1

How does SB obtain America’s

finest dark chocolate?

What type of process is it?

2

So, why is Jim Harris worried?

Planned capacity
increases:
• + 50%
• +100%
• +150%
3

Should Jim Harris buy the Ball Mill?

• Cost: $ 300k
• Capacity: 1,400 Kg
• Cycle time: 5 hours
• Must be used with the
conche
• Value of 1 full conche
batch = $ 30K
• Operating Margin:
40%
HOW SHOULD JIM MAKE A DECISION?
 SB process analysis:

• Is adding the ball mill the right decision?

• What is the new bottleneck?

 Process Map of Making Chocolate:
From Cocoa Pods to Chocolate Bars
Cocoa Unroasted
Cleaner Roaster Roasted beans
pods cocoa beans

Tempering
3-ounce bar and molding Conche Melangeur Winnower
 PayBack Analysis
• Ball Mill cost =
• Additional n. of conche-equivalent batches per
month =
• Additional profit per month =

• Pay back time =

7

PayBack Analysis

• Ball Mill cost = $ 300,000

• Additional n. of conche-equivalent batches per
month = 19
• Additional profit per month = 19 * $30,000 * 0.4 =
= $ 228,000
• Pay back time = 300,000/228,000 = 1.32 months
• Even with more conservative assumptions it
makes sense to invest in the new machine
 Bottleneck sequence
1. Conche  add Ball Mill
Bottleneck shifts to:

2. Roaster  Extra time (do more shifts)

Bottleneck shifts to:

3. Tempering and molding  what can be

done?
8
9

Should Schaffenberger outsource

tempering and molding?

• What are the advantages of outsourcing for Scharffen Berger?

• What are the risks / potential costs for Scharffen Berger?

10

Should Schaffenberger outsource

tempering and molding?

• What are the advantages of outsourcing for Scharffen Berger?

– Avoid bottleneck, leverage existing capacity and hence lower cost
solution (“prohibitively expensive to buy larger molding machine”)

• What are the risks / potential costs for Scharffen Berger?

– Quality control, transportation tricky and expensive, cost of bad batch
reaching market
11

Should Schaffenberger outsource

tempering and molding?

From Scharffen Berger’s (old) website:

• “Think of tempering as the alchemy of chocolate.
Tempered well, chocolate is glossy, breaks with a crisp
snap and melts smoothly in the mouth. If this step is
skipped or done poorly, chocolate becomes gray,
crumbly and almost dusty as it hardens.”
 Should Scharffenberger outsource
tempering and molding ?

• What should Jim worry about? What not (as much)?

Quality /
service
High
?
Scharffen
Berger

Hershey’s
Low Efficiency
Low High

12
 Epilogue 1

• Ball Mill installed successfully in July 2005

• Hershey went gourmet and in 2005 bought
ScharffenBerger and Joseph Schmidt to
form the Artisan Confections Company
(ACC)
• Harris promoted head of ACC

13
Presentation Title
14

Epilogue 2

“America's original craft

chocolate reclaims its bean-to-
bar roots with a founder-mentality
that will ensure quality and spur
innovation.”
Managing Service Systems
and Variability

Bilal Gokpinar
 Typical Questions
• How many workers should be hired?
• Consequences of a 20% increase in demand?
• How many counters should be opened to maintain customer wait
below 10 minutes?
• How many assembly stations are needed to maintain backorders
below 20?
• How often will all 6 surgery operating rooms be full?
• How will congestion at LHR change if a 3rd runway is built?
 Capacity

Too high:
• Capital expenses
• Labour expenses
Speed Capacity • Resource waste
• Environmental damage
Quality Cost
• Cash flow issues
Too low:
Flexibility
• Customer wait
• Brand damage
• Customer dissatisfaction
• Lost sales
• Low employee morale
• High turnover
Sources of Variability?
 Variability
DEMAND
• Variability in demand:
• Cyclical (predictable, such as seasonal, daily,
hourly)
• Stochastic (unpredictable)
• Variability in the size of the batch of
patients
SUPPLY
• Variability in the processing times
 Variability Leads to Waiting Time
Caller 1
Caller 2 • capacity: 15 calls/hour
Caller 3
Caller 4
• input rate: 12 calls/hour
Caller 5
Caller 6
Caller 7
Caller 8
 process without variability
Caller 9
Caller 10
Caller 11
Caller 12 Time
7:00 7:10 7:20 7:30 7:40 7:50 8:00

Service time

Caller 1
Caller 2
process with variability  Caller 3
Caller 4
Caller 5
Caller 6
Caller 7
Caller 8
Caller 9
Caller 10
Waiting time Caller 11
Caller 12
Time
7:00 7:10 7:20 7:30 7:40 7:50 8:00
Queueing Everywhere
 The Fundamental
Queueing Relationship

Waiting time

More variability

0 0.2 0.4 0.6 0.8 1

Utilisation rate

Waiting = (service time) x (utilisation) x (variability)

Reducing Variability

Bilal Gokpinar
Variability Reduction in a Restaurant?
Variability Reduction in a Restaurant? (cont.)

• Reduce cyclical variability in demand:

• Early-bird specials
• Lower lunch price
• Reduce stochastic variability in demand:
• Add an inventory buffer (bar)
Reservations: eliminate variability in batch size
• (Alternative: make capacity more flexible,
i.e. reconfigurable tables)
• Reduce variability in processing time
(short menu…)
What Does a Limited Menu Do?
• Shortens customer order time (increases
capacity)
• Reduces variability in processing (cooking) time
• Reduces total volume of food needed (pooling)
• Reduces food waste
• Reduces food storage space (rent, depreciation)
• Reduces food cost (volume purchasing)
• Increases quality (less to go wrong in
preparation, etc.)
Characterising and
Measuring Variability

Bilal Gokpinar
What Is a Good Measure of Variability?

Question: Which one of these processes

has a higher variability?
Average Process Standard Deviation of Coefficient of
Time (minutes) Process Time (minutes) Variation (c)
Process 1 20 15 c = 15/20 = 0.75
Process 2 20 10 c = 10/20 = 0.5
Process 3 10 10 c = 10/10 = 1

Standard Deviation
Coefficien t of Variation (c) 
Mean
 Characterising Arrival Variation
• CVa = 1 when arrivals are independent
• Customers walking into a store
• CVa < 1 when there is a schedule
• Example: doctor’s appointments
• CVa > 1 when arrivals occur in batches
• Teenage girls at a mall
• You do not need to calculate the coefficient of
variation (CV = standard deviation/mean)
exactly; an estimate is mostly sufficient; just
think if the process is more or less random than
“random”
Characterising Service Variation
• CVs < 1 when there is uniformity in processing a
job
• Repetitive processing, Jiffy-Lube, simple wills in legal
services
• CVs  1 when the job is customer-specific
• Car dealer service centre, surgery for particular
ailment, legal specialty such as family law
• CVs > 1 when jobs are very different in nature,
each job is customer specific
• Custom designs, general automotive repair shop,
general surgery, general practice of law
Variability Management Approaches

Queueing Theory

• Unpredictable Variability
• Utilization < 1
• Long Run Analysis

• stochastic analysis with inter-arrival and service time

distributions

All other cases Simulation / Experiments

Page
Queueing

 
Arriving  Departing
customers  customers

Customers Multiple
waiting for servers
service
▪ Demand side ▪ Supply side

The arrival rate (): the number Service capacity (): it is the
of jobs that arrive on average per maximum number of jobs that one
unit of time server can process per unit time
The inter-arrival time (a= 1/ ):
this is simply the average time Service time (=1/): it is the time
between two consecutive arrivals that on average one server needs
to process one job

Page
Queueing – some data and analysis

▪ Suppose:
— a = 20 seconds between customer
arrivals
— Arrival (input) rate () = (60/20) = 3 customers per minute

—  = 60 seconds per customer

— Capacity of each server (μ) = 1 customers per minute
— s = number of servers = 4
— Capacity of the system= 4 x 1= 4 customers per min

— Capacity Utilization = Input rate / Capacity

= 3 /4 = 75 %

Page
The implications of utilization

▪ A 75% utilization means:

— At any given moment, there is a 75% chance a
server is busy serving a customer and a 1-0.75
= 25% chance the server is idle.
— At any given moment, on average 75% of the
servers are busy serving customers and 1 –
0.75 = 25% are idle.

If utilization < 100% then:

 The system is stable which means that the size of the queue does not keep
growing over time. (Capacity is sufficient to serve all demand, despite wait)
If utilization >= 100% then:
 The system is unstable, which means that the size of the queue will
continue to grow over time.

Page
 The Operations Management Triangle

Waiting time
(or inventory)
High Inventory (b)
(or long lead time)

(a) (c)

0 0.2 0.4 0.6 0.8 1

Utilization rate

Waiting = (service time) x (utilization) x (variability)

Page
What determines time in queue in a stable
system? The capacity factor:

Average processing time of

the system = ( /s).
Demand does not influence
this factor.

 Service time   Utilization

2(s 1) 1 
  CVa2  CVs2 
Time in queue      
   
s 1 - Utilizatio n   2 

The variability factor:

The utilization factor:
This is how variability influences
Average demand influences this
time in queue – the more
factor because utilization is the
variability the more time in queue.
ratio of demand to capacity.

Page
How many customers are in the system?

▪ Remember Little’s Law: The number of

customers/jobs in a stationary system is equal
to the average arrival rate multiplied by the
average time that a customer/job spends in
the system.
▪ Using the same logic where:
: arrival rate to the system
Wq= time spent in the queue
W: total time spent in the system

Number of customers/jobs in the queue (Lq): it is

the average queue length (or average number of Lq =  x Wq
units backlogged)

Total number of customers/jobs in the system L =xW

(L); it is the queue length plus the number of jobs
currently being processed

Page
Utilization and system performance

 Service time   Utilization

2(s 1) 1 

 CV 2
 CV 2

Time in queue    
a s

 s   1 - Utilization   2 
200

180

160 Service time = 1

▪ Time in queue 140
s=1

increases Time in queue

120
CVa = 1
CVp = 1
dramatically as
100
the utilization
80
approaches
60
100%
40

20

0
80% 82% 84% 86% 88% 90% 92% 94% 96% 98% 100%
Utilization

Page
 Question

• Elite’s is a popular hair salon in Canary Wharf, which offers high-quality hair-styling and
physical relaxation services at a reasonable price. Customers arrive on average every 25
minutes. They are served by one of 3 employees, and the total service takes on average 60
minutes. The coefficient of variation of the time between arrivals as well as that of the
processing time is 1.2.

• How long do customers wait on average (if none of them leaves before being serviced),
and how many customers will on average be waiting to be serviced?
  Service time   Utilization
2(s 1) 1 
 CVa2  CVs2 
Question Time in queue  
 s

  1 - Utilization  
  
2



• Elite’s is a popular hair salon in Canary Wharf, which offers high-quality hair-styling and
physical relaxation services at a reasonable price. Customers arrive on average every 25
minutes. They are served by one of 3 employees, and the total service takes on average 60
minutes. The coefficient of variation of the time between arrivals as well as that of the
processing time is 1.2.

• How long do customers wait on average (if none of them leaves before being serviced),
and how many customers will on average be waiting to be serviced?

Input rate= every 25 mins a customer arrives input rate: 60/25 customer/hr = 2.4 customer/hr
Capacity= 1 server: 1 customer/hr 3 servers: 3 customer/hr
Hence the utilization rate is 2.4/3=80%.
The average waiting time Wq= ( 60/3) * ( 0.8^(Sqrt(2*(3+1))-1)/(1-0.8) )* (1.2^2+1.2^2)/2= 95.8 minutes=
1.6 hours

Using Little’s law, the average number of customers waiting is: WIP=TT*TR = 1.6hrs * 2.4 customers/hr=
3.83.
 Question

• Elite’s is a popular hair salon in Canary Wharf, which offers high-quality hair-styling and
physical relaxation services at a reasonable price. Customers arrive on average every 25
minutes. They are served by one of 3 employees, and the total service takes on average 60
minutes. The coefficient of variation of the time between arrivals as well as that of the
processing time is 1.2.

- How many additional servers would you need to hire so that customers wait on average no
longer than 10 minutes?
 Question

• Elite’s is a popular hair salon in Canary Wharf, which offers high-quality hair-styling and
physical relaxation services at a reasonable price. Customers arrive on average every 25
minutes. They are served by one of 3 employees, and the total service takes on average 60
minutes. The coefficient of variation of the time between arrivals as well as that of the
processing time is 1.2.

- How many additional servers would you need to hire so that customers wait on average no
longer than 10 minutes?

 Service time   Utilization

2(s 1) 1 
  CVa2  CVs2 
Time in queue      
 s   1 - Utilization   2 

If you hire one additional server (s=4) the waiting time is 17.9 minutes, if you hire two additional
servers (s=5) the waiting time is 5.45 minutes. Thus, we need two additional servers, if we want
a waiting time of less than 10 minutes.
How to reduce Waiting?
Levers to reduce waiting

Waiting = (service time) x (utilization) x (variability)

1. Reduce variability 2. Decrease utilization

3. Resource pooling 4. Manage customer’s perceptions of

waiting

Page 44
Levers to reduce waiting

Waiting = (service time) x (utilization) x (variability)

1. Reduce variability 2. Decrease utilization

Decrease arrival variability
Reliable suppliers
Better forecasts
Reservations
Appointments
Decrease service variability
Standardize operating procedures

3. Resource pooling 4. Manage customer’s perceptions of

waiting

Page 45
Levers to reduce waiting

Waiting = (service time) x (utilization) x (variability)

1. Reduce variability 2. Decrease utilization

Decrease arrival variability Increase capacity (servers)
Reliable suppliers Add servers
Better forecasts Increase capacity (speed of servers)
Reservations Simplify
Appointments Information technology
Decrease service variability Other technology
Standardize operating procedures Customer Participation
Decrease demand

3. Resource pooling 4. Manage customer’s perceptions of

waiting

Page 46
2. When is adding capacity most useful?
Decreasing utilization is more effective when
utilization is high! Search for the
bottleneck!
Wq , days

High returns

Low returns

Utilization
2: low utilization 1: high utilization
utilization, r
Page
3. Resource Pooling

The Line Dance

Queuing theory, the mathematical study of lines, helps businesses, call centres, computer networks and
others figure out how to keep things moving.

Multiple servers, multiple lines Multiple servers, single line

Page
THE WALL STREET JOURNAL
Resource Pooling (cont.)

The Line Dance

Queuing theory, the mathematical study of lines, helps businesses, call centres, computer networks and
others figure out how to keep things moving.

Multiple servers, multiple lines Multiple servers, single line

If there are delays If delays are random,

at one register, it delays will only affect
affects the entire a single register
line. This causes rather than the entire
the rate of service line.
to vary in different
lines

Page
THE WALL STREET JOURNAL
4. Managing the Psychology of Queueing

1. Unoccupied time feels longer than occupied time

2. Pre-process waits feel longer than in-process waits
3. Anxiety makes waits seem longer
4. Uncertain waits seem longer than known, finite waits
5. Unexplained waits are longer than explained ones
6. Unfair waits are longer than equitable waits
7. The more valuable the service, the longer the customer will wait
8. Solo waits feel longer than group waits

Page 50