You are on page 1of 13

Decision Support Models – 2020/2021

Exercises handout no. 2 - Multicriteria Evaluation

Exercise 1
Company MIO, Ltd has issued a call for tenders for the supply of technical equipment to which five
proposals (A1 to A5) were received. The proposals of the competitors will be evaluated based on the
following three criteria: technical quality (C1), guarantee period (C2) and price (C3).

The evaluation committee decided to evaluate the proposals using the simple additive value model,
after evaluating the proposals on a 0 to 100 scale on each criterion, being 0 and 100 the scores
respectively assigned to the proposals with the worst and the best performances.

The prices (criterion C3) of the proposals are shown in Table 1.

Table 1 - Prices of the received proposals.

Price (in
Proposal
thousand €)
A1 1550
A2 1500
A3 1410
A4 1380
A5 1310

It was decided to build a piecewise linear value function for the criterion price (C3), applying the
bisection method. Table 2 presents the results obtained.

Table 2 - Results from applying the bisection method to the criterion price.

Price (in Value Unit


thousand €)
1530 25
1490 50
1420 75

In addition, the evaluation committee applied the direct rating method to evaluate the proposals on
the criterion technical quality (C1), from which resulted the following information:

▪ The ranking of the proposals by decreasing order of technical quality performance is A2, A1, A3,
A5 and A4.
▪ The difference of value of technical quality between A2 and A4 is worth twice of the difference
of value of technical quality between the A2 and A3.
▪ The difference of value of technical quality between A2 and A3 is worth four times the
difference of value of technical quality between A2 and A1.
▪ The difference of value of technical quality between A5 and A4 is equal to the difference of
value of technical quality between A2 and A1.

Decision Support Models – 2020/2021, Exercises handout no. 2 1


The guarantee period (C2) for each equipment is listed in Table 3. The value function corresponding to
criterion C2 is linear.

Table 3 – Guarantee periods of the equipment considered in each proposal.

Guarantee period
Proposal of the equipment
(in months)
A1 14
A2 6
A3 24
A4 10
A5 10

a) Present a table with the values of the proposals on the three criteria. By analyzing that table,
say if there is any proposal (and which) that cannot win the contract, regardless of the weights
assigned to the criteria.
b) The evaluation committee applied the swing weighting method to assess the criteria weights.
They considered that a swing from the worst to the best performances on criterion C2 and on
criterion C3 are worth, respectively, 80% and 70% of the swing from worst to best
performances on criterion C1. Determine the criteria weights, considering that they must sum
up to one.
c) Calculate the overall value of each proposal by applying the additive value model.
d) Perform a sensitivity analysis on the weight of criterion C1 and specify the condition(s) that
could lead to a change of the first ranked proposal. Comment on the graph.
e) In light of the results obtained so far, what recommendation would you make to the evaluation
committee?

Exercise 2
During the process of building a multicriteria model to support the location of an office the following
dialogue occurs:

• Facilitator: Consider an office A located in the city center with a price of €2M. How much
should be the price for an office B, situated at 50 km from the city center, so that A and B
were indifferent?
• Decision-maker: The two offices would be indifferent if the price of the office B were equal
to €1.3M.

Considering the information presented in Table 4 and that the value functions of both criteria are
linear, determine the weights (round to one decimal place) of the criteria “Price” and “Centrality”.

Table 4 – Criteria, descriptors of performance, and respective performance reference levels.

Descriptor of
Criterion Best Worst
performance
Price € million (€M) 1 2
Distance to the
Centrality 0 50
center (in km)

Decision Support Models – 2020/2021, Exercises handout no. 2 2


Exercise 3
a) Explain what a descriptor of performance is.
b) Consider the following descriptor of performance presented in Table 5, which measures the
patients’ travel time to hospital. Indicate one problem associated with it.
c) How could this descriptor of performance be rectified?

Table 5 - Descriptor of performance measuring hospital accessibility.

Level 1 Level 2 Level 3

> 2 min 2 – 10 minutes > 10 minutes

Exercise 4
A decision-maker wants to choose one option among five possible options (P1 to P5). Using the data
presented in Table 6 and Table 7 draw a cost vs. benefit graph and:

a) Indicate, justifying, which are the efficient options taking into account their respective costs
and benefits.
b) Knowing that the decision-maker is willing to pay € 12 thousand for the difference of benefit
value between P1 and P3 on criterion C3, which option would you recommend to the decision-
maker? Justify your answer.
c) If the cost of P1 decreased from € 56 thousand to € 52 thousand, your answer to question b)
would be the same? Justify your answer.

Table 6 - Options’ costs.

Cost (in
Option
thousand €)

P1 56
P2 60
P3 57
P4 48
P5 58

Table 7 - Criteria weights (in %), partial and overall benefit value scores of the options.

Overall
C1 C2 C3 C4
Option Criterion benefit
(39%) (28%) (22%) (11%)
value score

P1 45 80 0 100 50.95
P2 100 0 40 50 53.30
P3 0 100 100 75 58.25
P4 50 50 40 0 42.30
P5 75 25 80 25 56.60

Decision Support Models – 2020/2021, Exercises handout no. 2 3


Exercise 5
In the value tree illustrated in Figure 1, R is the “Risk” criterion, which is composed by sub-criteria R1,
R2 and R3, and B is the “Benefit” criterion, which is composed by sub-criteria B1, B2 and B3. Performance
reference levels “Good” and “Neutral” were defined for the six bottom-level criteria (R1, R2, R3, B1, B2
and B3) and the weights of those criteria were assessed using the swing weighting method,
sequentially: first, to weigh R1, R2 and R3 among themselves (“swing weights” 100, 80 and 60,
respectively); second, to weigh B1, B2 and B3 among themselves (“swing weights” 70, 100 and 30,
respectively); third, to weigh R1 and B2 between themselves (“swing weights” 90 and 100,
respectively).

a) Calculate the weights of R and B knowing that they must sum to one (present all the necessary
calculations).
b) Explain other approach that could be applied to assess the weights of criteria P and Q.

R B

R1 R2 R3 B1 B2 B3

Figure 1 - Value tree.

Exercise 6
An international call for tenders was issued for the supply of a significant quantity of animal feed. Four
proposals (P1, P2, P3 and P4) were received and evaluated on three criteria: price (C1), delivery date
(C2) and food quality (C3). The proposals were evaluated using a 0 to 100 scale on each criterion, being
0 and 100 the scores respectively assigned to the proposals with the worst and the best performances.
Table 8 shows the criteria weights (in %), the value scores of the proposals that were obtained with
MACBETH and their corresponding overall value scores, which resulted from applying the additive
model.

Table 8 - Partial and overall value scores of the proposals and criteria weights (in %).

C1-Price C2-Delivery date C3-Food quality Overall


Proposal
(45%) (30%) (25%) value scores
P1 100 0 40 55.00
P2 31 25 100 46.45
P3 73 83 0 57.75
P4 0 100 60 45.00

Knowing that proposal P3 was removed from the evaluation process - as it did not comply with the
screening criteria - calculate, if needed, the new overall value scores of the remaining proposals. Justify
your answer.

Decision Support Models – 2020/2021, Exercises handout no. 2 4


Exercise 7
For building the value function on “colour printing quality” criterion the M-MACBETH judgements in
Figure 2 (left) was obtained.

Figure 2 – MACBETH judgments matrix (left) and its corresponding thermometer value scale (right).

a) Explain which question should be asked to the decision‐maker to obtain an answer for the
selected cell in the MACBETH matrix above presented?
b) Indicate whether the value scale presented in the MACBETH thermometer is compatible with
the M-MACBETH matrix of judgements. Justify your answer.
c) Explain the difference between the direct rating method and MACBETH in the construction of
value functions.

Decision Support Models – 2020/2021, Exercises handout no. 2 5


Exercise 8
The values in Figure 3 numerically express the attractiveness of five performance levels (L1, …, L5) in
an interval scale on criterion C1. Having this cardinal scale being constructed with the MACBETH
approach, it can be said that:

Figure 3 – MACBETH thermometer.

a) This scale is compatible with a decision-maker that expresses that the difference of
attractiveness between L1 and L2 is weak, and the difference of attractiveness between L1
and L3 is moderate.
b) This scale is compatible with a decision-maker that expresses that the difference of
attractiveness between L1 and L3 is moderate, and the difference of attractiveness between
L1 and L4 is strong.
c) This scale is compatible with a decision-maker that expresses that the difference of
attractiveness between L3 and L4 is strong, and the difference of attractiveness between L4
and L5 is moderate.
d) a) and b).
e) a), b) and c).

Decision Support Models – 2020/2021, Exercises handout no. 2 6


Additional Exercises

Exercise A.1
A company that is specialized in wrapping products wants to buy a new wrapping machine. For that
purpose, the company received five proposals concerning the supply of five different machines: P1,
P2, P3, P4 and P5. Four criteria were selected to evaluate the proposals: price (C1), delivery period (C2),
reliability (C3) and capacity (C4).

The company decided to apply the additive model to evaluate the proposals, using a 0 to 100 value
scale on each criterion, respectively assigned to the worst and to the best performances of the
proposals on each of the selected criteria.

Table 9 presents the prices (C1) of proposals.

Table 9 - Prices of the proposals.

Proposal Price (in


thousand €)
P1 360
P2 240
P3 160
P4 290
P5 180

It was decided to build a piecewise linear value function for the criterion price (C1), applying the
bisection method. Table 10 presents the results obtained.

Table 10 - Value units obtained by applying the bisection method to criterion price.

Price (in Value units


thousand €)
195 75
235 50
280 25

The delivery periods (C2) of the machines are listed in Table 11, being the corresponding value function
linear.

Table 11 - Delivery periods of the wrapping machines

Proposal Delivery period (in


days)
P1 180
P2 150
P3 120
P4 156
P5 120

Decision Support Models – 2020/2021, Exercises handout no. 2 7


To calculate the relative attractiveness of each of the proposals on the reliability criterion (C3), it was
applied the direct rating method, from which resulted the scores presented in Table 12.

Table 12 - Value scores on reliability for each proposal.

Proposal Value scores on


Reliability
P1 100
P2 80
P3 0
P4 40
P5 20

The capacities (C4) of the proposed machines are presented in Table 13.

Table 13 - Capacities of the proposed machines.

Proposal Capacity of the machines (in


units of capacity)
P1 2000
P2 1160
P3 1250
P4 1810
P5 1000

æ x - xmin ö
0.5

The value function for the capacity criterion was defined as v4 ( x ) = 100 × ç ÷ .
è xmax - xmin ø

To assess the criteria weights, consider the following conversation between the decision-maker (D)
and the facilitator (F):

“(…)

(F) – If a proposal that had the worst possible performance on every criterion were given
the opportunity to be improved from the worst to the best performance only in one
criterion, which criterion would provide the best improvement?

(D) – Criterion price.

(F) – Consider a proposal X with delivery period of 120 days and with a price of € 360 000.
How much should be the price of a proposal Y with a delivery period of 180 days in
order to turn proposals X and Y indifferent, knowing that both proposals have the
same performance on the remaining criteria?

(D) – Proposal Y should have a price of € 200 000 to be indifferent to proposal X.

(F) – Consider a proposal X that includes a machine with the best reliability and with a
price of € 360 000. How much should be the price of a proposal Y with a machine
with the worst reliability in order to turn proposals X and Y indifferent, knowing that
both proposals have the same performance on the remaining criteria?

(D) – Proposal Y should have a price of € 220 000 to be indifferent to proposal X.

Decision Support Models – 2020/2021, Exercises handout no. 2 8


(F) – To finalize, answer to a last question. Consider a proposal X that includes a machine
with the best capacity and with a price of € 360 000. How much should be the price
of a proposal Y with a machine with the worst capacity in order to turn proposals X
and Y indifferent, knowing that both proposals have the same performance on the
remaining criteria?

(D) – Proposal Y should have a price of € 295 000 to be indifferent to proposal X. (…)”

a) Present a multicriteria matrix with the partial value scores of the proposals.
b) Determine the criteria weights and calculate the overall value score of each proposal by applying
the additive model. Discuss on the results.

Exercise A.2
In the value tree depicted in Figure 4, R denotes the criterion “Risk”, which is composed by sub-criteria
R1, R2, R3, and B denotes criterion “Benefit”, which is composed by sub-criteria B1, B2 and B3. In all sub-
criteria of R and B the reference levels “Good” and “Neutral” were defined to which were assigned
scores of 100 and 0, respectively. On the sub-criteria of R and B the weights were obtained using the
swing weighing method, first weighing R1, R2 and R3 among themselves (swing weights 100, 80 and 60,
respectively); second, weighing B1, B2, and B3 among themselves (swing weights 100, 60 and 30,
respectively).

The trade-off procedure was applied to weigh R1 and B1 between themselves. Consider the following
dialog occurred between the facilitator (F) and the decision-maker (D):

“(…)

(F) – Consider an option X with a Good performance on R1, and Neutral performance
on B1. What would be the performance on B1 of an option Y, which has neutral
performance on R1, to make X and Y indifferent?

(D) – To make option Y indifferent to option X, option Y should have a performance on


criterion B1 which allows Y to obtain a score of 60 on that criterion. (…)”

Determine the weights of criteria R and B knowing that they must sum up to 1.

Figure 4 – Value tree.

Decision Support Models – 2020/2021, Exercises handout no. 2 9


Exercise A.3
In the framework of multicriteria value models, give an example in which two points of view are
preferentially dependent. Can these points of view be used as fundamental points of view (i.e. criteria)
of that model? Justify your answer. Which procedure (or procedures) may be used to consider these
points of view in an additive model?

Exercise A.4
Consider a decision problem with the criteria structure of the value tree presented in Figure 4. Table
14 shows, for this new problem, the weights (in %) and the partial scores of three options A1, A2 and
A3, as well as the scores assigned to the reference levels “Good” and “Neutral”.

a) Calculate the overall value scores of A1, A2 and A3 by applying the hierarchical additive model
to the data presented, and build a sensitivity analysis graph to show the variation of the overall
values of each of the options when the weight of criterion B varies between 0% and 100%.
Which is the best option for different values of this weight?
b) If option A1 were removed from the analysis what would be the overall value scores of A2 and
A3?
c) Explain why the definition of the reference levels is an essential step to the weighting of
criteria in a multicriteria additive model. Indicate advantages and disadvantages related to the
use of different types of reference revels.

Table 14 - Options’ scores and criteria weights.

R B
Criterion
58% 42%
R1 R2 R3 B1 B2 B3
Option
25% 30% 45% 40% 40% 20%
A1 100 0 0 100 120 50
A2 90 40 100 30 0 100
A3 150 100 40 0 75 -80
Good 100 100 100 100 100 100
Neutral 0 0 0 0 0 0

Decision Support Models – 2020/2021, Exercises handout no. 2 10


Exercise A.5
Regarding the matrix of judgements used to assess criteria weights with MACBETH presented in Figure
5, and when the worst and best references are used, one can say that:

a) The colour criterion is the most important, independently of the references in use.
b) The colour and design criteria have equal weights.
c) The difference of attractiveness between colour criterion and speed criterion is moderate.
d) All of the above.
e) None of the above.

Figure 5 - MACBETH weighting matrix.

Exercise A.6
Figure 6 shows an interval scale for criterion C1 built upon a descriptor
with five performance levels (L1, …, L5).

Which of the following sentences is true?

a) This scale has an arbitrary zero.


b) This scale is unique unless a single linear
transformation is done.
c) The ratio of any two intervals is independent of
the measuring unit and the zero point.
d) a) and b).
e) a), b) and c).

Figure 6 – Value scale.

Decision Support Models – 2020/2021, Exercises handout no. 2 11


Exercise A.7
Consider the MACBETH matrix of judgements presented in Figure 7 and the four possible value scales
shown in Figure 8.

Figure 7 – MACBETH matrix of judgments.

Figure 8 – Value scales.

Which of the value scales is compatible with the above MACBETH matrix?

a) Scale A
b) Scale B
c) Scale C
d) Scale D
e) Scale B and D

Decision Support Models – 2020/2021, Exercises handout no. 2 12


Exercise A.8
Considering the information presented in Figure 9, one can read that:

Figure 9 - M-MACBETH robustness analysis window.

a) The robustness analysis did not make use of information filled in the MACBETH matrices.
b) The Nomark option dominates the Espon option, independently of the set of weights used in
the model.
c) The Conan option additively dominates the Sister option, independently of the set of weights
used in the model.
d) a) and b)
e) b) and c)

Exercise A.9
Table 15 shows the performances, in € thousands and m2, of three apartments (A, B and C), on the
criteria Cost and Size, as well as the performances of the reference levels “Neutral” and “Good” on
each one of these criteria. The value scores of the apartments must be obtained with a decreasing
linear value function on criterion Cost, and with an increasing linear value function on criterion Size.
The performance reference levels “Neutral” and “Good” should have the scores 0 and 100,
respectively. Knowing that the weight of criterion Cost was fixed on 60% and the weight of criterion
Size was fixed on 40%:

a) Calculate the partial value scores of the apartments on each one of the criteria and their
respective overall value scores. Present all the calculations.
b) Knowing that the criteria weights cannot be changed and that the value functions must be
kept linear, say how the apartment ranked last may be ranked first. Justify your answer.
Table 15 - Apartments’ performances and performance reference levels.

Apartment Cost (in € thousands) Size (m2)


A 200 100
B 300 150
C 400 200
Neutral 400 100
Good 200 200

Decision Support Models – 2020/2021, Exercises handout no. 2 13

You might also like