MARKETING (UNR471)

Dr. Mohamed Sameh

DECISION ANALYSIS (CONTD.)
Decision making with experimentation

Fav. + Success P(Success|Fav.)

0..6*0.25=0.15 0.15/0.3 = 0.5

Renew the
Unfav. + Success P(Success|Unfav.
factory 0.4*0.25=0.1 )
Fav. + Failure 0.1/0.7 = 0.143
P(Failure|Fav.)
0.2*0.75=0.15 0.15/0.3 = 0.5

P(Fav. ) = 0.15+ 0.15 = 0.3 P(Failure|Unfav.)

Unfav. + Failure
0.8*0.75=0.6 0.6/0.7 = 0.857
P(Unfav. ) = 0.1+ 0.6 = 0.7
3
Decision making with experimentation
BAYES’ DECISION RULE
In case of a favorable survey:
Alternatives Payoff
Success Failure
Renew the Factory 70 -10
Ghazl Al-
Sell the Factory 9 9
Mehalla
Favorable survey 0.5 0. 5

• EP (Renew the Factory) = 0.5 * 70 + 0.5 *-10 - 3 = 27 (2,700,000 L.E.)

• EP (Sell the Factory) = 0.5 * 9 + 0.5 * 9 - 3 = 6 (600,000 L.E.)
• Best Decision: Renew the factory.
Decision making with experimentation
BAYES’ DECISION RULE
In case of an unfavorable survey:
Alternatives Payoff
Success Failure
Renew the Factory 70 -10
Ghazl Al-
Sell the Factory 9 9
Mehalla
Unfavorable survey 0.143 0. 857

• EP (Renew the Factory) = 0.143 * 70 + 0.857 *-10 - 3 = -1.56 (-156,000 L.E.)

• EP (Sell the Factory) = 0.143 * 9 + 0.857 * 9 - 3 = 6 (600,000 L.E.)
• Best Decision: Sell the factory.
Decision making with experimentation

VALUE OF EXPERIMENTATION
• Expected value of perfect information (EVPI):
EVPI = EP (with perfect information) – EP(without experimentation).
1. Provides a maximum limit of the potential value of the experiment.
2. If experiment cost > EVPI, No experimentation.
3. If experiment cost < EVPI, use Expected value of experimentation (EVE).
Alternatives Payoff
Success Failure
Renew the Factory 70 -10
Sell the Factory 9 9
Ghazl Al-Mehalla
Maximum Payoff 70 9
Chance 0.25 0.75

• Max. EP (with perfect information) = 0.25 * 70 + 0.75 * 9 = 24.25

• EP (without experimentation) = 0.25 * 70 + 0.75 * -10 = 10
• EVPI = 24.25 – 10 = 14.25 ( 1,425,000 L.E. > 300,000)
• EVPI > Experiment cost  use EVE
Decision making with experimentation
VALUE OF EXPERIMENTATION
• Expected value of experimentation (EVE):
EVE = EP (with experimentation) - EP(without experimentation).
1. Directly calculates the expected increase.
2. If experiment cost > EVE, No experimentation.
3. If experiment cost < EVE, Yes for experimentation.
For the example:
In case of favorable experimentation;
• P (Fav.) = 0.3, Max. EP = 0.5 * 70 + 0.5 *-10 = 30 (don’t include the experiment cost).
In case of unfavorable experimentation;
• P (Unfav.) = 0.7, Max. EP = 0.143 * 9 + 0.857 * 9 = 9 (don’t include the experiment cost).
• EP (with experimentation) = 0.3 * 30 + 0.7 * 9 = 15.3
• EP (without experimentation) = 0.25 * 70 + 0.75 * -10 = 10
• EVE = 15.3 – 10 = 5.3 ( 530,000 L.E. > 300,000)
• EVE > Experiment cost  Use experimentation.
Decision trees
8

Chance Node

A
Decision Node Terminal Node
(Outcomes or
pay-offs)

Branches B
(alternative pathways)

Time

Time (sequence of events) moves from left to right

Decision trees
9 Success(0.5) [+80]
+67
Renew
F
[- 10] Failure (0.5) [+0]
-13
Fav. E
(0.3) Sell
+6
Survey [+9]
[- 3]
B Success(0.143) [+80]
+67
Renew
[- 10]
H
Unfav. Failure(0.857) [+0]
(0.7)
G -13
Sell +6
A [+9]
Success +70
Renew (0.25) [+80]
[-10]
D

No survey
[+0]
C
Failure
-10
(0.75) [+0]
Sell +9
[+9]
Decision trees
1. Start at the right side of the decision tree and move left one column at a
time.
2. For each column, perform either step 3 or step 4 depending upon whether
the nodes in that column are chance nodes or decision nodes.
3. For each chance node, calculate its EP ( = Σ EP * P)
4. For each decision node, compare the EP of its branches and choose the
alternative whose branch has the largest EP In each case.
5. Remove low EP branches.
6. Rinse and repeat.

10
Decision trees
11 Success(0.5) [+80]
+67
Renew EP = 67 * 0.5 + (-13 * 0.5) = 27
F
[- 10] Failure (0.5) [+0]
-13
Fav. E
(0.3) Sell
+6
Survey [+9] Success(0.143) [+80]
[- 3]
B +67
Renew
[- 10]
H EP = 67 * 0.143 + (-13 * 0.857) = -1.56
Unfav. Failure(0.857) [+0]
(0.7)
G -13
Sell +6
A [+9]
Success +70
Renew (0.25) [+80]
[-10]
D
EP = 70 * 0.25 + (-10 * 0.75) = 10
No survey
[+0]
C
Failure
-10
(0.75) [+0]
Sell +9
[+9]
OPTIMAL POLICY

• Do the survey.
• If the result is favorable, Renew the factory.
• If the result is unfavorable, Sell the factory.
• The EP= 12.3 (1, 230,000 L.E.)