ORESTE (Organisation, Rangement Et Synthèse de données rElaTionnEllEs)

ORESTE is a ranking method used when you hold ranking information rather than exact measurement. It asks the decision-maker for neither a precise numerical weight for the criteria nor an exact measurement for the alternatives; it uses only "which is ahead of which" information. Its output is an ascending score for the alternatives; the lower the score, the further ahead the alternative. Roubens proposed it in 1982, and the computation procedure used today was clarified by Pastijn and Leysen in 1989.

The idea behind ORESTE is this: asking people for exact numbers often creates an artificial precision. An expert can say "this supplier is more reliable than that one," but forced to say "exactly twelve per cent more reliable," they make the number up. ORESTE accepts this reality and combines two kinds of rank information: the rank of the alternatives on each criterion, and the rank of the criteria's own importance relative to one another. It reduces both to a single order with the same logic, by computing an average rank (the Besson method).

The philosophical consequence of this approach is fully ordinal. ORESTE looks not at numerical magnitude but at rank; the gap between coming first and second on a criterion, whether small or large in the actual measurement, is the same "one rank" difference to the method. This carries less information than methods that use exact figures, such as TOPSIS, but requires fewer assumptions; it claims no more than the true precision of the information you actually hold.

The method proceeds through four steps.

First, ranking per criterion. For each criterion the alternatives are ranked against one another: on a "lower is better" criterion the smallest value takes first rank, on a "higher is better" criterion the largest value takes first rank.

Second, the combined position. ORESTE combines every alternative's rank on every criterion with that criterion's importance rank. This combined position is the average of the alternative's rank on the criterion and the criterion's importance rank. In this way both "how far ahead is this alternative on this criterion" and "how important is this criterion overall" meet in a single number.

Third, the global rank. The combined positions of all alternative-criterion pairs are gathered together and reordered into a single rank; in this ordering, equal values receive the average rank (the Besson method). Every alternative's combined position on every criterion now corresponds to a value in this global rank.

Fourth, the total score and ranking. The global rank values an alternative receives across all criteria are summed. The lower this total, the further ahead the alternative; ORESTE ranks the alternatives by this total from lowest to highest.

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

The total score is a sum of ranks, not a distance or a percentage. The gap between two alternatives' scores does not say how large the real performance difference between them is; it says only how many rank steps separate them. An alternative scoring 20 is not "twice as bad" as one scoring 10; it has simply fallen to positions further back in the global rank.

The method also carries a threshold, δ. This threshold determines whether the gap between two alternatives counts as a meaningful advantage or as a difference too small for the method to distinguish. If δ is chosen small, many pairs come out "incomparable"; if chosen large, almost everything is deemed equal. The ranking given by the total score should therefore be read together with the choice of δ.

Therefore, instead of writing:

"ORESTE proved A2 was the best"

the report should read:

"With this ranking information, A2 is furthest ahead in the global rank; but because ORESTE uses rank rather than distance, it says nothing about the size of this advantage"

ORESTE works with crisp data, but unlike other crisp methods it does not use the data directly, only in ranked form. DecisionMind holds no extension of this base method.

You need, for each criterion, the alternatives' performance rank (or a measurement convertible to rank) and, for each criterion, whether higher or lower is better. You also need the criteria's own importance rank relative to one another; this is not an exact weight but a rank, such as "first is most important, second less so." Finally, a δ threshold is set. ORESTE does not produce weights, it asks for a rank; decimal weights summing to 1 are not entered here. A minimum of two alternatives and two criteria is required; three to twelve criteria work comfortably.

If you can confidently express your criteria and alternatives only by rank rather than by exact figures, or if your numerical data is not precise, ORESTE is a suitable choice. Expert panels, vote-based assessments, and situations where "there is no exact measurement, but who is ahead of whom is known" are its typical territory.

The case where it should not be used is where you hold reliable, exact numerical data and want to use that data in full. ORESTE reduces numbers to rank; if two alternatives have a small but real numerical difference, this difference is reflected in the ranking merely as "one step," and its magnitude is lost. If you have exact data, cardinal methods such as TOPSIS or SAW use the information more fully.

Only ranking is reliable, exact measurement is uncertain → ORESTE

Exact, reliable numerical data, the full information should be used → TOPSIS, SAW, VIKOR

The importance of the criteria is also known only by rank, no numerical weight → ORESTE

Weights are numerical and clear → weighting methods such as AHP, BWM, SWARA

ORESTE's greatest advantage is the realism of what it asks for. It does not demand a precision the decision-maker does not have; ranking is the natural form of human judgement, and most experts are more comfortable ranking than assigning figures. There is no scale or unit problem either, since no normalisation is needed: ranks are already unit-free. Because it takes the importance of criteria as a rank rather than a numerical weight, much of the arbitrariness in the weight-setting process is removed.

Its limitations arise from how little data it demands. Ordinal reduction loses information: a small difference and a large difference in the real measurement can appear as the same "one step" in the ranking (Bourguignon and Massart, 1994). The choice of the δ threshold is left largely to the decision-maker, and this choice can change the result; a small δ produces many incomparable pairs, while a large δ makes almost every alternative equal. In the global ranking, equal values take the average rank; this is called the Besson method. When a new alternative is added to the dataset, all ranks must be recomputed, carrying the same rank-reversal risk seen in other ranking methods.

The most common mistake is entering the criteria's importance rank as though it were a numerical weight, in decimal form. ORESTE wants a rank such as "first, second, third"; decimal weights summing to 1 are meaningless in this method and produce the wrong result.

A second mistake is reading the total-score gap as a percentage or a ratio. The score is a sum of ranks, not a distance; an interpretation such as "the alternative with the lower score is such-and-such per cent better" cannot be made. A third mistake is leaving the δ threshold at a default without ever considering it; yet this threshold determines whether two alternatives are genuinely distinguishable or should be deemed incomparable, and the report must explain this choice.

The governing principle is this:

The order ORESTE produces is only a summary of the rank information in the input; it carries no claim about numerical magnitude, and an ORESTE ranking presented without explaining the δ threshold is an incomplete report.