# 144

Dear Hamed Taherdoost, Mitra Madanchian

I read your paper

A Comprehensive Overview of the ELECTRE Method in MultiCriteria Decision-Making

My comments

1-In page 2you say “The ELECTRE evaluation methods have manifold application areas and are widely used decision-making methodsthat can be applied in a vast range of areas from transport to environmental protection programs”

I am afraid that I do not concur with you on the underlined sentence. It is true that ELECTRE is in the great league of MCDM main rational methods such as PROMETHEE, TOPSIS and SIMUS used to solve complex problems, but it is not widely used as TOPSIS and PROMETHEE.

In my opinion, it is due to its complexity and the subjectivity in choosing thresholds. Regarding complexity remember that MCDM methods are intended for practitioners use not by mathematical scientists. I do not think that the average user is not familiar or interested in mathematical concepts and formulas, and from this point of view I find the method difficult to understand

2- In page 4 the article mentions as a disadvantage that it can produce rank reversal. True, but it can happen in all MCDM methods

3- The necessity to use different versions according to the purpose of the study, complicates even more its use. The ideal is a method that can produce selection, sorting and ranking with the same software. This is somehow also valid for PROMETHEE due to its different versions

4- In page 6, What is a ‘bloomlean matrix’. Perhaps you wanted to say ‘Boolean matrix’

5- In page 6, what is ’over ranking’? You do not explain is

6-- In page 7, you say “when there is strong and weak outranking”

At first sight it seems obviously important, but not from my point of view, because the decision is arbitrary, and the reason is that we do not know what influence can have that discrimination based on the difference between alternatives scores. A MCDM method may tell for instance that the difference in scores between alternatives A3 and A4 is say 5.6 – 2.9. It is a large difference, but from the practical point of view, what does it means? Nothing, only that there is a strong outranking.

It does not take into account the proportional number of criteria that each alternative is linked to. It only indicates that A3 > A4.

In my opinion, in analysing several alternatives subject to a set of criteria - that in reality are objectives - common sense says that the best solution is the one that satisfies the largest number of objectives.

For instance, a water supply project is designed to benefit the largest number of objectives like agriculture and cattle rising, forestry, water for people, recreation, etc.

However, the same computational method may show that A3 has a score of 5.6, linked to only one criterion. If we have, say eight criteria, the weighted score of A3 will be 5.6 x 1/8 = 0.7. If A4 is linked to say five criteria, its weighted score will be A4 is 2.9 x 5/8 = `1.81, and then A4 > A3.

Observe that the weighting factor is not related with the relative importance of each criterion. This demonstrates that it does not matter if there is a strong or weak outranking, at least from my point of view.

I would like to know yours.

In my opinion, irrelevant of the score values, the best alternative of the ranking must be tested in two ways that have different purposes.

6.1- By a well-known rational sensitivity analysis using all criteria simultaneously AAT (All at A Time), and not an arbitrary chosen one OAT (known as ONE at A Time). This is fundamental for finding the strength of the selected alternative

6.2- By determining in what extent, it satisfies objectives that are crucial for the company, owner or entrepreneur developing a project. Fundamentally for finding in which extent each criterion or objective has been fulfilled, because that is what the DM and stakeholders are interested on.

As an example. In studying an investment for a project, A, the stakeholders have established 9.6 % as a minimum Internal Rate of Return (IRR), or target. The question is: If alternative A is the best, which is the real IRR? It could be higher or lower. If it is too low the company probably will discard A and select the second-best alternative B, that yields 10. 6%.

If in this case the score for A = 3.56 and for B =1.17; evidently A > B. However, as seen, this strong outranking has no effect.

This is fundamental for finding if the most important objective has been reached, and if not, in what percentage it is in excess or in defect.

Why the real IRR is normally different to the wished? Because a person cannot compute all the interrelationships between multiples aspects.

If you are a manufacturer of a bundle of cables for the car industry, and there three different processes for it (alternatives), you compute your costs considering raw material, electricity, wages, tests and rejects and arrive at a certain figure, say 24.18 Euros/bundle. You want to determine which is the manufacturing alternative or process that will produce that piece at that cost, or slightly higher.

Suppose you run a MCDM method and find the resulting production costs considering all the elements direct and indirect than intervene in its fabrication, that is, the above mentioned, plus: Risk in delays from suppliers, unexpected machine failure, strikes, taxes, storage costs, absences, etc., you will get the real cost of your production, and this is what really is important when you compare it with your calculation of 24.18 Euros/bundle.

What is important is in how much it surpasses or falls behind your calculation, because it will affect diverse areas of your company. In other words, you need to know how well your expectations on cost, sales, image, market penetration, etc., are achieved.

Sorry for the length of my comment, but I always try to justify what I assert.

7- In page 7 “Strong outranking relies on solid bases, whereas weak outranking has questionable grounds”

Where is the justification of this assertion?

8- Why do you use ‘nodes’ which are components of networks and no defining them? I can understand them in outranking as sources and sinks, but you do not explain.

9- In page 12 “the absence of detailed real-world examples limits the ability to assess the method’s effectiveness and identify potential challenges or areas for improvement in specific contexts”

I agree with this, but did you realize that you contradict yourself with this comment, because you said at the beginning about the wide usage of the method. If it is so widely used, couldn’t you find an example?

These are my comments, I hope they can help

Nolberto Munier