waiting time models: The
need for excess capacity
Prof. Christian Terwiesch
Modeling Variability in Flow
Outflow
N l iti l
Flow Rate
Minimum{Demand, Capacity}=Demand =1/a
Processing Buffer
No loss, waiting only
This requires u<100%
Outflow=Inflow
Inflow
Demand process is random
Look at the inter-arrival times
Processing
p: average processing time
a: average inter-arrival time
St Dev(inter arrival times)
Time
IA1 IA2 IA3 IA4
Same as activity time and service time
CVp =
St-Dev(processing times)
Average(processing times)
CVa =
Often Poisson distributed:
CVa =1
Constant hazard rate (no memory)
St-Dev(inter-arrival times)
Average(inter-arrival times)
Can have many distributions:
CVp depends strongly on standardization
Often Beta or LogNormal
Prof. Christian Terwiesch
Exponential inter-arrivals
Difference between seasonality and variability
Average flow
time T
Flow rate
The Waiting Time Formula
Inflow
Outflow
Inventory
waiting I
q
Increasing
Variability
Entry to system Departure Begin Service
Theoretical Flow Time
Utilization 100%
Time in queue T
q
Service Time p
Flow Time T=T
q
+p
Utilization 100%
Waiting Time Formula
|
|
| +
| |
2 2
CV CV
n utilizatio
Variability factor
|
|
.
|
\
| +
-
|
.
|
\
|
- =
2 1
p a
CV CV
n utilizatio
n utilizatio
Time Activity queue in Time
Prof. Christian Terwiesch
Service time factor
Utilization factor
Example: Walk-in Doc
Newt Philly needs to get some medical advise. He knows that his Doc, Francoise, has a patient arrive every 30
minutes (with a standard deviation of 30 minutes). A typical consultation lasts 15 minutes (with a standard
deviation of 15 minutes). The Doc has an open-access policy and does not offer appointments.
If Newt walks into Francoiss practice at 10am, when can he expect to leave the practice again?
Prof. Christian Terwiesch
Summary
Even though the utilization of a process might be less than 100%, it might still require long customer wait time
Variability is the root cause for this effect
As utilization approaches 100%, you will see a very steep increase in the wait time
If you want fast service, you will have to hold excess capacity
Prof. Christian Terwiesch
M W i i i d l /
Response Time
More on Waiting time models /
Staffing to Demand
Prof. Christian Terwiesch
Inventory
Waiting Time Formula for Multiple, Parallel Resources
Inflow
Outflow
Inventory
waiting I
q
in service I
p
Inflow
Outflow
Flow rate
E t t t D t B i S i Entry to system Departure Begin Service
Time in queue T
q
Service Time p
Flow Time T=T
q
+p
|
|
| +
|
|
|
| |
+
2 2
1 ) 1 ( 2 m
CV CV
n utilizatio time Activity
Waiting Time Formula for Multiple (m) Servers
Prof. Christian Terwiesch
|
|
.
|
\
| +
-
|
|
.
|
\
|
-
|
.
|
\
|
=
2 1
p a
CV CV
n utilizatio
n utilizatio
m
time Activity
queue in Time
Example: Online retailer
Customers send emails to a help desk of an online retailer every 2
minutes, on average, and the standard deviation of the inter-arrival time minutes, on average, and the standard deviation of the inter arrival time
is also 2 minutes. The online retailer has three employees answering
emails. It takes on average 4 minutes to write a response email. The
standard deviation of the service times is 2 minutes.
Estimate the average customer wait before being served.
Prof. Christian Terwiesch
Server Flow unit
Utilization(Note: make sure <1)
Summary of Queuing Analysis
am
p
p
m
a
u = =
1
*
1
Inventory
Utilization (Note: make sure <1)
CV CV
p
| | | |
| |
+
2 2
1 ) 1 ( 2
Inventory
waiting I
q
in service I
p
Time related measures
q
p a
m
q
p T T
CV CV
u
u
m
p
T
+ =
|
|
.
|
\
| +
-
|
|
.
|
\
|
-
|
.
|
\
|
~
+
2 1
2 2
1 ) 1 ( 2
Inflow
Outflow
T I *
1
Inventory related measures (Flow rate=1/a)
q p
p
q q
I I I
m u I
T
a
I
+ =
=
=
*
*
Entry to
system
Departure Begin
Service
Prof. Christian Terwiesch
q p
Waiting Time T
q
Service Time p
Flow Time T=T
q
+p
Staffing Decision
Customers send emails to a help desk of an online retailer every 2
minutes, on average, and the standard deviation of the inter-arrival time minutes, on average, and the standard deviation of the inter arrival time
is also 2 minutes. The online retailer has three employees answering
emails. It takes on average 4 minutes to write a response email. The
standard deviation of the service times is 2 minutes.
How many employees would we have to add to get the average wait
time reduced to x minutes?
Prof. Christian Terwiesch
What to Do With Seasonal Data
Measure the true demand data
Apply waiting model in each slice Apply waiting model in each slice
Slice the data by the hour (30min, 15min) Slice the data by the hour (30min, 15min)
Level the demand
Assume demand is stationary within a slice
Prof. Christian Terwiesch
Service Levels in Waiting Systems
0.8
1
Fraction of
customers who
have to wait x
seconds or less
Waiting times for those customers
h d t t d i di t l
90% of calls had to
wait 25 seconds or
less
0.4
0.6
who do not get served immediately
Fraction of customers who get served
0
0.2
0.4 Fraction of customers who get served
without waiting at all
0
0 50 100 150 200
Waiting time [seconds]
Target Wait Time (TWT)
Service Level = Probability{Waiting TimesTWT}
Example: Big Call Center
- starting point / diagnostic: 30%of calls answered within 20 seconds
Prof. Christian Terwiesch
starting point / diagnostic: 30% of calls answered within 20 seconds
- target: 80% of calls answered within 20 seconds
Response Time
Capacity Pooling
Prof. Christian Terwiesch
I d d t R
Managerial Responses to Variability: Pooling
Independent Resources
2x(m=1)
Example:
Processing time=4 minutes
Inter-arrival time=5 minutes (at each server)
m=1 Cva=CVp=1 m 1, Cva CVp 1
Tq=
Pooled Resources
(m=2)
Processing time=4 minutes
Inter-arrival time=2.5 minutes
m=2, Cva=CVp=1
Tq= Tq=
Prof. Christian Terwiesch
Managerial Responses to Variability: Pooling
Waiting
Time T
q
50.00
60.00
70.00
m=1
30.00
40.00
m=2
0.00
10.00
20.00
m=5
m=10
0.00
60% 65% 70% 75% 80% 85% 90% 95%
Utilization u
Prof. Christian Terwiesch
Pooling: Shifting the Efficient Frontier
Prof. Christian Terwiesch
Summary
What is a good wait time?
Fire truck or IRS?
Prof. Christian Terwiesch
Limitations of Pooling
Assumes flexibility
Increases complexity of work-flow Increases complexity of work flow
Can increase the variability of service time
I t t th l ti hi ith th t / f t th t Interrupts the relationship with the customer / one-face-to-the-customer
Group clinics Group clinics
Electricity grid / smart grid
Flexible production plants
Prof. Christian Terwiesch
The Three Enemies of Operations
Additional costs due to variability in
demand and activity times
Is associatedwith longer wait times
Use of resources beyond what is
needed to meet customer
requirements
Not addingvalue to the product
Is associated with longer wait times
and / or customer loss
Requires process to hold excess
capacity(idle time)
Variability
Not adding value to the product,
but adding cost
Reducing the performance of the
production system
7 different types of waste
Waste
capacity (idle time)
yp
Inflexibility
Waste Work Value-
adding
Waste Work Value-
adding
C t
Additional costs incurred because of supply
demand mismatches
Waiting customers or
Waiting(idle capacity)
Capacity
Customer
demand
Prof. Christian Terwiesch
Waiting (idle capacity)
Response Time
Scheduling / Access
Prof. Christian Terwiesch
Managerial Responses to Variability: Priority Rules
in Waiting Time Systems
Flowunits are sequenced in the waiting area (triage step) Flow units are sequenced in the waiting area (triage step)
Provides an opportunity for us to move some units forwards and some backwards
First-Come-First-Serve
- easy to implement
- perceived fairness
- lowest variance of waiting time lowest variance of waiting time
Sequence based on importance
- emergency cases
id tif i fit bl fl it - identifying profitable flow units
Prof. Christian Terwiesch
Managerial Responses to Variability: Priority Rules
in Waiting Time Systems
Service times:
A: 9 minutes
B: 10 minutes B: 10 minutes
C: 4 minutes
D: 8 minutes
A
B
9 min.
D
C
4 min.
D
C
19 min.
23 min.
Total wait time: 9+19+23=51min
B
A
12 min.
21 min.
Total wait time: 4+13+21=38 min Total wait time: 9+19+23=51min Total wait time: 4+13+21=38 min
Shortest Processing Time Rule
- Minimizes average waiting time
- Problem of having true processing times
Prof. Christian Terwiesch
Appointments
OpenAccess Open Access
Appointment systems
Prof. Christian Terwiesch
Response Time
Redesign the Service
PProcess
Prof. Christian Terwiesch
Reasons for Long Response Times
(And Potential Improvement Strategies)
Insufficient capacity on a permanent basis
=> Understand what keeps the capacity low
Demand fluctuation and temporal capacity shortfalls
Unpredictable wait times => Extra capacity / Reduce variability in demand
Predictable wait times =>Staff to demand / Takt time Predictable wait times > Staff to demand / Takt time
Long wait times because of low priority
=> Align priorities with customer value g p
Many steps in the process / poor internal process flow (often driven by handoffs
and rework loops)
=> Redesign the service process
Prof. Christian Terwiesch
http://www.minyanville.com/businessmarkets/articles/drive-thrus-emissions-fast-food-mcdonalds/5/12/2010/id/28261
The Customers Perspective
20 minutes
How much time does a patient spend on a primary care encounter?
Driving Parking CheckinVitals WaitingPCPAppt. Checkout LabsDrivehome
20minutes
Two types of wasted time:
Auxiliary activities required to get to value add activities (result of process location / lay-out)
Wait time (result of bottlenecks / insufficient capacity)
Total value add time of a unit
Flow Time Efficiency (or %VAT) =
Prof. Christian Terwiesch
Total time a unit is in the process
Flow Time Efficiency (or %VAT) =
Process Mapping / Service Blue Prints
Customer
actions
Walk into the
branch / talk to
agent
Customer
supplies more
data
Customer
supplies more
data
Sign contracts
Line of interaction
Onstage
actions
Collect basic
information
Request for more
data
Request for more
data
Explain final
document act o s
Line of visibility
Backstage
Pre Approval
t
data
Pre Approval
t
Backstage
actions
Line of internal
interaction
process; set up
workflow / account
responsibility
process; set up
workflow / account
responsibility
Support
processes
Run formal credit
scoring model
Prof. Christian Terwiesch
Source: Yves Pigneur
Process Mapping / Service Blue Prints
How to Redesign a Service Process
Move work off the stage
Example: online check-in at an airport
Reduce customer actions / rely on support processes y pp p
Example: checking in at a doctors office
Instead of optimizing the capacity of a resource, try to eliminate the step altogether
Example: Hertz Gold Check-in offers no value; go directly to the car
Avoid fragmentation of work due to specialization / narrow job responsibilities
Example: Loan processing / hospital ward
If customers are likely to leave the process because of long wait times, have the wait occur
later in the process / re-sequence the activities
Example: Starbucks Pay early, then wait for the coffee
Have the waiting occur outside of a line
Example: Restaurants in a shoppingmalls usingbuzzers Example: Restaurants in a shopping malls using buzzers
Example: Appointment
Communicate the wait time with the customer (set expectations)
Example: Disney
Prof. Christian Terwiesch
Response Time
L M d l Loss Models
Prof. Christian Terwiesch
Different Models of Variability
Waiting problems
Utilization has to be less than 100%
Impact of variability is on Flow Time
Loss problems
Demand can be bigger than capacity
Impact of variability is on Flow Rate
Pure waiting
problem, all customers
are perfectly patient.
All customers
enter the process,
some leave due to
their impatience
Customers do not
enter the process once
buffer has reached a
certain limit
Customers are lost
once all servers are
busy
Same if customers are patient Same if buffer size=0
S if b ff i i t l l Same if buffer size is extremely large
Variability is always bad you pay through lower flow rate and/or longer flow time
Prof. Christian Terwiesch
Buffer or suffer: if you are willing to tolerate waiting, you dont have to give up on flow rate
Analyzing Loss Systems
Resources
3 trauma bays (m=3) y ( )
Ambulances /
Helicopters
Trauma center moves to
diversion status once all
servers are busy
Demand Process Service Process
y
incoming patients
are directed to
other locations
One trauma case comes
in every 3 hours
(a=3 hours)
Patient stays in trauma bay
for an average of 2 hours
(p=2 hours) (a 3 hours)
a is the interarrival time
Exponential interarrival times
(p 2 hours)
p is the service time
Can have any distribution
Prof. Christian Terwiesch
Exponential interarrival times Can have any distribution
What is P
m
, the probability that all m resources are utilized?
Analyzing Loss Systems: Finding P
m
(r)
Define r = p / a
1 2 3 4 5
m
Example: r= 2 hours/ 3 hours
r=0.67
0.10 0.0909 0.0045 0.0002 0.0000 0.0000
0.20 0.1667 0.0164 0.0011 0.0001 0.0000
0.25 0.2000 0.0244 0.0020 0.0001 0.0000
0.30 0.2308 0.0335 0.0033 0.0003 0.0000
0.33 0.2500 0.0400 0.0044 0.0004 0.0000
Recall m=3
Use Erlang Loss Table
0.33 0.2500 0.0400 0.0044 0.0004 0.0000
0.40 0.2857 0.0541 0.0072 0.0007 0.0001
0.50 0.3333 0.0769 0.0127 0.0016 0.0002
0.60 0.3750 0.1011 0.0198 0.0030 0.0004
0.67 0.4000 0.1176 0.0255 0.0042 0.0006
0.70 0.4118 0.1260 0.0286 0.0050 0.0007
r = p / a
Find that P
3
(0.67)=0.0255
0.70 0.4118 0.1260 0.0286 0.0050 0.0007
0.75 0.4286 0.1385 0.0335 0.0062 0.0009
0.80 0.4444 0.1509 0.0387 0.0077 0.0012
0.90 0.4737 0.1757 0.0501 0.0111 0.0020
1.00 0.5000 0.2000 0.0625 0.0154 0.0031
Given P
m
(r) we can compute:
Time per day that systemhas to deny access
Prof. Christian Terwiesch
Time per day that system has to deny access
Flow units lost = 1/a * P
m
(r)
Implied utilization vs probability of having all servers
utilized: Pooling Revisited
Probability
0.6
utilized: Pooling Revisited
Probability
that all servers
are utilized
0.4
0.5
m=1
m=2
m=5
m=10
0.2
0.3
m=3
m 10
m=20
0
0.1
Implied utilization
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
Prof. Christian Terwiesch
Erl ang Loss Tabl e
m
1 2 3 4 5 6 7 8 9 10
0.10 0.0909 0.0045 0.0002 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.20 0.1667 0.0164 0.0011 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.25 0.2000 0.0244 0.0020 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.30 0.2308 0.0335 0.0033 0.0003 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.33 0.2500 0.0400 0.0044 0.0004 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
0.40 0.2857 0.0541 0.0072 0.0007 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000
Erlang Loss Table
0.40 0.2857 0.0541 0.0072 0.0007 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000
0.50 0.3333 0.0769 0.0127 0.0016 0.0002 0.0000 0.0000 0.0000 0.0000 0.0000
0.60 0.3750 0.1011 0.0198 0.0030 0.0004 0.0000 0.0000 0.0000 0.0000 0.0000
0.67 0.4000 0.1176 0.0255 0.0042 0.0006 0.0001 0.0000 0.0000 0.0000 0.0000
0.70 0.4118 0.1260 0.0286 0.0050 0.0007 0.0001 0.0000 0.0000 0.0000 0.0000
0.75 0.4286 0.1385 0.0335 0.0062 0.0009 0.0001 0.0000 0.0000 0.0000 0.0000
0.80 0.4444 0.1509 0.0387 0.0077 0.0012 0.0002 0.0000 0.0000 0.0000 0.0000
0.90 0.4737 0.1757 0.0501 0.0111 0.0020 0.0003 0.0000 0.0000 0.0000 0.0000
1.00 0.5000 0.2000 0.0625 0.0154 0.0031 0.0005 0.0001 0.0000 0.0000 0.0000
1 10 0 5238 0 2237 0 0758 0 0204 0 0045 0 0008 0 0001 0 0000 0 0000 0 0000 1.10 0.5238 0.2237 0.0758 0.0204 0.0045 0.0008 0.0001 0.0000 0.0000 0.0000
1.20 0.5455 0.2466 0.0898 0.0262 0.0063 0.0012 0.0002 0.0000 0.0000 0.0000
1.25 0.5556 0.2577 0.0970 0.0294 0.0073 0.0015 0.0003 0.0000 0.0000 0.0000
1.30 0.5652 0.2687 0.1043 0.0328 0.0085 0.0018 0.0003 0.0001 0.0000 0.0000
1.33 0.5714 0.2759 0.1092 0.0351 0.0093 0.0021 0.0004 0.0001 0.0000 0.0000
1.40 0.5833 0.2899 0.1192 0.0400 0.0111 0.0026 0.0005 0.0001 0.0000 0.0000
1.50 0.6000 0.3103 0.1343 0.0480 0.0142 0.0035 0.0008 0.0001 0.0000 0.0000
1.60 0.6154 0.3299 0.1496 0.0565 0.0177 0.0047 0.0011 0.0002 0.0000 0.0000
1.67 0.6250 0.3425 0.1598 0.0624 0.0204 0.0056 0.0013 0.0003 0.0001 0.0000
Probability{all m servers busy}=
1.70 0.6296 0.3486 0.1650 0.0655 0.0218 0.0061 0.0015 0.0003 0.0001 0.0000
1.75 0.6364 0.3577 0.1726 0.0702 0.0240 0.0069 0.0017 0.0004 0.0001 0.0000
r = p/a 1.80 0.6429 0.3665 0.1803 0.0750 0.0263 0.0078 0.0020 0.0005 0.0001 0.0000
1.90 0.6552 0.3836 0.1955 0.0850 0.0313 0.0098 0.0027 0.0006 0.0001 0.0000
2.00 0.6667 0.4000 0.2105 0.0952 0.0367 0.0121 0.0034 0.0009 0.0002 0.0000
2.10 0.6774 0.4156 0.2254 0.1058 0.0425 0.0147 0.0044 0.0011 0.0003 0.0001
2.20 0.6875 0.4306 0.2400 0.1166 0.0488 0.0176 0.0055 0.0015 0.0004 0.0001
2.25 0.6923 0.4378 0.2472 0.1221 0.0521 0.0192 0.0061 0.0017 0.0004 0.0001
2.30 0.6970 0.4449 0.2543 0.1276 0.0554 0.0208 0.0068 0.0019 0.0005 0.0001
y{ y}
!
) (
2 1
r r r
m
r
r P
m
m
m
=
2.30 0.6970 0.4449 0.2543 0.1276 0.0554 0.0208 0.0068 0.0019 0.0005 0.0001
2.33 0.7000 0.4495 0.2591 0.1313 0.0577 0.0220 0.0073 0.0021 0.0005 0.0001
2.40 0.7059 0.4586 0.2684 0.1387 0.0624 0.0244 0.0083 0.0025 0.0007 0.0002
2.50 0.7143 0.4717 0.2822 0.1499 0.0697 0.0282 0.0100 0.0031 0.0009 0.0002
2.60 0.7222 0.4842 0.2956 0.1612 0.0773 0.0324 0.0119 0.0039 0.0011 0.0003
2.67 0.7273 0.4923 0.3044 0.1687 0.0825 0.0354 0.0133 0.0044 0.0013 0.0003
2.70 0.7297 0.4963 0.3087 0.1725 0.0852 0.0369 0.0140 0.0047 0.0014 0.0004
2.75 0.7333 0.5021 0.3152 0.1781 0.0892 0.0393 0.0152 0.0052 0.0016 0.0004
2.80 0.7368 0.5078 0.3215 0.1837 0.0933 0.0417 0.0164 0.0057 0.0018 0.0005
2 90 0 7436 0 5188 0 3340 0 1949 0 1016 0 0468 0 0190 0 0068 0 0022 0 0006
!
...
! 2 ! 1
1
m
r r r
+ + + +
2.90 0.7436 0.5188 0.3340 0.1949 0.1016 0.0468 0.0190 0.0068 0.0022 0.0006
3.00 0.7500 0.5294 0.3462 0.2061 0.1101 0.0522 0.0219 0.0081 0.0027 0.0008
3.10 0.7561 0.5396 0.3580 0.2172 0.1187 0.0578 0.0249 0.0096 0.0033 0.0010
3.20 0.7619 0.5494 0.3695 0.2281 0.1274 0.0636 0.0283 0.0112 0.0040 0.0013
3.25 0.7647 0.5541 0.3751 0.2336 0.1318 0.0666 0.0300 0.0120 0.0043 0.0014
3.30 0.7674 0.5587 0.3807 0.2390 0.1362 0.0697 0.0318 0.0130 0.0047 0.0016
3.33 0.7692 0.5618 0.3843 0.2426 0.1392 0.0718 0.0331 0.0136 0.0050 0.0017
3.40 0.7727 0.5678 0.3915 0.2497 0.1452 0.0760 0.0356 0.0149 0.0056 0.0019
3.50 0.7778 0.5765 0.4021 0.2603 0.1541 0.0825 0.0396 0.0170 0.0066 0.0023
Prof. Christian Terwiesch
3.60 0.7826 0.5848 0.4124 0.2707 0.1631 0.0891 0.0438 0.0193 0.0077 0.0028
3.67 0.7857 0.5902 0.4191 0.2775 0.1691 0.0937 0.0468 0.0210 0.0085 0.0031
3.70 0.7872 0.5929 0.4224 0.2809 0.1721 0.0960 0.0483 0.0218 0.0089 0.0033
3.75 0.7895 0.5968 0.4273 0.2860 0.1766 0.0994 0.0506 0.0232 0.0096 0.0036
3.80 0.7917 0.6007 0.4321 0.2910 0.1811 0.1029 0.0529 0.0245 0.0102 0.0039
3.90 0.7959 0.6082 0.4415 0.3009 0.1901 0.1100 0.0577 0.0274 0.0117 0.0046
4.00 0.8000 0.6154 0.4507 0.3107 0.1991 0.1172 0.0627 0.0304 0.0133 0.0053
Response Time
R i Review
Prof. Christian Terwiesch
(My-law.com)
My-law.com is a recent start-up trying to cater to customers in search of legal services online. Unlike traditional
law firms, My-law.com allows for extensive interaction between lawyers and their customers via telephone and
the Internet This process is usedin the upfront part of the customer interaction largelyconsistingof answering the Internet. This process is used in the upfront part of the customer interaction, largely consisting of answering
some basic customer questions prior to entering a formal relationship. In order to allow customers to interact
with the firms lawyers, customers are encouraged to send e-mails to my-lawyer@My-law.com. From there, the
incoming e-mails are distributed to the lawyer who is currently on call. Given the broad skills of the lawyers,
each lawyer can respond to each incoming request.
E-mails arrive from 8 A.M. to 6 P.M. at a rate of 10 e-mails per hour (coefficient of variation
for the arrivals is 1). At each moment in time, there is exactly one lawyer on call,
that is, sitting at his or her desk waiting for incoming e-mails. It takes the lawyer, on average,
5 minutes to write the response e-mail The standarddeviationof this is 4 minutes 5 minutes to write the response e mail. The standard deviation of this is 4 minutes.
a. What is the average time a customer has to wait for the response to his/her e-mail, ignoring any transmission
times? Note: This includes the time it takes the lawyer to start writing the e-mail and the actual writing time.
b. How many e-mails will a lawyer have received at the end of a 10-hour day?
c. When not responding to e-mails, the lawyer on call is encouraged to actively pursue
cases that potentially could lead to large settlements. How much time on a 10-hour day
can a My-law.com lawyer dedicate to this activity
Prof. Christian Terwiesch
Jims Computer
J im wants to find someone to fix his computer. PC Fixers (PF) is a local service that offers such computer
repairs. A new customer walks into PF every 10 minutes (with a standard deviation of 10 minutes). PF has a
staff of 5 computer technicians Service times average around40 minutes (with a standarddeviationof 40 staff of 5 computer technicians. Service times average around 40 minutes (with a standard deviation of 40
minutes).
J C1. If J im walks into PF, how long must he wait in line before he can see a technician? (Only include the
waiting time, not any service time)
J C2. How many customers will, on average, be waiting for their computer to be fixed?
Prof. Christian Terwiesch
Real Compute
RealCompute offers real-time computing services. The company owns 4 supercomputers that can be accessed
through the internet. Their customers send jobs that arrive on average every 4 minutes (inter-arrival times are
exponentially distributed and, thus, the standard deviation of the inter-arrival times is 4 minutes). p y , , )
Each job takes on average 10 minutes of one of the supercomputers (during this time, the computer cannot
perform any other work). Customers pay $20 for the execution of each job. Given the time-sensitive nature of the
calculations, if no supercomputer is available, the job is redirected to a supercomputer of a partner company
calledOnComp which charges $40 per job to Real Compute (OnCompalways has supercomputer capacity called OnComp, which charges $40 per job to Real Compute (OnCompalways has supercomputer capacity
available).
RC1. What is the probability with which an incoming job can be executed by one of the supercomputers owned
by RealCompute?
RC2. How much does RealCompute pay on average to OnComp (in $s per hour)?
Prof. Christian Terwiesch
Contractor
A contractor building houses and doing renovation work has currently six projects planned for the season. Below
are the items, and the estimated times to complete them:
New construction at Springfield - 60 days
Bathroom remodeling at Herne - 10 days
Training time for solar roof installation - 2 days
Update web-site - 6 days y
Renovation of deck at Haverford - 8 days
New kitchen at Rosemont - 20 days
Suppose the contractor starts immediately with the first project, no other projects get added to this list, and the
contractor sequences themso as to minimize the average time the project waits before it gets started What will contractor sequences them so as to minimize the average time the project waits before it gets started. What will
the contractor be doing in 30 days from the start date of the first project?
Prof. Christian Terwiesch
Call Center
Consider a call center that has a constant staffing level. Because of increased demand in the morning,
the call center has a very high utilization in the morning and a very low utilization in the afternoon. Which
of the following will decrease the average waiting time in the call center?
(a) Add more servers
(b) Decrease the service time coefficient of variation
(c) Decrease the average service time
(d) Level the demandbetweenthe morninghours andthe afternoonhours (d) Level the demand between the morning hours and the afternoon hours
(e) All of the above
Prof. Christian Terwiesch
Prof. Christian Terwiesch
Quality
Introduction
Prof. Christian Terwiesch
Quality Introduction
I said that the worst thing about healthcare would be waiting, not true; worst thing are defects
Two dimensions of quality: conformance and performance
Our focus will be on conformance quality
Motivating example: the sinking ship / swiss cheese logic
Prof. Christian Terwiesch
Assembly Line Defects
Assembly operations for a Lap-top
9 Steps
Each of them has a 1% probability of failure
What is the probability of a defect?
Prof. Christian Terwiesch
The Duke Transplant Tragedy
Source: http://www.cbsnews.com/2100-18560_162-544162.html
17 year old J esica Santillandied following an organ transplant (heart+lung)
Mismatch in blood type between the donor and J esica
Experienced surgeon, high reputation health system
About one dozen care givers did not notice the mismatch
The offering organization did not check, as they had contacted the surgeon with another recipient in mind
The surgeon did not check and assumed the organization offering the organ had checked
It was the middle of the night / enormous time pressure / aggressive time line
A system of redundant checks was in place
A single mistake would have been caught
But if a number of problems coincided, the outcome could be tragic
Prof. Christian Terwiesch
Swiss Cheese Model
Source: J ames Reason
Barriers
Example:
3 redundant steps
Each of them has a 1% probability of failure
What is the probability of a defect?
Prof. Christian Terwiesch
The Nature of Defects
Assembly line example: ONE thing goes wrong and the unit is defective
Swiss cheese situations: ALL things have to go wrong to lead to a fatal outcome
Compute overall defect probability / process yield
When improving the process, dont just go after the bad outcomes, but also
after the internal process variation (near misses)
Prof. Christian Terwiesch
Quality
Defects / impact on flow
Prof. Christian Terwiesch
Impact of Defects on Flow
5 min/unit
4 min/unit
50% defect
Scrap
6 min/unit
Prof. Christian Terwiesch
Impact of Defects on Flow
5 min/unit
4 min/unit
30% defect
Rework
2 min/unit
Prof. Christian Terwiesch
Impact of Defects on Variability: Buffer or Suffer
Processing time of 5 min/unit at each resource (perfect balance)
With a probability of 50%, there is a defect at either resource and it takes 5 extra min/unit at the resource to rework
=>What is the expected flow rate?
Prof. Christian Terwiesch
The Impact of Inventory on Quality
Inventory takes pressure off the resources (they feel buffered): demonstrated behavioral effects
Expose problems instead of hiding them
I
n
v
e
n
t
o
r
y
i
n
p
r
o
c
e
s
s
Buffer argument:
Increase inventory
Toyota argument:
Decrease inventory
Prof. Christian Terwiesch
Operations of a Kanban System: Demand Pull
Visual way to implement a pull system
Amount of WIP is determined by
number of cards
Kanban = Sign board
Work needs to be authorized by demand
Authorize
production
of next unit
Prof. Christian Terwiesch
Quality
Six sigma and process
capability
Prof. Christian Terwiesch
Gurkenverordnung:
http://de.wikipedia.org/wiki/Verordnung_(EWG)_Nr._1677/88_(Gurkenverordnung)
Failure of a pharmacy
Intro: two types of variability
Prof. Christian Terwiesch
M&M Exercise
A bag of M&Ms should be between 48 and 52g
Measure the samples on your table:
Measure x1, x2, x3, x4, x5
Compute the mean (x-bar) and the standard deviation
Number of defects
All data will be compiled in master spread sheet
Yield = %tage of units according to specifications
How many defects will we have in 1MM bags?
Prof. Christian Terwiesch
Process capability measure
Estimate standard deviation in excel
Look at standard deviation relative to specification limits
3
Upper
Specification
Limit (USL)
Lower
Specification
Limit (LSL)
X-3
A
X-2
A
X-1
A
X X+1
A
X+2 X+3
A
X-6
B
X X+6
B
Process A
(with st. dev
A
)
Process B
(with st. dev
B
)
6
LSL USL
C
p
x C
p
P{defect} ppm
1 0.33 0.317 317,000
2 0.67 0.0455 45,500
3 1.00 0.0027 2,700
4 1.33 0.0001 63
5 1.67 0.0000006 0,6
6 2.00 2x10
-9
0,00
Measure Process Capability:
Quantifying the Common Cause Variation
Prof. Christian Terwiesch
Not just the mean is important, but also the variance
Need to look at the distribution function
The Concept of Consistency:
Who is the Better Target Shooter?
Prof. Christian Terwiesch
Quality
Two types of variation
Prof. Christian Terwiesch
Common Cause Variation (low level)
Common Cause Variation (high level)
Assignable Cause Variation
Need to measure and reduce common cause variation
Identify assignable cause variation as soon as possible
What is common cause variation for one person might be
assignable cause to the other
Two Types of Variation
Prof. Christian Terwiesch
M&M Exercise
Analysis of new sample in production environment
=> Show this in Excel
Prof. Christian Terwiesch
Time
Process
Parameter
Upper Control Limit (UCL)
Lower Control Limit (LCL)
Center Line
Track process parameter over time
- average weight of 5 bags
- control limits
- different from specification limits
Distinguish between
- common cause variation
(within control limits)
- assignable cause variation
(outside control limits)
Detect Abnormal Variation in the Process:
Identifying Assignable Causes
Prof. Christian Terwiesch
Statistical Process Control
Capability
Analysis
Conformance
Analysis
Investigate for
Assignable Cause
Eliminate
Assignable Cause
Capability analysis
What is the currently "inherent" capability of my process when it is "in control"?
Conformance analysis
SPC charts identify when control has likely been lost and assignable cause
variation has occurred
Investigate for assignable cause
Find Root Cause(s) of Potential Loss of Statistical Control
Eliminate or replicate assignable cause
Need Corrective Action To Move Forward
Prof. Christian Terwiesch
Quality
Detect / Stop / Alert
Prof. Christian Terwiesch
7
1
2 3 4 5
6
8
ITAT=7*1 minute
3
1
2
4
ITAT=2*1 minute
Good unit
Defective unit
Information Turnaround Time
Inventory leads to a longer ITAT (Information turnaround time) => slow feed-back and no learning
Assume a 1 minute processing time
Prof. Christian Terwiesch
Cost of a Defect: Catching Defects Before the Bottleneck
What is the cost of a defect?
Defect detected before bottleneck
Defect detected after bottleneck
Bottleneck
Buy pasta /
ingredients for $2
per meal
Prepare
Cook
Serve
Serve food for $20
per meal
Prof. Christian Terwiesch
Detecting Abnormal Variation in the Process at
Toyota: Detect Stop - Alert
Source: www.riboparts.com, www.NYtimes.com
Jidoka
If equipment malfunctions / gets out of
control, it shuts itself down automatically to
prevent further damage
Requires the following steps:
Detect
Alert
Stop
Andon Board / Cord
A way to implement J idoka in an assembly line
Make defects visibly stand out
Once worker observes a defect, he shuts down
the line by pulling the andon / cord
The station number appears on the andon
board
Prof. Christian Terwiesch
Detect,
stop,
alert
J idoka
Andon cord
Root-
cause
problem-
solving
Ishikawa Diagram
Kaizen
Avoid
Poka Yoke
Build-in quality
Two (similar) Frameworks for Managing Quality
Toyota Quality System
Capability
Analysis
Conformance
Analysis
Investigate for
Assignable
Cause
Eliminate
Assignable
Cause
Six Sigma System
Some commonalities:
Avoid defects by keeping variation out of the process
If there is variation, create an alarm and trigger process improvement actions
The process is never perfect you keep on repeating these cycles
Prof. Christian Terwiesch
Quality
Problem solve / improve
Prof. Christian Terwiesch
Root Cause Problem Solving
Ishikawa Diagram
A brainstorming technique of what might
have contributed to a problem
Shaped like a fish-bone
Easy to use
Pareto Chart
Maps out the assignable causes of a problem
in the categories of the Ishikawa diagram
Order root causes in decreasing order of
frequency of occurrence
80-20 logic
Prof. Christian Terwiesch
The Power of Iterative Problem-solving
M
o
d
e
l
s
R
e
a
l
i
t
y
Prof. Christian Terwiesch
Root Cause Problem Solving
Ishikawa Diagram
A brainstorming technique of what might
have contributed to a problem
Shaped like a fish-bone
Easy to use
Pareto Chart
Maps out the assignable causes of a problem
in the categories of the Ishikawa diagram
Order root causes in decreasing order of
frequency of occurrence
80-20 logic
Prof. Christian Terwiesch
Conclusion
Lean Operations
Prof. Christian Terwiesch
The Ford Production System
Influenced by Taylor; optimization of work
The moving line / big machinery => focus on utilization
Huge batches / long production runs; low variety
Produced millions of cars even before WW2
Model built around economies of scale
=> Vehicles became affordable to the middle class
Prof. Christian Terwiesch
The Toyota Production System
Toyota started as a maker of automated looms
Started vehicle production just before WW2
No domestic market, especially following WW2
Tried to replicate the Ford model (produced about 10k vehicles)
No success due to the lack of scale
Around 1950, TPS was born and refined over the next 30 years
Systematic elimination of waste
Operating system built around serving demand
Prof. Christian Terwiesch
Introduction
1903
1st car
1908
1st Model
T
1911
F.W.
Taylor
1913
1st
moving
line
1923
2.1
million
vehicles/
year
Cost
USD/
unit
1916 1904 1926
950
360
290
Key idea of TPS: systematic elimination
of non-value-adding activities
1933
Founded
1946
Major
strike
1950
Start of
TPS
1960s
Supplier
develop-
ment
1980s
Trans-
plants
Mass production driven by economies
of scale impossible
Low production volume (1950):
GM 3,656,000 Toyota 11,000
Low productivity (J apan 1/9 of US)
Lack of resources
Taylorism: Standardized parts and work
patterns (time studies)
Moving line ensuring working at same pace
Process driven by huge, rapid machinery
with inflexible batch production
Source: McKinsey
Key idea of Ford: cost reduction through
cheap labor and economies of scale
Prof. Christian Terwiesch
Zero non-value added
activities (muda)
Production flow synchronized with demand (JIT)
One-unit-at-a-time flow
Mixed model production (heijunka)
Match production demand based on Takt time
Pull instead of push
Supermarket / Kanban
Make-to-order
Quality methods to reduce defects
Fool-proofing (poka-yoke) and visual feed-back
Detect-stop-alert (J idoka)
Defects at machines (original J idoka)
Defects in assembly (Andoncord)
Flexibility
Standardization of work
Worker involvement
Quality circles (Kaizen)
Fishbone diagrams (Ishikawa)
Skill development / X-training
Reduction of Variability
Quartile Analysis
Standard operating procedures
Adjustment of capacity to meet takt-time
Reduce inventory to
expose defects
Toyota Production System: An Overview
Prof. Christian Terwiesch
The Three Enemies of Operations
Is associated with longer wait times
and / or customer loss
Requires process to hold excess
capacity (idle time)
Buffer or suffer
Often times: quality issues
Variability
Use of resources beyond what is
needed to meet customer
requirements
7 different types of waste
OEE framework
Lean: do more with less
Waste Work Value-
adding
Waste Work Value-
adding
Waste
Inflexibility
Additional costs incurred because of supply
demand mismatches
Waiting customers or
Waiting (idle capacity)
Capacity
Customer
demand
Source: Reinecke / McKinsey
Prof. Christian Terwiesch
Quality
Review Questions
Prof. Christian Terwiesch
Pharmacy Medication Error
A pharmacy in a Philadelphia suburb wants to investigate the likelihood of making a medication error. There
are two ways in which a patient can end up with the wrong medication:
- In about 2% of the cases, the doctor fills out the prescription incorrectly. Nobody in the pharmacy catches
these errors
- In about 1% of the cases, the pharmacist makes a mistake in picking the medication according to the
prescription. The pharmacy has an internal quality inspection process that catches about 97% of the errors
made by the pharmacist.
Another source of quality control is the patients. The pharmacy estimates that about half of the errors made by
the physician are recognized by the patient. However, the patient is only able to recognize 10% of the
mistakes done at the pharmacy.
What is the likelihood that the patient is presented with a wrong medication?
What is the likelihood that the patient leaves the pharmacy with the wrong medication?
Prof. Christian Terwiesch
Four Step Process with Rework and Scrap
Consider the following four step assembly operation with quality problems. All resources are staffed with one
operator.
- The first resource has a processing time of 4 minutes per unit
- The second resource has a processing time of 3 minutes per unit. This process suffers from a high yield
loss and 50% of all products have to be scrapped after this step.
- The third resource also suffers from quality problems. However, instead of scrapping the product, the third
resource reworks it. The processing time at the third resource is 5 minutes per unit. In the 30% of the
products in which the product needs to be reworked, this extends to a total (initial processing time plus
rework) processing time of 10 minutes per unit. Rework always leads to a non-defective unit.
- No quality problems exist at the first and final resource. The processing time is 2 minutes per unit.
For every unit of demand, how many units have to flow through the third step in the process?
Where in the process is the bottleneck?
What is the process capacity?
Prof. Christian Terwiesch
Chicken Eggs
A farmer focusing on the production of eco-friendly chicken eggs collects the following data about his output.
In a sample of 50 eggs, the farmer finds the average egg to weigh 47 grams. The standard deviation of the
egg weight is 2 grams and the distribution of weights resembles a normal distribution reasonably closely.
The farmer can sell the eggs to a local distributor. However, they have to be in the interval between 44
grams and 50 grams (i.e., the lower specification limit is 44 grams and the upper specification limit is 50
grams).
What is the capability score of the eco-friendly chicken egg operation?
What percentage of the produced eggs fall within the specification limits provided by the local distributor?
By how much would the farmer have to reduce the standard deviation of the operation if his goal were to
obtain a capability score of Cp=2/3 (i.e., get 4.5% defects)?
Prof. Christian Terwiesch
Process capability measure
Estimate standard deviation in excel
Look at standard deviation relative to specification limits
3
Upper
Specification
Limit (USL)
Lower
Specification
Limit (LSL)
X-3
A
X-2
A
X-1
A
X X+1
A
X+2 X+3
A
X-6
B
X X+6
B
Process A
(with st. dev
A
)
Process B
(with st. dev
B
)
6
LSL USL
C
p
x C
p
P{defect} ppm
1 0.33 0.317 317,000
2 0.67 0.0455 45,500
3 1.00 0.0027 2,700
4 1.33 0.0001 63
5 1.67 0.0000006 0,6
6 2.00 2x10
-9
0,00
Measure Process Capability:
Quantifying the Common Cause Variation
Prof. Christian Terwiesch
Toyota Word Matching
Please write the letter corresponding to the most appropriate example or definition from choices (a k below)
on the blank line next to each word below.
a) Examples of this include: workers having to make unnecessary movements (i.e. excessive reaching or
walking to get tools or parts), working on parts that are defective and idle time.
b) A system that enables a line worker to signal that he or she needs assistance from his or her supervisor, for
example in the case of a defect. Used to implement the J idoka principle.
c) A brainstorming technique that helps structure the process of identifying underlying causes of an (usually
undesirable) outcome
d) As an example of this philosophy, workers at Toyota often times make suggestions for process improvement
ideas.
e) A method that controls the amount of work-in-process inventory
f) If an automotive assembly plant used this technique, the adjacent cars on an assembly line would be mixed
models (e.g. Model A with sunroof, Model A without sunroof, Model B, Model B with sunroof), in proportions
equal to customer demand.
g) Making production problems visible and stopping production upon detection of defects
Please only add ONE LETTER to each of the following terms:
Kanban ____
Muda ____
Heijunka ____
Andon cord ____
Kaizen ____
Ishikawa ____
J idoka ____