CIMAS (Criteria Importance Assessment)

CIMAS is not a ranking method. It scores criteria rather than alternatives, and its output is a set of weights summing to 1. The method answers the question "how important is this criterion" by asking several experts and drawing on the disagreement among them. Bošković, Jovčić, Simić, Švadlenka, Dobrodolac and Bacanin proposed the method in 2023 and published it in the journal Facta Universitatis in 2025. It is still a young method and has already begun to be applied in areas such as supplier selection, crowd-sourced transport and innovation-capacity assessment.

CIMAS produces a weight; it does not ask for one. It does the same job as other subjective weighting methods such as AHP, BWM and SWARA, but through a different mechanism. It scores directly rather than using pairwise comparison, and it rests on the spread of scores between experts rather than on a best-worst choice.

The idea behind CIMAS is this: a criterion's importance is not just its average score across experts, but also how much the experts disagree on it. If every expert gives a criterion the same score, there is no "information" in that criterion; everyone already agrees, so the criterion does not distinguish between decision-makers. If experts give a criterion widely differing scores, there is a genuine disagreement on it. The width of that disagreement is a signal of how discriminating the criterion is. CIMAS turns the width of the spread between experts' scores, for each criterion, into that criterion's weight. This spread is the difference between the highest and the lowest expert score.

The philosophical consequence is that CIMAS is a consensus mechanism sensitive to expert experience: a more experienced expert's score carries more weight in the process. This rejects the assumption that "everyone's opinion counts equally" and foregrounds experience with an explicit justification. If that justification is unacceptable, CIMAS's philosophy will clash with the decision's own.

The method proceeds through six steps.

First, expert weights. Each expert is given a weight proportional to their years of experience; these weights sum to 1. If no experience information is available, every expert receives equal weight.

Second, scale equalisation. The scores each expert has given to each criterion are divided by that criterion's column total, making them unit-free.

Third, the expert-weighted matrix. The equalised scores are multiplied by the relevant expert's weight, so that an experienced expert's score carries more weight in the outcome than an inexperienced one's.

Fourth, column extremes. For each criterion column in the expert-weighted matrix, the highest and lowest value is found.

Fifth, the difference. For each criterion, the difference between this highest and lowest value is calculated; this difference represents the (weighted) magnitude of disagreement among experts.

Sixth, converting to weight. Each criterion's difference is divided by the sum of all the differences, converting them into final weights that sum to 1.

The formulas behind each step are given on the DecisionMind method page; this card carries no formulas.

CIMAS's output is a weight, not a ranking. The criterion with the highest weight is not the "best" criterion. It is the criterion that creates the most disagreement among experts and that experienced experts emphasise the most. A low weight on a criterion does not mean it is "unimportant." It means that all the experts agree on that criterion, so it does not distinguish between decision-makers.

The quality of the weight depends on whether the experts are genuinely well-informed. If the experts are well-informed, the disagreement is a genuine signal. If the experts are biased, meaning the disagreement stems from position rather than knowledge, CIMAS converts that bias into weight all the same. The method does not question the source of the disagreement, only its magnitude. Bošković and colleagues (2025) therefore proposed a reliability index. If this index is low, that is, below 0.1, the results are considered acceptable. If the index is high, reviewing the expert panel is recommended.

For this reason:

"CIMAS found the most important criterion"

should be written as:

"According to this expert panel's scores, the criterion generating the most disagreement, and most emphasised by the most experienced experts, is this one; the weights change if the panel changes"

CIMAS works with crisp data: a single score given by each expert to each criterion. DecisionMind holds three CIMAS members alongside the base method (classical CIMAS and picture-fuzzy extensions); which one fits your data situation is explained on the relevant data-type cards.

You need experts in rows, criteria in columns, an importance score in every cell (for example on a 1–10 scale), and no empty cells. You also need, for each expert, a years-of-experience figure or a similar measure of seniority. If this information is unavailable, DecisionMind assumes equal weight. CIMAS does not ask for weights, it produces them. These weights then feed into ranking methods such as TOPSIS, SAW and WISP. A minimum of two experts and two criteria is required. Three to twelve criteria work comfortably. If all experts give exactly the same score on a criterion, that criterion's weight comes out at zero. This is not an error; it is a zero-information signal.

CIMAS is a suitable choice when you have a table of criterion-importance scores given independently by several experts and want to explicitly build expert experience into the weighting process. Its typical territory is supplier selection and expert-focused group weighting.

The cases where it should not be used are as follows. If you have only a single decision-maker's opinion, CIMAS is not suitable. The method's logic of disagreement between experts does not work with one person; AHP, BWM or SWARA are more suitable here. If you suspect your experts disagree because of position or interest rather than knowledge, CIMAS is again unsuitable, because it does not distinguish this; the reliability index measures only the magnitude of disagreement, not its source. If your data consists of directly measured figures rather than expert scores, objective weighting methods such as Entropy or CRITIC are more suitable.

Multi-expert importance scoring, experience should be weighted → CIMAS

A single decision-maker, pairwise comparison preferred → AHP, BWM, SWARA

The data itself, not a weight, should determine criterion importance → Entropy, CRITIC

Suspicion that experts disagree from bias rather than knowledge → review the panel first, then CIMAS

Ranking is needed, not weighting → TOPSIS, SAW, WISP, CRADIS and other ranking methods

CIMAS's chief feature is this: it derives criterion importance from disagreement among experts rather than from a single "average score." This automatically gives a low weight to criteria all the experts agree on, since these criteria are not discriminating in the first place. Building expert experience into the process explicitly and traceably gives a concrete answer to the question of whose opinion should count for how much. The computational burden is small, and the steps can be followed on the table. Because the method proposes its own reliability index, it also signals when a result should be treated with suspicion (Bošković et al., 2025).

Its limitations largely stem from the logic of disagreement between experts. First, CIMAS does not distinguish the source of disagreement. A healthy difference of opinion arising from knowledge and one arising from bias look identical to the method and turn into the same weight. Bošković and colleagues (2025) proposed the reliability index precisely for this reason. Second, the result is quite sensitive to the expert experience weights. If the expert with the highest experience weight scores very differently from the rest of the panel, that single expert can dominate the outcome. Third, if all experts agree on a criterion, that criterion's weight falls to zero. This is methodologically correct but may be counter-intuitive to decision-makers and mistaken for an "error." Fourth, classical CIMAS works with crisp scores. It cannot capture an expert's own uncertainty, such as a range like "between 6 and 8." Because of this gap, a picture-fuzzy extension was published shortly after the method was proposed (Kara, Yalçın, Kaygısız, Simić, Örnek and Pamučar, 2024).

The most common mistake is giving every expert equal weight without accounting for experience and presenting this as "neutrality." But equal weight is also a decision and must be justified.

A second mistake is mistaking a criterion's zero weight for a data error and removing that criterion from the table. In fact, a zero weight shows that the experts fully agree on that criterion, and this is a legitimate result. A third mistake is reporting the weights without checking how much a single expert with the highest experience weight is driving the outcome. Without removing that expert from the panel and recalculating, the robustness of the result cannot be known. A fourth mistake is presenting the weight CIMAS produces as a "true importance" without noting that the panel could be small or one-sided. A fifth mistake is treating the result as final without ever calculating the reliability index.

The governing principle is this:

CIMAS weights are an experience-weighted summary of the disagreement among that panel's experts. If the panel is small or a single expert dominates, the weights carry that dominance too, and the report must show this.