SWING (Swing Weighting)

You have a set of criteria and want to elicit their importance from an expert in a way that answers not just "which is more important" but "how much more important." SWING works as follows: the criterion judged most important is identified first and given a score of 100. For every other criterion, the expert is then asked: "how valuable is swinging this criterion from its worst value to its best value, compared with swinging the most important criterion the same way?" The expert answers with a score between 0 and 100. Von Winterfeldt and Edwards defined the method in 1986. The output is a weight vector summing to 1. SWING does not produce a ranking; it distributes weight among criteria.

The idea behind SWING is that a criterion's "importance" depends on how far it can actually be changed, that is, on its range. The greater the benefit gained from "swinging" a criterion from its worst state to its best state, the more important that criterion is. This differs from simply asking "which criterion is more important in the abstract"; SWING forces the decision-maker to picture a concrete range (worst to best) and to compare the value of that range against the ranges of the other criteria.

This idea carries a philosophical consequence. If a criterion's range is kept narrow (for instance, if the alternatives already differ little on it), that criterion's swing value comes out low; if the range is kept wide, it comes out high. This means SWING weights depend not only on a criterion's "abstract importance" but also on how much variability that criterion shows within this particular decision situation. If the same criterion has a different range in a different alternative set, its SWING weight changes too.

The method proceeds through a single step.

First step, normalising the swing scores. The expert first identifies the most important criterion and gives it a score of 100. Then, for every other criterion, they score the value of swinging that criterion from worst to best, relative to the most important criterion, on a scale of 0 to 100. The method divides these raw scores by their sum, giving the weight vector, which sums to 1.

The formula, intermediate values and citation format for this single step are given on the DecisionMind method page; this card carries no formulas.

A weight shows a criterion's swing from worst to best relative to the same swing on the most important criterion; it does not show the criterion's abstract or absolute importance. The criterion receiving the highest weight is the one the expert judged to make "the biggest difference when moving from worst to best."

This also reveals a limit of SWING weights. The same criterion can take a different weight if the range between its worst and best value changes; it is the criterion's range within that decision situation that is assessed, not the criterion itself. The report should therefore state clearly which range definition the weights were given against.

For this reason:

"The SWING analysis showed this criterion to be absolutely the most important"

should be written instead as:

"This weight reflects the value of swinging this criterion between the defined worst and best values, relative to the same swing on the most important criterion; if the range definition changes, the weight changes too"

SWING works with crisp data: a single swing score between 0 and 100 for every criterion. DecisionMind holds no extension of this method; the base swing method is used on its own.

You need the list of criteria to be weighted, a defined worst and best value range for every criterion, and the swing scores the expert gives by comparing these ranges against the most important criterion (100 for the most important, 0 to 100 for the rest). The method produces weights and does not require them from outside; it needs no alternative data. At least two criteria are required; three to twelve criteria work comfortably.

SWING is a suitable choice when the expert can concretely define every criterion's worst and best value and can score the value of these ranges comparatively. It works well when the criteria's real ranges (the worst and best values in the data set) are known and the decision-maker can picture them.

There are two cases where it should not be used. SWING should not be used if the criteria's worst and best values cannot be defined clearly, because the swing question then rests on an ambiguous range and the scores come out inconsistent. If the expert can only give an ordering and cannot give a ratio or swing value, rank-based methods such as ROC or REVISED-SIMOS should be preferred.

Criterion ranges are clearly defined, comparative swing scores can be given → SWING

Only an ordering can be given, ranges cannot be defined → ROC, REVISED-SIMOS

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

No expert available, let the weights be derived from the data → Entropy, CRITIC (objective)

SWING's most important advantage is that it grounds weights not in an abstract sense of importance alone, but in the value of a concrete range (worst to best). This stops the decision-maker from using the word "important" independently of the size of the range, and ties the weights to the data's actual variability. The method is simple, requires only a single round of scoring, and is directly compatible with multi-criteria utility methods such as SMART and SMARTS (Edwards and Barron, 1994).

Its limitations stem from this same structure. First, the weights depend on the chosen range definition; if the same criterion has a different worst-to-best range in a different alternative set, its weight changes too. Second, fixing the most important criterion's score at 100 can anchor the decision-maker to their first judgement; Rezaei, Arab and Mehregan (2024) show that SWING is exposed to this anchoring effect in a different way from SMART and BWM. Third, if the expert defines a range incorrectly or unrealistically, for instance by setting a best value that would never actually occur, the weight inflates or shrinks unrealistically as well. Fourth, only the endpoints (worst, best) are assessed; the utility of the values in between is assumed to be linear.

The most common mistake is asking for a swing score directly without defining the criterion's worst and best value; a score given without a defined range does not show what is actually being swung.

A second mistake is confusing the swing score with "the criterion's general importance"; the score reflects only the value of the range within this particular decision situation. A third mistake is directly comparing the same criterion's SWING weight across different analyses; if the range definitions differ, the weights cannot be compared. A fourth mistake is leaving unjustified why the most important criterion was given 100; this choice anchors the entire scale and must be explained in the report.

The governing principle is this:

SWING weights reflect the swing of every criterion between its defined worst and best values, relative to the same swing on the most important criterion; if the range definition changes, the weights change too, and the report must state these ranges explicitly.