Business Analytics

By
Nikunj Kumar Jain
Associate Professor – POM, IIM Nagpur
Fellow – IIM Indore, B.Tech - NIT, Bhopal
Room No. 19, Ground Floor; Mob. 9827440301
Email: [email protected]
Session Plan

 Decision Making
 Certainty
 Uncertainty
 Risk
 Pay-off Matrix
 EMV and EOL Criteria
 Value of Information
 Value of Perfect Information
 Value of Imperfect Information
Decision-making Environments

 Certainty
 Uncertainty
 Risks
Five Steps

1. Clearly define the problem

2. List all possible alternatives
3. Identify all possible outcomes for each alternative
4. Identify the payoff for each alternative and outcome combination
5. Use a decision modeling technique to choose an alternative
Thompson Lumber Company

 John Thompson is the founder and president of Thompson Lumber

Company, a profitable firm located in Portland, Oregon. The problem
that John Thompson identifies is whether to expand his product line by
manufacturing and marketing a new product, backyard storage sheds.
John decides that his alternatives are to construct:
 A large new plant to manufacture the storage sheds, or
 A small plant
John determines that all his decision alternatives have the same three
possible outcomes:
 Demand for the sheds will be high
 Demand for the sheds will be moderate
 Demand for the sheds will be low
Thompson Lumber

1. Decision: Should he make and sell storage sheds

2. Alternatives:
1. Build a large plant
2. Build a small plant
3. Do nothing
3. Outcomes: Demand for sheds will be
1. High
2. Moderate
3. Low
Payoffs

 John wants to use net profits to measure his payoffs. He has already
evaluated the potential profits associated with various combinations
of alternatives and outcomes, as follows:
 If John decides to build a large plant, he thinks that with high demand
for sheds, the result would be a net profit of $200,000 to his firm. The
net profit would, however, be only $100,000 if the demand were
moderate. If the demand were low, there would actually a net loss of
$120,000.
 If he builds a small plant, the results would be a net profit of $90,000
if there were high demand for sheds, a net profit of $50,000 if there
were moderate demand, and a net loss of $20,000 if there were low
demand.
Thompson Lumber

4. Payoff table

OUTCOMES
HIGH MODERATE LOW
ALTERNATIVES DEMAND DEMAND DEMAND
Build large plant $200,000 $100,000 –$120,000
Build small plant $ 90,000 $ 50,000 –$ 20,000
No plant $ 0 $ 0 $ 0

5. Select and apply decision analysis model

Decision-Making Environments

Type 1: Decision making under certainty

Type 2: Decision making under uncertainty
Type 3: Decision making under risk
??? (Decision Making Under
Certainty)
 If John knows with certainty that the demand for sheds will be high,
what will he do? Will his decision change if he knows with certainty
that the demand for sheds will be moderate?
Certainty

 Consequence of every alternative is known

 Usually only one outcome for each alternative
 Seldom occurs in reality
??? (Decision Making Under
Uncertainty)
 What will John do if he can’t predict demand for storage sheds?
Uncertainty

 Probabilities of possible outcomes not known

 Decision making methods:
1. Maximax

2. Maximin

3. Criterion of realism

4. Equally likely

5. Minimax regret
Thompson Lumber
 Maximax Criterion
 Maximizes the maximum payoff

OUTCOMES
HIGH MODERATE LOW
ALTERNATIVES DEMAND DEMAND DEMAND
Build large plant $200,000 $100,000 –$120,000
Build small plant $ 90,000 $ 50,000 –$ 20,000
No plant $ 0 $ 0 $ 0

Maximax
Thompson Lumber
 Maximin Criterion
 Maximizes the minimum payoff

OUTCOMES
HIGH MODERATE LOW
ALTERNATIVES DEMAND DEMAND DEMAND
Build large plant $200,000 $100,000 –$120,000
Build small plant $ 90,000 $ 50,000 –$ 20,000
No plant $ 0 $ 0 $ 0

Maximin
Thompson Lumber
 Criterion of Realism (Hurwicz)

Thompson’s coefficient of realism a = 0.45

Realism payoff for
= a x (Maximum payoff for alternative)
alternative
+ (1 – a) x (Minimum payoff for alternative)

OUTCOMES
HIGH MODERATE LOW WT. AVG.
FOR
ALTERNATIVES DEMAND DEMAND DEMAND
ALTERNATIVE

Build large plant $200,000 $100,000 –$120,000 $24,000

Build small plant $ 90,000 $ 50,000 –$ 20,000 $29,500
No plant $ 0 $ 0 $ 0 $ 0
Realism
Thompson Lumber
 Equally Likely (Laplace) Criterion
 Highest average payoff

OUTCOMES
HIGH MODERATE LOW AVERAGE FOR
ALTERNATIVES DEMAND DEMAND DEMAND ALTERNATIVE
Build large plant $200,000 $100,000 –$120,000 $60,000
Build small plant $ 90,000 $ 50,000 –$ 20,000 $40,000
No plant $ 0 $ 0 $ 0 $ 0

Equally
likely
Minmax Regret

 Minimax regret finds the alternative that minimizes the maximum

opportunity loss within each alternative.
 The final decision criterion that we discuss is based on opportunity
loss, also called regret.
 Opportunity loss is defined as the difference between the optimal
payoff and the actual payoff received.
 In other words, it’s the amount lost by not picking the best
alternative.
Thompson Lumber
 Minimax Regret Criterion

OUTCOMES
ALTERNATIVES HIGH DEMAND
Build large plant $200,000 – $200,000 = $ 0
Build small plant $200,000 – $ 90,000 = $110,000
No plant $200,000 – $ 0 = $200,000
MODERATE DEMAND
Build large plant $100,000 – $100,000 = $ 0
Build small plant $100,000 – $ 50,000 = $ 50,000
No plant $100,000 – $ 0 = $100,000
LOW DEMAND
Build large plant $0 – (–$120,000) = $120,000
Build small plant $0 – (–$ 20,000) = $ 20,000
No plant $0 – $ 0 =$ 0
Thompson Lumber
• Minimax Regret Criterion

OUTCOMES
HIGH MODERATE LOW MAXIMUM FOR
ALTERNATIVES DEMAND DEMAND DEMAND ALTERNATIVE
Build large plant $ 0 $ 0 $120,000 $120,000
Build small plant $110,000 $ 50,000 $ 20,000 $110,000
No plant $200,000 $100,000 $ 0 $200,000

Minimax
??? (Decision Making Under Risks)

 John decides to consult his father, who has started the business and
he has good knowledge of the storage shed industry. John’s father
suggest that the probabilities of high, moderate and low demand
would be 0.3, 0.5 and 0.2 respectively. With these risky estimates,
determine the alternative that would give John the greatest expected
monetary value (EMV)?
 Refer above problem, determine the best alternative for John using
minimizing expected opportunity loss (EOL) criteria.
 Under Risk

 Expected Monetary Value (EMV)

EMV (Alternative i) = (Payoff of first outcome)

x (Probability of first outcome)
+ (Payoff of second outcome)
x (Probability of second outcome)
+ … + (Payoff of last outcome)
x (Probability of last outcome)
Thompson Lumber
OUTCOMES
HIGH MODERATE LOW EMV FOR
ALTERNATIVES DEMAND DEMAND DEMAND ALTERNATIVE
Build large plant $200,000 $100,000 –$120,000 $200,000 x 0.3
+ $100,000 x 0.5
+ (–$120,000) x 0.2
= $86,000
Build small plant $ 90,000 $ 50,000 –$ 20,000 $90,000 x 0.3
+ $50,000 x 0.5
+ (–$20,000) x 0.2
= $48,000
No plant $ 0 $ 0 $ 0 $0 x 0.3 + $0 x 0.5
+ $0 x 0.2 = $ 0
Probabilities 0.3 0.5 0.2
 Under Risk

 Expected Opportunity Loss (EOL)

EOL (Alternative i) = (Regret of first outcome)

x (Probability of first outcome)
+ (Regret of second outcome)
x (Probability of second outcome)
+ … + (Regret of last outcome)
x (Probability of last outcome)
Thompson Lumber
OUTCOMES
HIGH MODERATE LOW EOL FOR
ALTERNATIVES DEMAND DEMAND DEMAND ALTERNATIVE
Build large plant $ 0 $ 0 $120,000 $0 x 0.3 + $0 x 0.5
+ $120,000 x 0.2
= $24,000
Build small plant $110,000 $ 50,000 $ 20,000 $110,000 x 0.3
+ $50,000 x 0.5
+ $20,000 x 0.2
= $62,000
No plant $200,000 $100,000 $ 0 $200,000 x 0.3
+ $100,000 x 0.5
+ $0 x 0.2 =
$110,000
Probabilities 0.3 0.5 0.2
Decision Trees

 Presents decision alternatives and outcomes in a sequential manner

Decision node
Outcome node
Payoff node
Decision Trees
 Thompson Lumber

Payoffs
High Demand
$200,000
Outcome Node
Moderate Demand
1 $100,000
t Low Demand
an
Decision Node Pl –$120,000
a rge
L High Demand
$90,000
Small Plant 2 Moderate Demand
$50,000
No Low Demand
P l an –$20,000
t
All Demands
3 $0
Decision Trees
 Thompson Lumber
Probability

Payoffs
High Demand (0.30)
$200,000
Outcome Node $86,000
Moderate Demand (0.50)
1 $100,000
l a nt Low Demand (0.20)
Decision Node eP –$120,000
rg
La High Demand (0.30)
$90,000
$48,000
Small Plant 2 Moderate Demand (0.50)
$50,000
No Low Demand (0.20)
P l an –$20,000
t
$0
All Demands
3 $0

EMV 1 = $200,000 x 0.3 + $100,000 x 0.05 + (–$120,000) x 0.2 = $86,000

EMV 2 = $90,000 x 0.3 + $50,000 x 0.05 + (–$20,000) x 0.2 = $48,000

Decision Trees
 Thompson Lumber

1 EMV = $86,000
t
an
Decision Node e Pl
rg
La

Small Plant 2 EMV = $48,000

No
P l an
t
$0
3 EMV = $0
??? (Value of Information)

 John Thompson has been approached by Scientific Marketing, Inc., a

market research firm, with a proposal to help him make the right
decision regarding the size of the new plant. Scientific claims that its
analysis will tell John with certainty whether the demand for storage
sheds will be high, moderate, or low. In other words, it will change John’s
problem environment from one of decision making under risk to one of
decision making under certainty. Obviously, this information could
prevent John from making an expensive mistake.
 Scientific would charge $30,000 for the information.
 What should John do?
 Should he hire Scientific to do the marketing study? Is the information
worth $30,000?
 If not, what is it worth?
Perfect Information

 Expected Value with Perfect Information (EVwPI)

EVwPI = (Best payoff of first outcome)

x (Probability of first outcome)
+ (Best payoff of second outcome)
x (Probability of second outcome)
+ … + (Best payoff of last outcome)
x (Probability of last outcome)

EVwPI is the long run average re turn if we have perfect

information before a decision is made
 Perfect Information
 Expected Value of Perfect Information (EVPI)
EVPI = EVwPI – Maximum EMV
EVPI = Minimum EOL
Perfect Information
Expected Value with Perfect Information
EVwPI = $200,000 x 0.3 + $100,000 x 0.5 + $0 x 0.20
= $110,000

Expected Value of Perfect Information

EVPI = EVwPI – Maximum EMV

= $110,000 – $86,000 = $24,000
Exercise

 Airline industry is highly competitive and price of air tickets are

volatile in nature. Table 1 shows the payoffs (in $ million) to the Air
India airline over a period of time, for a given fare set by the AI and by
its competitor.
Competitor’s Fare (State of Nature)
AI’s fare (Action)
$200 $300
$200 8 9
$300 4 10

 Assume that the probability of competitor choosing low price is 0.6 and
probability of high price is 0.4. AI is looking forward to hire a
consultant to obtain information about the operating philosophy of the
competitor airline. How much should AI pay for this information?
Airline Industry: The Decision Tree

Airline Competitor’s Payoff

Fare Fare
$8 million
Competitor:$200
$200 Pr = 0.6
Fare
8.4 Competitor:$300
Pr = 0.4 $9 million

Competitor:$200 $4 million
$300 Pr = 0.6
Fare
6.4
Competitor:$300
Pr = 0.4
$10 million
Airline Industry: Value of Additional
Information
• If no additional information is available, the best strategy is to set the fare at
$200
 E(Payoff|200) = (.6)(8)+(.4)(9) = $8.4 million
 E(Payoff|300) = (.6)(4)+(.4)(10) = $6.4 million
• With further information, the expected payoff could be:
 E(Payoff|Information) = (.6)(8)+(.4)(10)=$8.8 million
• EVPI=8.8-8.4 = $.4 million
New Product Introduction

 ABC is thinking to launch a new product. After doing some research, the management
came to the conclusion there is a 75 percent probability of product being successful and
the profits will be $ 100,000 and there is 25 percent probability that the product may
not be success incurring a loss of $ 20,000.
 What should ABC do?
Decision Tree: New-Product Introduction

Chance
Chance Final
Final
Decision
Decision Occurrence
Occurrence Outcome
Outcome
Product
successful $100,000
(P = 0.75)
Market
Product -$20,000
unsuccessful
Do not (P = 0.25)
market
$0
Payoff Table and Expected Values of Decisions:
New-Product Introduction
Productisis
Product
Action
Action Successful Not
Successful NotSuccessful
Successful
Marketthe
Market theproduct
product $100,000
$100,000 -$20,000
-$20,000
Donot
Do notmarket
marketthe
theproduct
product $0
$0 $0
$0

The expected value of X , denoted E ( X ):

E ( X )   xP ( x )
all x
E ( Outcome)  (100,000)( 0.75)  (  20,000)( 0.25)
= 750000 -5000 = 70,000
Solution to the New-Product Introduction
Decision Tree
Clipping the Nonoptimal Decision Branches
Product
Expected
Expected successful
Payoff
Payoff (P = 0.75) $100,000
$70,000
$70,000
Market

Product -$20,000
unsuccessful
(P = 0.25)
Nonoptimal Do not
Nonoptimal
decisionbranch
branch market Expected
decision Expected $0
isisclipped
clipped Payoff
Payoff
$0
$0
New Product Introduction

 The management decided to consider the extended possibilities regarding the success of
the new product and came with the following outcomes with their probabilities. What is
the expected payoff of ABC?

Outcome Payoff Probability

Extremely successful 150000 0.1
Very successful 120000 0.2
Successful 100000 0.3
Somewhat successful 80000 0.1
Barely successful 40000 0.1
Breakeven 0 0.1
Unsuccessful -20000 0.05
Disastrous 50000 0.05
New-Product Introduction:
Extended-Possibilities
Outcome
Outcome Payoff Probability
Payoff Probability xP(x)
xP(x)
Extremelysuccessful
Extremely successful $150,000
$150,000 0.1
0.1 15,000
15,000
Verysuccessful
Very successful 120.000
120.000 0.2
0.2 24,000
24,000
Successful
Successful 100,000
100,000 0.3
0.3 30,000
30,000
Somewhatsuccessful
Somewhat successful 80,000
80,000 0.1
0.1 8,000
8,000
Barelysuccessful
Barely successful 40,000
40,000 0.1
0.1 4,000
4,000
Breakeven
Break even 00 0.1
0.1 00
Unsuccessful
Unsuccessful -20,000
-20,000 0.05
0.05 -1000
-1000
Disastrous
Disastrous -50,000
-50,000 0.05
0.05 -2,500
-2,500

ExpectedPayoff:
Expected Payoff: $77,500
$77,500
New-Product Introduction:
Extended-Possibilities Decision Tree
Chance
Chance Payoff
Decision
Decision Occurrence
Occurrence
0.1
$150,000
Expected
Expected 0.2 $120,000
Payoff
Payoff 0.3
$77,500 $100,000
$77,500 0.1
$80,000
0.1 $40,000
Market
0.1 $0
0.05
-$20,000
0.05
-$50,000
Do not
Nonoptimal
Nonoptimal market
decisionbranch
decision branch $0
isisclipped
clipped
Exercise 1

 Suppose you can earn a little bit of spending money selling falafel
sandwiches in a busy square on Saturday afternoons. On a normal
Saturday, a day’s worth of effort nets you $150. Unfortunately, there
is a 30 percent chance of rain. If it rains, you will only make $50 (a
few desperate souls will still buy soggy falafel sandwiches). Suppose
you can rent an extra large umbrella for $25/day. In the event it rains,
your falafel sandwiches will not be soggy, and you can expect to
make $100 (exclusive of umbrella costs).
 Should you rent the umbrella? You must decide to rent before you find
out whether it rains. Furthermore, you must pay for the umbrella
whether or not it rains.
 What is the most you would pay for a perfectly accurate weather
forecasts?
Exercise 2

 ONGC obtained the rights to drill the KG basin reserves. If ONGC

drills, there is a 20% chance that it will strike oil, in which case the
payoff is $200,000. The cost of drilling is $20,000.
 a) Should ONGC drill?
 b) What is the most you would pay for the services of a geologist who
can assess the oil prospects with perfect accuracy?
Exercise 3

 ABC company is under contract to make 10,000 blenders. It must

decide whether to make or buy the blender motor. Unfortunately, ABC
does not presently have motor technology. Motor development is a
two-stage process. The first stage has an 80% chance of success and
will cost $30,000. The second stage has a 60% chance of success and
will cost $20,000. If development is successful, then the variable cost
of producing a motor will be $2.50/motor. If either development stage
is unsuccessful, ABC must purchase motors outside at a cost of
$10/motor.
 a) Should ABC purchase the motor outside or begin development?
 b) What is the most ABC would pay to know, before making the first
decision, whether the first stage would succeed? the second stage?
Investment Decision

An individual has $1 million dollars and wishes to make a one-year investment. Suppose his/her
possible actions are:
 a1: buy a guaranteed income certificate paying 10%
 a2: buy bond with a coupon value of 8%
 a3: buy a well-diversified portfolio of stocks
Return on investment in the diversified portfolio depends on the behavior of the interest rate next
year. Suppose there are three possible states of nature:
s1: interest rate increases
s2: interest rate stays the same
s3: interest rate decreases
Suppose further that the subjective probabilities for these states are 0.2, 0.5, and 0.3, respectively.
Pay-off table

Based on historical data, the payoff table is:

State of Nature Action

A1 A2 A3

S1 100 -50 150

S2 100 80 90
S3 100 180 40

Which action should he/she take?

Pay-off table (Expected Value approach)

State of Nature Action

A1 A2 A3

S1 100 -50 150

S2 100 80 90
S3 100 180 40
Expected Pay-off 100 84 87

Hence, if one wishes to maximize expected payoff, then action A1 should be

taken.
Pay-off table (Expected Opportunity Loss
approach)
An equivalent concept is to minimize expected opportunity loss (EOL). Consider any given
state. For each possible action, the opportunity loss is defined as the difference between
what the payoff could have been had the best action been taken and the payoff for that
particular action. Thus,
Pay-off table (Expected Opportunity Loss
approach)
State of Nature Action
A1 A2 A3

S1 50 200 0
S2 0 20 10
S3 80 0 140
EOL 34 50 47

Again A1 is optimal
The Law of Total Probability and Bayes’
Theorem
The law of total probability:
P( A) P( A  B)  P( A  B )

In terms of conditional probabilities:

P( A) P( A  B)  P( A  B )
P ( A B ) P ( B )  P ( A B ) P ( B )

More generally (where Bi make up a partition):

P( A)  P( A  B )
i
 P( AB ) P( B )
i i
Example

 An analyst believes the stock market has a 0.75 probability of going up in the next year
if the economy should do well, and a 0.3 probability of going up if the economy should
not do well during the year. The analyst further believes there is a 0.8 probability that
the economy will do well in the coming year. What is the probability that the stock
market will go up next year (using the analyst’s assessment)?
The Law of Total Probability-

Event U: Stock market will go up in the next year

Event W: Economy will do well in the next year
P(U W ) .75
P(U W ) 30
P(W ) .80  P(W ) 1 .8 .2

P(U ) P(U W )  P(U W )

P(U W ) P(W )  P(U W ) P(W )
(.75)(.80)  (.30)(.20)
.60.06 .66
Bayes’ Theorem
• Bayes’ theorem enables you, knowing just a little more than the
probability of A given B, to find the probability of B given A.
• Based on the definition of conditional probability and the law of total
probability.

P ( A B)
P ( B A) 
P ( A)
P ( A  B) Applying the law of total
 probability to the denominator
P ( A B)  P ( A  B )
P ( A B) P ( B)
 Applying the definition of
P ( A B) P ( B)  P ( A B ) P ( B ) conditional probability throughout
Bayes’ Theorem Extended

• Given a partition of events B1,B2 ,...,Bn:

P( A  B )
P ( B A)  1

P ( A)
1

Applying the law of total

P( A  B ) probability to the denominator
 1

 P( A  B ) i
Applying the definition of
P( A B ) P( B ) conditional probability throughout
 1 1

 P( A B ) P( B )
i i
Bayes’ Theorem Extended -
Example

An economist believes that during periods of high economic growth, the U.S.
dollar appreciates with probability 0.70; in periods of moderate economic
growth, the dollar appreciates with probability 0.40; and during periods of
low economic growth, the dollar appreciates with probability 0.20.

During any period of time, the probability of high economic growth is 0.30,
the probability of moderate economic growth is 0.50, and the probability of
low economic growth is 0.20.

Suppose the dollar has been appreciating during the present period. What is
the probability we are experiencing a period of high economic growth?
Partition: Event A  Appreciation
H - High growth P(H) = 0.30 P ( A H )  0.70
M - Moderate growth P(M) = 0.50 P ( A M )  0.40
L - Low growth P(L) = 0.20 P ( A L)  0.20
Example (Tree Diagram)
Prior Conditional Joint
Probabilitie Probabilities Probabilities
s P ( A  H )  ( 0.30)( 0.70)  0.21
P ( A H )  0.70

P ( A H )  0.30
P ( H )  0.30 P ( A  H )  ( 0.30)( 0.30)  0.09

P ( A M )  0.40 P ( A  M )  ( 0.50)( 0.40)  0.20

P ( M )  0.50

P ( A M )  0.60 P ( A  M )  ( 0.50)( 0.60)  0.30

P ( A L )  0.20
P ( L )  0.20 P ( A  L )  ( 0.20)( 0.20)  0.04

P ( A L )  0.80 P ( A  L )  ( 0.20)( 0.80)  0.16

Example (Tree Diagram)
Posterior
Prior Conditional Joint
Probabilities
Probabilitie Probabilities Probabilities
s P ( A  H )  ( 0.30)( 0.70)  0.21 P( H A) 
0.21

0.21
0.467
P ( A H )  0.70 0.21 0.20  0.04 0.45

P ( A H )  0.30
P ( H )  0.30 P ( A  H )  ( 0.30)( 0.30)  0.09

P ( A M )  0.40 P ( A  M )  ( 0.50)( 0.40)  0.20

P ( M )  0.50

P ( A M )  0.60 P ( A  M )  ( 0.50)( 0.60)  0.30

P ( A L )  0.20
P ( L )  0.20 P ( A  L )  ( 0.20)( 0.20)  0.04

P ( A L )  0.80 P ( A  L )  ( 0.20)( 0.80)  0.16

Example (continued)

P( H  A)
P( H A) 
P( A)
P( H  A)

P( H  A)  P( M  A)  P( L  A)
P( A H ) P( H )

P ( A H ) P ( H )  P ( A M ) P ( M )  P ( A L) P ( L)
( 0.70)( 0.30)

( 0.70)( 0.30)  ( 0.40)( 0.50)  ( 0.20)( 0.20)
0.21 0.21
 
0.21 0.20  0.04 0.45
 0.467
Bayesian Analysis

• Probability values can be revised based on new information

• Prior probabilities
• Revised probabilities

• Prior Probabilities

P(HD) = 0.30, P(MD) = 0.50, P(LD) = 0.30

 Before deciding about building a new plant, let’s suppose John Thompson has
been approached by Smart Services, another market research firm. Smart will
charge John $4,000 to conduct a market survey. The results of the survey will
indicate either positive or negative market conditions for storage sheds. What
should John do? (Assume positive or negative market with 100% accuracy)
 John determines the probability of 0.57 that the survey will indicate positive
market conditions for storage sheds, and probability of 0.43 that the survey
will indicate negative market conditions. In order to evaluate the reliability of
the survey, John asks Smart Services to provide him with information regarding
its performance on past surveys. Specifically, he wants to know how many
similar surveys the company has conducted in the past, what it predicted each
time, and what the actual result turned out to be eventually in each case.
Smart has data on 75 past surveys that it has conducted. In these 75 surveys,
Smart had predicted high demand in 30 cases, moderate demand in 15 cases,
and low demand in 30 cases. Table 1 shows this information. Table 1 reveals,
for example, that in 29 of 30 past cases where a product’s demand
subsequently turned out to be high, Smart’s surveys had predicted positive
market conditions. That is, the probability of positive survey results, given high
demand, P (PS/ HD), is 0.967. Likewise, in 7 of 15 past cases where a product’s
demand subsequently turned out to be moderate, Smart’s surveys had
predicted negative market conditions. That is, the probability of negative
survey results, given moderate demand, P (NS/MD), is 0.467. How does John
use this information to gauge the accuracy of Smart’s survey in his specific
case? Determine whether John should use Smart’s offer? What should John do
Table 1: Reliability of Smart Services
Survey in Predicting Actual
Outcomes
SURVEY RESULT
WAS
WHEN ACTUAL
OUTCOME WAS POSITIVE (PS)
NEGATIVE (NS)
High (HD) P(PS | HD) = 29/30 = 0.967
P(NS | HD) = 1/30 = 0.333
Moderate (MD) P(PS | MD) = 8/15 = 0.533
P(NS | MD) = 7/15 = 0.467
Low (LD) P(PS | LD) = 2/30 = 0.067
P(NS | LD) = 28/30 = 0.933
Revised Probabilities
Revised Probabilities
Revised Probabilities
 Survey reliability

SURVEY RESULT WAS

WHEN ACTUAL
OUTCOME WAS POSITIVE (PS) NEGATIVE (NS)

High (HD) P(PS | HD) = 29/30 = 0.967 P(NS | HD) = 1/30 = 0.333
Moderate (MD) P(PS | MD) = 8/15 = 0.533 P(NS | MD) = 7/15 = 0.467
Low (LD) P(PS | LD) = 2/30 = 0.067 P(NS | LD) = 28/30 = 0.933
Revised Probabilities
 Positive survey results

CONDITIONAL REVISED
PROB. PRIOR JOINT PROB.
OUTCOME P(PS | OUTCOME) PROB. PROB. P(OUTCOME | PS)
High (HD) 0.967 X 0.30 = 0.290 0.290/0.57 = 0.509
Mod. (MD) 0.533 X 0.50 = 0.267 0.267/0.57 = 0.468
Low (LD) 0.067 X 0.20 = 0.013 0.013/0.57 = 0.023
P(PS) = P(Positive Survey) = 0.570 1.000
Revised Probabilities

 Negative survey results

CONDITIONAL REVISED
PROB. PRIOR JOINT PROB.
OUTCOME P(NS | OUTCOME) PROB. PROB. P(OUTCOME | NS)
High (HD) 0.033 X 0.30 = 0.010 0.010/0.43 = 0.023
Mod. (MD) 0.467 X 0.50 = 0.233 0.233/0.43 = 0.543
Low (LD) 0.933 X 0.20 = 0.187 0.187/0.43 = 0.434
P(NS) = P(Negative Survey) = 0.430 1.000
Revised Probabilities
CONDITIONAL REVISED
PROB. PRIOR JOINT PROB.
OUTCOME P(NS | OUTCOME) PROB. PROB. P(OUTCOME | NS)
High (HD) 0.033 X 0.30 = 0.010 0.010/0.43 = 0.023
Mod. (MD) 0.467 X 0.50 = 0.233 0.233/0.43 = 0.543
Low (LD) 0.933 X 0.20 = 0.187 0.187/0.43 = 0.434
P(NS) = P(Negative Survey) = 0.430 1.000

CONDITIONAL REVISED
PROB. PRIOR JOINT PROB.
OUTCOME P(NS | OUTCOME) PROB. PROB. P(OUTCOME | NS)
High (HD) 0.033 X 0.30 = 0.010 0.010/0.43 = 0.023
Mod. (MD) 0.467 X 0.50 = 0.233 0.233/0.43 = 0.543
Low (LD) 0.933 X 0.20 = 0.187 0.187/0.43 = 0.434
P(NS) = P(Negative Survey) = 0.430 1.000
 Second Decision Payoffs
Point

Multistage Decision
High Demand (0.3)
$200,000
Moderate Demand (0.5)
1 $100,000
t
lan Low Demand (0.2)

Trees La
rg
e P

Small Plant
High Demand (0.30)
Moderate Demand (0.50)
$120,000

$90,000
2 $50,000
No Low Demand (0.20)
Pla $20,000
nt
All Demands
3 $0

ey
rv
First Decision High Demand (0.509)

Su
Point $196,000

o
N
Moderate Demand (0.486)
4 $96,000
t
Plan Low Demand (0.023)
–$124,000
e
rg
La High Demand (0.509)
$86,000
Small Plant Moderate Demand (0.486)
5 $46,000

7)
Co

.5
No Low Demand (0.023)

(0
–$24,000

nd
Pla

lts
uc
nt

su
t

Re
Su
All Demands

rv

ive
6 –$4,000

ey

sit
Po
High Demand (0.023)
$196,000
Moderate Demand (0.543)

Ne
7 $96,000
t
ga
lan Low Demand (0.434)
tive
P –$124,000
e
rg
Re
La High Demand (0.023)
su
$86,000
lts Small Plant Moderate Demand (0.543)
(0
8 $46,000
.4
3)
No Low Demand (0.434)
Pla –$24,000
nt
All Demands
9 –$4,000
Folding Back
• No market survey
EMV (node 1) = EMV (Large plant)
= $200,000 x 0.30 + $100,000 x
0.50
+ (–$120,000) x 0.20
= $86,000
EMV (node 2) = EMV (Small plant)
= $90,000 x 0.30 + $50,000 x 0.50
+ (–$20,000) x 0.20
= $48,000
EMV (node 3) = EMV (No plant)
= $0
Folding Back
• Positive market survey
EMV (node 4) = EMV (Large plant | Positive
result)
= $196,000 x 0.509 + $96,000 x 0.468
+ (–$124,000) x 0.023
= $141,840
EMV (node 5) = EMV (Small plant | Positive
result)
= $86,000 x 0.509 + $46,000 x 0.468
+ (–$24,000) x 0.023
= $64,750
EMV (node 6) = EMV (No plant | Positive
result)
= –$4,000
Folding Back
• Negative market survey
EMV (node 7) = EMV (Large plant | Negative
result)
= $196,000 x 0.023 + $96,000 x 0.543
+ (–$124,000) x 0.434
= $2,820
EMV (node 8) = EMV (Small plant | Negative
result)
= $86,000 x 0.023 + $46,000 x 0.543
+ (–$24,000) x 0.434
= $16,540
EMV (node 9) = EMV (No plant | Negative
result)
= –$4,000
Folding Back
• Node 10 analysis
EMV (node 10) = EMV (Conduct survey)
= $141,840 x 0.57 + $16540 x 0.43
= $87,961

• Decision is to conduct the survey and await

the results
• If positive, build large plant
• If negative, build small plant
 Second Decision Payoffs
Point

Multistage Decision
High Demand (0.3)
$86,000 $200,000
Moderate Demand (0.5)
1 $100,000
e Low Demand (0.2)

Trees ||
rg $120,000
La ant
Pl High Demand (0.30)
$86,000 $48,000 $90,000
Small Plant Moderate Demand (0.50)
2 $50,000
No Low Demand (0.20)
Pla $20,000
nt
$0
All Demands
3 $0

ey
rv
First Decision High Demand (0.509)

Su
Point $141,000 $196,000

o
N
Moderate Demand (0.486)
4 $96,000

||
Low Demand (0.023)
e –$124,000
rg
La ant High Demand (0.509)
$141,840 Pl $64,750 $86,000
$87,961 Small Plant Moderate Demand (0.486)
5 $46,000

7)
Co

.5
No Low Demand (0.023)

(0
–$24,000

nd
Pla

lts
uc
nt

su
–$4,000

Re
Su
All Demands

rv

ive
6 –$4,000

ey

sit
Po
||
10 High Demand (0.023)
$2,820 $196,000
$87,961 Moderate Demand (0.543)

Ne
7 $96,000
t
ga
lan Low Demand (0.434)
tive
P –$124,000
e
rg
Re
La High Demand (0.023)
su
$16,540 $86,000
lts Small Moderate Demand (0.543)
(0 Plant | | 8 $46,000
.4
3)
No Low Demand (0.434)
$16,540 Pla –$24,000
nt
–$4,000
All Demands
9 –$4,000
Expected Value
 Sample Information (EVSI)

EVSI = (EMV of best decision with sample

information, assuming no cost to get it)
– (EMV of best decision without any
information)
EVSI = $91,961 – $86,000 = $5,961

• Efficiency of Sample Information

Efficiency of sample information = EVSI / EVPI
= $5,961 / $24,000 = 0.2484 or 24.84%
Problem

 Suppose that 60% of all successful entrepreneurs have MBAs. Further,

20% of unsuccessful entrepreneurs have MBAs. Moreover, only 5% of
entrepreneurs are successful.
 What proportions of MBAs who are entrepreneurs are also successful?