SIWEC (Simple Weight Calculation)

You have several experts and several criteria, and you want to ask the experts to score the criteria's importance directly, on a number scale (1 to 10, say). No pairwise comparison, no card sorting and no best-worst selection is required. SIWEC was proposed for exactly this situation by Puška, Nedeljković, Pamučar, Božanić and Simić in 2024. The method scales each expert's scores internally, then measures how discriminating that expert's scores are, that is, how spread out they are. An expert who scores in a discriminating way, clearly separating the criteria from one another, has more influence on the result. The output is a weight vector summing to 1. SIWEC does not rank; it distributes weight across criteria.

SIWEC's underlying idea is that not every expert's scoring style deserves equal trust. If an expert gives most criteria similar scores (around 7-8 for everything, say), that expert either genuinely sees little difference between the criteria or struggles to discriminate between them. If another expert gives markedly different scores across criteria (9 for some, 3 for others), that expert clearly perceives the differences between criteria and reflects this in their scores. SIWEC gives more weight to the opinion of the expert whose scores have a higher spread, that is, a higher standard deviation.

This idea carries a philosophical consequence. SIWEC gives weight not to the question "how many experts say the same thing" but to "which expert speaks in a discriminating way." This is meaningful in small groups, where the number of experts is low but every expert's opinion is taken. But where an expert genuinely regards the criteria as close in importance (if they truly are all equally important), a low spread ends up penalised; this is not the expert's error but a consequence of the criteria genuinely being close in importance to that expert.

The method proceeds through six steps.

First, direct scoring. Every expert scores all criteria directly on a numerical scale (1 to 10, say). The result is a score table with experts in rows and criteria in columns. No pairwise comparison or prior ranking is required.

Second, within-expert scaling. Each expert's scores are divided by that expert's own highest score. This brings every expert's scores within the range 0 to 1 on their own terms; different experts' scoring habits (one always scoring high, another always low) are removed at this step.

Third, the discrimination measure. For every expert, the standard deviation of the scaled scores across the criteria is calculated. A high standard deviation shows that expert clearly separates the criteria from one another.

Fourth, weighting the experts' scores. Each expert's scaled score is multiplied by that expert's own discrimination measure. This means a discriminating expert's scores contribute more to the result, and a flat scorer's scores contribute less.

Fifth, summation. These weighted scores are summed across all experts for every criterion, giving that criterion's total importance signal.

Sixth, normalisation. Every criterion's total signal is divided by the total signal across all criteria. The result is a weight vector summing to 1.

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

The weight shows how much a criterion stands out in the experts' discriminating scores; it does not simply mean "the average score is high." Even if a criterion scores high on average, if the experts giving that score discriminate little between criteria (giving them all similar scores), that criterion's contribution to the weight stays limited. Conversely, a criterion with a low average score can receive a higher weight if the expert scoring it low discriminates very clearly between criteria.

For this reason, the weight should not be confused with "the experts' shared view"; it is a combination that gives more share to the opinion of experts who score in a discriminating way. If an expert gives the same score to every criterion (a standard deviation of zero), that expert makes no contribution at all to the result; this is not an error, it follows from the method's own definition.

Thus instead of writing:

"The SIWEC analysis showed the experts found this criterion the most important"

the report should read:

"This weight is the result of a combination that gives more share to the experts who scored this criterion in the most discriminating way; it does not reflect all the experts' shared view equally"

SIWEC works with crisp data: a single numerical score from each expert for each criterion. Alongside the base method, DecisionMind also offers an extension that works with fuzzy linguistic scores (Fuzzy SIWEC); there are two SIWEC members in total.

You need a list of the criteria to be weighted, at least one expert, and a complete table in which this expert (or experts) has scored every criterion on the same numerical scale, with no empty cells. The method produces weights; it does not take weights from outside, and it needs no alternative data. A minimum of two criteria is required; three to fifteen criteria work comfortably. As the number of experts grows, the discrimination differences emerge more reliably.

SIWEC is a suitable choice if you want to take direct scores quickly from several experts, and you need to combine a group's opinion without the extra load of pairwise comparison or card sorting. If the experts' scoring styles differ from one another (one always scoring high, another always cautious), SIWEC's within-expert scaling step removes this difference automatically.

There are two situations where it should not be used. If there is only a single expert and that expert gives similar scores to every criterion, the discrimination measure comes out close to zero and the weights end up nearly equal; the method adds no extra information in that case. If what is wanted is a justification measured through pairwise comparisons between criteria rather than consistency of scoring, AHP or BWM should be preferred.

Several experts are available, quick direct scoring is wanted → SIWEC

Experts can confidently rank the criteria and also give gap information → REVISED-SIMOS

Pairwise comparison and a consistency check are wanted → AHP, BWM

No experts, weights should come from the data → Entropy, CRITIC (objective)

Scores come as fuzzy linguistic expressions → Fuzzy SIWEC

SIWEC's greatest strength is that it asks the expert for nothing more than a direct, single-step scoring; no pairwise comparison, ranking or best-worst selection is required. This is a practical advantage when fast group work is needed with a large number of experts. The method brings experts with different scoring habits onto the same footing through within-expert scaling, and by giving more share to the opinion of the discriminating scorer, it prevents flat, non-discriminating scoring from diluting the result.

Its limitations arise from this same structure. First, reading standard deviation as "discrimination" cannot distinguish a case where an expert genuinely sees the criteria as close in importance from a case where that expert is simply indecisive or careless. Second, the method is a new one proposed in 2024; independent, long-term testing by other research groups is still limited. Third, SIWEC is a subjective weighting method, and the result depends on the scoring style of the experts chosen; a different group of experts can produce different weights. Fourth, in single-expert applications the discrimination measure's power is limited, because there is no other expert's scores to compare against.

The most common mistake is assuming that an expert who gives similar scores to every criterion is "indecisive" or "has not done the job properly"; this can sometimes also show that the criteria genuinely are close in importance to that expert.

A second mistake is skipping within-expert scaling and comparing the scores directly; if this step is skipped, one expert's habit of always scoring high and another's habit of always scoring low mix wrongly into the weights. A third is presenting SIWEC weights based on a single expert's scores as "the group's view"; the discrimination measure only becomes meaningful in comparison across several experts. A fourth is reporting the weight directly as "the criterion's objective importance"; SIWEC is a subjective method and the result depends on the experts chosen.

The governing principle is this:

SIWEC weights are a combination of the experts' direct scores, each weighted by that expert's own discriminating power; that a flat-scoring expert contributes little to the result is not an error but a defined feature of the method, and the report must explain this.