Decision Tree Worked Examples
Decision Tree Worked Examples
J E Beasley
OR-Notes are a series of introductory notes on topics that fall under the broad heading of the
field of operations research (OR). They were originally used by me in an introductory OR
course I give at Imperial College. They are now available for use by any students and
teachers interested in OR subject to the following conditions.
Your company is considering whether it should tender for two contracts (MS1 and MS2) on
offer from a government department for the supply of certain components. The company has
three options:
If tenders are to be submitted the company will incur additional costs. These costs will have
to be entirely recouped from the contract price. The risk, of course, is that if a tender is
unsuccessful the company will have made a loss.
The cost of tendering for contract MS1 only is £50,000. The component supply cost if the
tender is successful would be £18,000.
The cost of tendering for contract MS2 only is £14,000. The component supply cost if the
tender is successful would be £12,000.
The cost of tendering for both contract MS1 and contract MS2 is £55,000. The component
supply cost if the tender is successful would be £24,000.
For each contract, possible tender prices have been determined. In addition, subjective
assessments have been made of the probability of getting the contract with a particular tender
price as shown below. Note here that the company can only submit one tender and cannot, for
example, submit two tenders (at different prices) for the same contract.
In the event that the company tenders for both MS1 and MS2 it will either win both contracts
(at the price shown above) or no contract at all.
What do you suggest the company should do and why?
What are the downside and the upside of your suggested course of action?
A consultant has approached your company with an offer that in return for £20,000 in
cash she will ensure that if you tender £60,000 for contract MS2 only your tender is
guaranteed to be successful. Should you accept her offer or not and why?
Solution
Below we carry out step 1 of the decision tree solution procedure which (for this example)
involves working out the total profit for each of the paths from the initial node to the terminal
node (all figures in £'000).
Step 1
path to terminal node 12, we tender for MS1 only (cost 50), at a price of 130, and win
the contract, so incurring component supply costs of 18, total profit 130-50-18 = 62
path to terminal node 13, we tender for MS1 only (cost 50), at a price of 130, and lose
the contract, total profit -50
path to terminal node 14, we tender for MS1 only (cost 50), at a price of 115, and win
the contract, so incurring component supply costs of 18, total profit 115-50-18 = 47
path to terminal node 15, we tender for MS1 only (cost 50), at a price of 115, and lose
the contract, total profit -50
path to terminal node 16, we tender for MS2 only (cost 14), at a price of 70, and win
the contract, so incurring component supply costs of 12, total profit 70-14-12 = 44
path to terminal node 17, we tender for MS2 only (cost 14), at a price of 70, and lose
the contract, total profit -14
path to terminal node 18, we tender for MS2 only (cost 14), at a price of 65, and win
the contract, so incurring component supply costs of 12, total profit 65-14-12 = 39
path to terminal node 19, we tender for MS2 only (cost 14), at a price of 65, and lose
the contract, total profit -14
path to terminal node 20, we tender for MS2 only (cost 14), at a price of 60, and win
the contract, so incurring component supply costs of 12, total profit 60-14-12 = 34
path to terminal node 21, we tender for MS2 only (cost 14), at a price of 60, and lose
the contract, total profit -14
path to terminal node 22, we tender for MS1 and MS2 (cost 55), at a price of 190, and
win the contract, so incurring component supply costs of 24, total profit 190-55-
24=111
path to terminal node 23, we tender for MS1 and MS2 (cost 55), at a price of 190, and
lose the contract, total profit -55
path to terminal node 24, we tender for MS1 and MS2 (cost 55), at a price of 140, and
win the contract, so incurring component supply costs of 24, total profit 140-55-
24=61
path to terminal node 25, we tender for MS1 and MS2 (cost 55), at a price of 140, and
lose the contract, total profit -55
Hence we can arrive at the table below indicating for each branch the total profit involved in
that branch from the initial node to the terminal node.
We can now carry out the second step of the decision tree solution procedure where we work
from the right-hand side of the diagram back to the left-hand side.
Step 2
Hence the best decision at decision node 2 is to tender at a price of 115 (EMV=32.45).
Hence the best decision at decision node 4 is to tender at a price of 140 (EMV=20.4).
Hence at decision node 1 have three alternatives:
Hence the best decision is to tender for MS1 only (at a price of 115) as it has the highest
expected monetary value of 32.45 (£'000).
With regard to the consultants offer then, ignoring ethical considerations, we could of course,
tender 60 for MS2 only without her help and if we were to do that we would have a 0.95
probability of having our tender accepted. Hence there are essentially three options:
as before, tender for MS1 only at a price of 115: EMV 32.45, downside -50
(probability 0.15), upside 47 (probability 0.85)
tender for MS2 only at a price of 60, unaided by the consultant: EMV 31.6, downside
-14 (probability 0.05), upside 34 (probability 0.95)
tender for MS2 only at a price of 60, with the consultants help, then (assuming she
can fulfil her promise of guaranteeing we will be successful), we have a certain
outcome with a profit of 34 (terminal node 20) - 20 (cash paid to the consultant) = 14
On an EMV basis we would still support our original decision. Looking at the risks
(probabilities) of loosing money, and considering tendering for MS2 only at 60, we would
essentially be paying the consultant 20 to avoid a 0.05 chance of loosing 14, the downside of
tendering unaided.
Paying 20 to guarantee not incurring a loss of 14 which will occur with a probability of 0.05
(one in twenty) does not seem like an awfully good investment and so we should reject her
offer (or offer her a smaller sum of money in return for her guarantee!).
The Metal Discovery Group (MDG) is a company set up to conduct geological explorations
of parcels of land in order to ascertain whether significant metal deposits (worthy of further
commercial exploitation) are present or not. Current MDG has an option to purchase outright
a parcel of land for £3m.
If MDG purchases this parcel of land then it will conduct a geological exploration of the land.
Past experience indicates that for the type of parcel of land under consideration geological
explorations cost approximately £1m and yield significant metal deposits as follows:
manganese 1% chance
gold 0.05% chance
silver 0.2% chance
Only one of these three metals is ever found (if at all), i.e. there is no chance of finding two
or more of these metals and no chance of finding any other metal.
If manganese is found then the parcel of land can be sold for £30m, if gold is found then the
parcel of land can be sold for £250m and if silver is found the parcel of land can be sold for
£150m.
MDG can, if they wish, pay £750,000 for the right to conduct a three-day test exploration
before deciding whether to purchase the parcel of land or not. Such three-day test
explorations can only give a preliminary indication of whether significant metal deposits are
present or not and past experience indicates that three-day test explorations cost £250,000 and
indicate that significant metal deposits are present 50% of the time.
If the three-day test exploration indicates significant metal deposits then the chances of
finding manganese, gold and silver increase to 3%, 2% and 1% respectively. If the three-day
test exploration fails to indicate significant metal deposits then the chances of finding
manganese, gold and silver decrease to 0.75%, 0.04% and 0.175% respectively.
Solution
Below we carry out step 1 of the decision tree solution procedure which (for this example)
involves working out the total profit for each of the paths from the initial node to the terminal
node (all figures in £'000000).
Step 1
Hence we can arrive at the table below indicating for each branch the total profit involved in
that branch from the initial node to the terminal node.
We can now carry out the second step of the decision tree solution procedure where we work
from the right-hand side of the diagram back to the left-hand side.
Step 2
Consider chance node 7 with branches to terminal nodes 15-21 emanating from it. The
expected monetary value for this chance node is given by
The EMV for chance node 6 is given by 0.03(25) + 0.02(245) + 0.01(145) + 0.94(-5) = 2.4
The EMV for chance node 2 is given by 0.01(26) + 0.0005(246) + 0.002(146) + 0.9875(-4) =
-3.275
abandon EMV=0
purchase and explore EMV=-3.275
3-day test EMV=0.7
Hence the best decision is the 3-day test as it has the highest expected monetary value of 0.7
(£m).
Sharing the costs and revenues on a 50:50 basis merely halves all the monetary figures in the
above calculations and so the optimal EMV decision is exactly as before. However in a wider
context by accepting to share costs and revenues the company is spreading its risk and from
that point of view may well be a wise offer to accept.
A company is trying to decide whether to bid for a certain contract or not. They estimate that
merely preparing the bid will cost £10,000. If their company bid then they estimate that there
is a 50% chance that their bid will be put on the "short-list", otherwise their bid will be
rejected.
Once "short-listed" the company will have to supply further detailed information (entailing
costs estimated at £5,000). After this stage their bid will either be accepted or rejected.
The company estimate that the labour and material costs associated with the contract are
£127,000. They are considering three possible bid prices, namely £155,000, £170,000 and
£190,000. They estimate that the probability of these bids being accepted (once they have
been short-listed) is 0.90, 0.75 and 0.35 respectively.
What should the company do and what is the expected monetary value of your suggested
course of action?
Solution
Step 1
Total profit = 0
path to terminal node 8 - the company prepare the bid but fail to make the short-list
path to terminal node 9 - the company prepare the bid, make the short-list and their
bid of £155K is accepted
path to terminal node 10 - the company prepare the bid, make the short-list but their
bid of £155K is unsuccessful
path to terminal node 11 - the company prepare the bid, make the short-list and their
bid of £170K is accepted
path to terminal node 13 - the company prepare the bid, make the short-list and their
bid of £190K is accepted
path to terminal node 14 - the company prepare the bid, make the short-list but their
bid of £190K is unsuccessful
path to terminal node 15 - the company prepare the bid and make the short-list and
then decide to abandon bidding (an implicit option available to the company)
Hence we can arrive at the table below indicating for each branch the total profit involved in
that branch from the initial node to the terminal node.
We can now carry out the second step of the decision tree solution procedure where we work
from the right-hand side of the diagram back to the left-hand side.
Step 2
Consider chance node 4 with branches to terminal nodes 9 and 10 emanating from it. The
expected monetary value for this chance node is given by 0.90(13) + 0.10(-15) = 10.2
Similarly the EMV for chance node 5 is given by 0.75(28) + 0.25(-15) = 17.25
Hence at the bid price decision node we have the four alternatives
Hence the best alternative is to prepare the bid leading to an EMV of £3625. In the event that
the company is short-listed then (as discussed above) it should bid £170,000.
A householder is currently considering insuring the contents of his house against theft for one
year. He estimates that the contents of his house would cost him £20,000 to replace.
Local crime statistics indicate that there is a probability of 0.03 that his house will be broken
into in the coming year. In that event his losses would be 10%, 20%, or 40% of the contents
with probabilities 0.5, 0.35 and 0.15 respectively.
An insurance policy from company A costs £150 a year but guarantees to replace any losses
due to theft.
An insurance policy from company B is cheaper at £100 a year but the householder has to
pay the first £x of any loss himself. An insurance policy from company C is even cheaper at
£75 a year but only replaces a fraction (y%) of any loss suffered.
Solution
Step 1
Total profit = 0
path to terminal node 10 - we have no insurance policy but suffer a theft resulting in a
loss of 10% of the contents.
Similarly for terminal nodes 11 and 12 total profit = -4000 and -8000 respectively.
path to terminal node 13 - we have an insurance policy with company A costing £150
but suffer no theft.
Total revenue = 2000 Total cost = 2000 + 150 Total profit = -150
It is clear from this calculation that when the reimbursement equals the amount lost the total
profit will always be just the cost of the insurance.
Continuing in a similar manner we can arrive at the table below indicating for each branch
the total profit involved in that branch from the initial node to the terminal node.
We can now carry out the second step of the decision tree solution procedure where we work
from the right-hand side of the diagram back to the left-hand side.
Step 2
Consider chance node 5 with branches to terminal nodes 10, 11 and 12 emanating from it.
The expected monetary value for this chance node is given by
0.5(-2000) + 0.35(-4000) + 0.15(-8000) = -3600
Hence the EMV for chance node 1 is given by 0.97(0) + 0.03(-3600) = -108
= -139.8 since y = 40
Hence at the initial decision node we have the four alternatives
1. no policy EMV = -108
2. company A policy EMV = -150
3. company B policy EMV = -101.5
4. company C policy EMV = -139.8
Hence the best alternative is the policy from company B leading to an EMV of - £101.5
minimise x
s.t. -100 - 0.03x >= -108 i.e. EMV B >= EMV no policy
-100 - 0.03x >= -150 EMV B >= EMV A
-100 - 0.03x >= -183 + 1.08y EMV B >= EMV C
i.e.
minimise x
s.t.
x <= 266.67
x <= 1666.67
1.08y + 0.03x <= 83
x >= 0
y >= 0 and y <= 100
maximise x
s.t. the same constraints as above
Solution
Step 1
path to terminal node 9 - we carry out no program and flu does not strike
Total revenue = 0
Total cost = 0
Total profit = 0
path to terminal node 10 - we carry out no program and flu strikes costing the
government £7m
Total revenue = 0
Total cost = 7
Total revenue = 0
Total cost = 7
Total profit = -7
path to terminal node 14 - we carry out a program costing £7m and flu strikes. Now
we would have lost £7m with this flu outbreak but because of the program (which we
assume to be 100% effective) we do not.
The key here is to regard the £7m paid for the program as "insurance" which reimburses the
government for whatever losses are suffered as a result of flu striking. Hence we have
Total profit = -7
It is clear from the above calculation that since (in this case) the reimbursement always
exactly equals the amount lost the total profit will just be the cost of the "insurance" (-£7m).
The situation with the vaccination program is very similar to household insurance where a
single payment guarantees replacement of any losses suffered. Whatever happens the effect
of the insurance will be "as if" nothing had occurred. Under these circumstances the only
expense (in effect) is the cost of the insurance.
path to terminal nodes 15 and 16 similar to the case above where we carry out a
program costing £7m and this insures us against losses. Hence
path to terminal node 17 - we carry out an early warning program costing £3m and flu
does not strike giving
Total revenue = 0
Total cost = 3
Total profit = -3
path to terminal nodes 18, 19 and 20 - we carry out an early warning program costing
£3m, flu strikes and we decide to vaccinate costing £10m. Hence for a total cost of
£13m we are insured against losses so that we have
Hence we can form the table below indicating for each branch the total profit involved in that
branch from the initial node to the terminal node.
We can now carry out the second step of the decision tree solution procedure where we work
from the right-hand side of the diagram back to the left-hand side.
Step 2
Consider chance node 2 (with branches to terminal nodes 10, 11 and 12 emanating from it).
The expected monetary value (EMV) for this chance node is given by 0.1 x (-7) + 0.3 x (-10)
+ 0.6 x (-15) = -12.7
Hence the EMV for chance node 1 is given by 0.25 x (0) + 0.75 x (-12.7) = -9.525
Similarly the EMV for chance node 7 is given by 0.1 x (-7) + 0.3 x (-7) + 0.6 x (-7) = -7
which leads to an EMV for chance node 3 of 0.25 x (-7) + 0.75 x (-7) = -7
The EMV for chance node 8 is 0.1 x (-13) + 0.3 x (-13) + 0.6 x (-13) = -13
and the EMV for chance node 6 is 0.1 x (-10) + 0.3 x (-13) + 0.6 x (-18) = -15.7
Hence the best alternative here is to vaccinate (alternative 4) with an EMV of -13.
The EMV for chance node 4 is therefore 0.25 x (-3) + 0.75 x (-13) = -10.5
and at the initial decision node (node 0) we have the three alternatives:
(1) no program EMV = -9.525
Hence the best alternative is alternative 2, institute a program costing £7m, leading to an
EMV of -£7m.
Note here that it is clear that the concept of the vaccination program being an insurance
against all possible losses could have enabled us to have drawn a much simpler decision tree
(e.g. chance node 3 could be transformed into a "terminal" node of cost -£7m and nodes
7,13,14,15 dropped altogether (similarly for nodes 8,18,19,20)). However, for clarity, we
have presented the decision tree as given above.
With respect to the last part of the question mention discounting, alternative value for a
chance node (other than EMV), changing the decision node ("choose highest EMV
alternative") rule and utility and briefly discuss whether appropriate/inappropriate.
Your company is considering whether it should tender for two contracts (C1 and C2) on offer
from a government department for the supply of certain components. If tenders are submitted,
the company will have to provide extra facilities, the cost of which will have to be entirely
recouped from the contract revenue. The risk, of course, is that if the tenders are unsuccessful
then the company will have to write off the cost of these facilities.
The extra facilities necessary to meet the requirements of contract C1 would cost £50,000.
These facilities would, however, provide sufficient capacity for the requirements of contract
C2 to be met also. In addition the production costs would be £18,000. The corresponding
production costs for contract C2 would be £10,000.
If a tender is made for contract C2 only, then the necessary extra facilities can be provided at
a cost of only £24,000. The production costs in this case would be £12,000.
It is estimated that the tender preparation costs would be £2,000 if tenders are made for
contracts C1 or C2 only and £3,000 if a tender is made for both contracts C1 and C2.
For each contract, possible tender prices have been determined. In addition, subjective
assessments have been made of the probability of getting the contract with a particular tender
price as shown below. Note here that the company can only submit one tender and cannot, for
example, submit two tenders (at different prices) for the same contract.
Possible Probability
tender of getting
prices (£) contract
Tendering for C1 only 120,000 0.30
110,000 0.85
Tendering for C2 only 70,000 0.10
65,000 0.60
60,000 0.90
Tendering for both C1 and C2 190,000 0.05
140,000 0.65
100,000 0.95
In the event that the company tenders for both C1 and C2 it will either win both contracts (at
the price shown above) or no contract at all.
Solution
Below we carry out step 1 of the decision tree solution procedure which (for this example)
involves calculating the total profit for each of the paths from the initial node to the terminal
nodes.
Step 1
path to terminal node 12 - we decide to tender for C1 only at a price of 120K and are
successful
Total cost = 50 + 18 + 2 = 70
Total profit = 50 (all figures in £K)
path to terminal node 13 - we decide to tender for C1 only at a price of 120K but are
unsuccessful
Total revenue = 0
Total cost = 50 + 2 = 52
path to terminal node 14 - we decide to tender for C1 only at a price of 110K and are
successful
Total cost = 50 + 18 + 2 = 70
Total profit = 40
path to terminal node 15 - we decide to tender for C1 only at a price of 110K but are
unsuccessful
Total revenue = 0
Total cost = 50 + 2 = 52
path to terminal node 16 - we decide to tender for C2 only at a price of 70K and are
successful
Total revenue = 70
Total cost = 24 + 12 + 2 = 38
Total profit = 32
path to terminal node 17 - we decide to tender for C2 only at a price of 70K but are
unsuccessful
Total revenue = 0
Total cost = 24 + 2 = 26
path to terminal node 18 - we decide to tender for C2 only at a price of 65K and are
successful
Total revenue = 65
Total cost = 24 + 12 + 2 = 38
Total profit = 27
path to terminal node 19 - we decide to tender for C2 only at a price of 65K but are
unsuccessful
Total revenue = 0
Total cost = 24 + 2 = 26
path to terminal node 20 - we decide to tender for C2 only at a price of 60K and are
successful
Total revenue = 60
Total cost = 24 + 12 + 2 = 38
Total profit = 22
path to terminal node 21 - we decide to tender for C2 only at a price of 60K but are
unsuccessful
Total revenue = 0
Total cost = 24 + 2 = 26
path to terminal node 22 - we decide to tender for C1/C2 at a price of 190K and are
successful
Total cost = 50 + 18 + 10 + 3 = 81
path to terminal node 23 - we decide to tender for C1/C2 at a price of 190K but are
unsuccessful
Total revenue = 0
Total cost = 50 + 3 = 53
path to terminal node 24 - we decide to tender for C1/C2 at a price of 140K and are
successful
Total cost = 50 + 18 + 10 + 3 = 81
Total profit = 59
path to terminal node 25 - we decide to tender for C1/C2 at a price of 140K but are
unsuccessful
Total revenue = 0
Total cost = 50 + 3 = 53
path to terminal node 26 - we decide to tender for C1/C2 at a price of 100K and are
successful
Total cost = 50 + 18 + 10 + 3 = 81
Total profit = 19
path to terminal node 27 - we decide to tender for C1/C2 at a price of 100K but are
unsuccessful
Total revenue = 0
Total cost = 50 + 3 = 53
Total revenue = 0
Total cost = 0
Total profit = 0
Hence we can form the table below indicating for each branch the total profit involved in that
branch from the initial node to the terminal node.
We can now carry out the second step of the decision tree solution procedure where we work
from the right-hand side of the diagram back to the left-hand side.
Step 2
Consider chance node 1 (with branches to terminal nodes 12 and 13 emanating from it). The
expected monetary value (EMV) for this chance node is given by 0.3 x (50) + 0.7 x (-52) = -
21.4
Consider chance node 2, the EMV for this chance node is given by 0.85 x (40) + 0.15 x (-52)
= 26.2
Then for the decision node relating to the price for C1 we have the two alternatives:
It is clear that, in £ terms, alternative 6 is the most attractive alternative and so we can discard
the other alternative.
Continuing the process the EMV for chance node 3 is given by 0.10 x (32) + 0.9 x (-26) = -
20.2
The EMV for chance node 4 is given by 0.60 x (27) + 0.40 x (-26) = 5.8
The EMV for chance node 5 is given by 0.90 x (22) + 0.10 x (-26) = 17.2
Hence for the decision node relating to the price for C2 we have the three alternatives:
It is clear that, in £ terms, alternative 9 is the most attractive alternative and so we can discard
the other two alternatives.
Continuing the process the EMV for chance node 6 is given by 0.05 x (109) + 0.95 x (-53) = -
44.9
The EMV for chance node 7 is given by 0.65 x (59) + 0.35 x (-53) = 19.8
The EMV for chance node 8 is given by 0.95 x (19) + 0.05 x (-53) = 15.4
Hence for the decision node relating to the price to charge for C1 and C2 we have the three
alternatives:
It is clear that, in £ terms, alternative 11 is the most attractive alternative and so we can
discard the other two alternatives.
Hence for the decision node relating to the tender decision we have the four alternatives:
It is clear that, in £ terms, alternative 1 is the most attractive alternative and so we can discard
the other three alternatives.
Hence we recommend that the company tenders for contract C1 only, with a tender price of
£110K because this alternative has the highest EMV of £26.2K.
If the company follows this recommendation the actual outcome will be one of the terminal
nodes 14 or 15 (depending upon chance events) i.e. the outcome will be one of [40, -52].
Hence the downside is that the company may lose £52K (if their tender is unsuccessful).
P(failure) = 0.40
The company is also considering consulting an expert from a research institution to study the
project for a fee of $30,000. The expert will develop a prototype solvent to determine if it
would work at all temperatures. Based on previous experience with the expert, the following
conditional probabilities are realistic appraisals of the expert’s evaluation accuracy.
1) What is the recommended decision if the expert is not used? What is the expected value?
2) What is the expected value of perfect information? Explain what it means.
3) Construct a decision tree for this problem, and determine all probabilities. What is the
optimal decision strategy and its expected value?
4) What is the maximum amount the company should be willing to pay to the expert?
2. A product manager for a soap manufacturer must decide whether or not to offer a new,
biodegradable laundry detergent. The projected profit from a successful detergent is $2
million, whereas failure of the product would result in a loss of $1 million. The manager
currently thinks there is a 40% chance that the product will be successful. Not offering the
product would not change profits.
At a cost of $100,000, the product can be tested. Consumer testing can be favorable, a 50%
chance, or unfavorable. Given a favorable test result, the chance of product success is judged
to be 80%. However, for an unfavorable test result, the chance of product success is judged
to be only 30%.
Construct a decision tree for this problem. What is the optimal decision strategy and its
expected value? Compute the manager's expected value of sample information (EVSI) and
explain what it means. What is the manager's expected net gain of sample information
(ENGSI)?
3. A cab company uses 250,000 gallons of gasoline per year. It may satisfy its fuel
requirement either by one annual contract or separate monthly contracts for the year. The
cost for the annual contract is $1.25 per gallon. The cost for the monthly contract will
average $1.00 if oil production is increased, but the average cost will be $1.75 if oil
production is not increased. The manager estimates a 0.1 probability of no increase in oil
production.
It is also possible to get a professional forecast on oil production at a cost of $10,000. The
manager believes that the following conditional probabilities are realistic appraisals of the
forecast’s accuracy.
Answers:
1. 1) Develop the solvent. Expected value = $340,000 2) $280,000. Chemco should not
be willing to pay more than $280,000 for a perfect information 3) Consult the expert, if
the prototype works, develop the solvent. If the prototype does not work, don’t develop the
solvent. Expected value = $347, 244 4) $37,244
2. Conduct the test, if it is favorable, market the detergent. If the test result is unfavorable,
don’t market the detergent. Expected value = $600,000.
3. 1) Don’t buy the forecast, choose the monthly-contract option. Expected value =
$268,750 2) $1250. The cab company should not be willing to pay more than $1250 for
the forecast