FARE (Factor Relationship)

FARE is not a ranking method; it does not rank alternatives, it produces criterion weights. What sets it apart from other objective weighting methods (Entropy, CRITIC, CILOS) is its input: not a decision table (alternatives by criteria) but a criterion relationship table (criteria by criteria). Every cell in this table shows how strongly one criterion affects another. FARE is therefore not purely objective: the relationship table itself has to come from somewhere, and it is usually filled in from expert judgement. DecisionMind lists it alongside the objective weighting methods, but because its input rests partly on expert opinion, it should be read as a hybrid sitting between subjective and objective methods.

Ginevičius proposed the method in 2011. The idea is close to AHP's pairwise-comparison logic, but instead of asking about every pair of criteria separately, it starts from the influence relationship between criteria.

FARE's underlying idea is this: in a decision, criteria do not stand independently of one another, they influence each other. The more strongly a criterion affects other criteria, the more central it sits within the decision, and the higher a weight it deserves. In Ginevičius's original design, part of these relationships is taken from an expert and the rest is derived analytically from the system's own operating logic, so that every criterion pair need not be asked about individually.

A consequence of this idea is that FARE defines weight not as "the criterion's own inherent value" but as "the criterion's influence within the system." If a criterion looks important on its own but does not affect other criteria, FARE assigns it a low weight. If this distinction is acceptable, that is, if you hold that "what matters is the criterion that affects a great many things," FARE is the right tool; if the weight needs to come directly from the spread in the data or from an expert's bare priority, Entropy, CRITIC or AHP fits better.

FARE in DecisionMind proceeds in a single step; the user supplies the relationship table already fully completed.

A single step: row sums and normalisation. For every criterion, the row is summed to show how strongly that criterion affects all the other criteria (and itself). This sum is divided by the grand total of every criterion's row sum. The result is a weight vector that sums to 1.

An important limitation applies here. In Ginevičius's original method, only part of the relationship table is taken from an expert; the remaining cells are completed analytically from the criteria's own operating conditions. DecisionMind does not implement this completion step; it expects the user to supply the table filled in from start to finish. This simplifies the method, but it also hands part of the original FARE's "saying a lot from little data" advantage back to the user.

The formula behind this step and the intermediate tables are given on the DecisionMind method page; this card carries no formulas.

The FARE weight measures a criterion's total influence power within the relationship table; it has nothing to do with how "good" or "bad" the criterion itself is measured to be. A weight of 0.60 does not mean "this criterion is 60 per cent of the decision"; it means "in this relationship table, this criterion's total share of influence is roughly 60 per cent." The weight is a property of the relationship table. If an expert assesses the relationships differently and changes the table, the weight changes with it.

FARE weights are therefore not independent of the opinion of the expert who filled in the relationship table; a different expert's table can give a different weight distribution. The table need not be symmetric; how strongly one criterion affects another need not equal the reverse. This flexibility also means that consistency is not checked from outside.

Thus instead of writing:

"FARE proved that C1 is the most important criterion"

the report should read:

"According to this relationship table, C1 is the criterion that most strongly affects the others; its weight of 0.60 reflects this share of influence, and the table itself rests on expert judgement"

FARE works with crisp numbers, but its input is not a decision table but a relationship table: a square table whose rows and columns share the same list of criteria, with an influence-power number in every cell. DecisionMind holds no other member of FARE; it stands as a method on its own.

You need: a complete list of criteria; an influence-power value for every pair of criteria, including a criterion's relationship with itself; and all of these values must be positive. The table need not be reciprocal, as in AHP (one value the inverse of the other); this means both flexibility and a lack of oversight. A minimum of two criteria is required; as the number of criteria grows, filling in the relationship table consistently becomes harder, because the number of cells to be filled grows with the square of the criterion count.

FARE is a sound choice if a real interaction exists between criteria, that is, if one enlarges or shrinks another, and if this interaction can be reasonably estimated by expert judgement. It is favoured in studies that want to capture the relative influence power of criteria without taking on the full pairwise-comparison burden that AHP requires.

The situations where it should not be used follow from its own philosophy. If criteria are independent of one another (none affecting another), the relationship table becomes meaningless, and a data-driven method such as Entropy or CRITIC fits better. If no expert is available to fill in the relationship table, FARE cannot be applied; DecisionMind does not complete the table analytically. If consistency needs checking (as with AHP's consistency ratio), FARE provides no such check.

Criteria influence one another, an expert can estimate this influence → FARE

No influence relationship, the weight should come directly from the data → Entropy, CRITIC, CILOS

Full pairwise comparison and a consistency check are wanted → AHP

Criteria should be ranked and the number of comparisons reduced → SWARA, BWM

FARE's greatest strength is that it captures the interaction between criteria without asking for the full set of n(n-1)/2 pairwise comparisons that AHP requires; in the original design, only part of the relationships is taken from an expert. The calculation is simple and can be followed on the table. Where criteria form a system that influences itself rather than standing independently, it offers a more realistic picture than methods that ignore this interaction.

Its limitations stem largely from the relationship table itself. First, because the table carries no reciprocity requirement, inconsistencies can go unnoticed; DecisionMind's own engineering note flags this risk explicitly. Second, the weight depends heavily on the opinion of the expert who filled in the table; a different expert can produce a different weight distribution, and the method does not hide this subjectivity. Third, as the number of criteria grows, the number of cells to fill grows with the square of that count, and the table may not be filled in consistently. Fourth, the simplified form DecisionMind implements does not include the original method's "analytical completion from missing data" step; the user must fill in the entire table alone.

The most common mistake is confusing the relationship table with a decision table; FARE must be given data on influence between criteria, not alternative performance data.

A second mistake is reporting the weight as the criterion's "natural importance"; the weight reflects the influence assessment of the expert who filled in the table. A third is presenting a result filled in by a single person as an "objective analysis"; because the input is subjective, the result carries that subjectivity too. A fourth is feeding implausible row sums into the calculation without noticing them, such as marking a criterion as affecting itself excessively strongly. Whether the row sums are plausible should be checked. A fifth is trusting a single table without comparing it against tables filled in by different experts.

The governing principle is this:

A FARE weight reflects the assessment of influence between criteria made by the expert who filled in the relationship table; if the table changes, the weight changes, and the report must present this as an expert judgement, not a data finding.