SMART (Simple Multi-Attribute Rating Technique)

SMART is the simplest application of multi-attribute utility theory, ranking alternatives by a single utility score once you hold a decision table filled with numbers. Its output is a utility score between 0 and 1 for every alternative and a descending ranking based on it. Edwards introduced it in 1977. Rather than estimating complex utility functions, the method proposes a simple linear scaling the decision-maker can understand directly, together with weight derivation through ratio rating; it was designed for use in social and public decisions. It produces weights, but these rest on externally supplied importance scores; it does not derive weights from the data itself.

SMART's underlying idea is simple. Rather than asking the decision-maker to estimate a complex utility function, it asks two simple questions: "what is the worst and best case on this criterion, and where does every value in between sit relative to these two ends?" and "what is the relative importance between the criteria?" The method places every criterion's value linearly onto the range 0 to 1 within its own worst-to-best range (this is the default crisp scale). It determines importance directly through ratio rating: the least important criterion is given a low baseline score, the others are scaled up relative to it, and the result is then normalised to sum to 1. These two simple steps turn the decision-maker's intuitive judgement directly into numbers.

Its philosophical consequence places it in the same family as TOPSIS and SAW. SMART is compensatory; a weakness on one criterion can be closed by strength on another, and the final score is a weighted sum. Its difference lies not in the normalisation but in the logic of weight derivation. SMART obtains its weights not through pairwise comparison, as AHP does, but directly through ratio rating. The decision-maker is asked not "how many times more important is this criterion than that one?" but "how many points would you give this criterion?"

The method proceeds through four steps.

First, determining each criterion's end points. For a benefit criterion, the lowest value in the table becomes the "worst" end and the highest becomes the "best"; for a cost criterion, this direction is reversed.

Second, direct scoring (rescaling). The method places the value in every cell linearly onto the range 0 (worst) to 1 (best) between its own criterion's worst and best end. In DecisionMind's classical SMART, this scaling is linear (crisp); the decision-maker may define a non-linear curve if desired, but the default is fixed.

Third, deriving the importance weights. The decision-maker gives every criterion a relative importance score; for instance, the least important criterion is given a low baseline score, and the others are scaled up relative to it. The method normalises these scores into weights summing to 1.

Fourth, the final score. The method multiplies every alternative's criterion scores by their own weight and sums them; it ranks the alternatives from the highest total score to the lowest.

The formulas behind each step, the intermediate tables and the citation formats are given on the DecisionMind method page; this card carries no formulas. (Note: Edwards himself, in later work, also defined two variants, SMARTS, which simplifies ratio rating further, and SMARTER, which differentiates the extreme weighting; DecisionMind's SMART engine uses the classical ratio-rating form.)

The utility score says how well an alternative stands relative to these criterion ranges and the given weights, and nothing more. A score of 0.57 does not mean "57 per cent successful," only a relative position within this particular alternative set and these criterion ranges. A score of 1 does not mean "perfect" but "the best in this set on every criterion." It cannot be compared with a score from a different analysis, because each criterion's worst/best end is built from that analysis's own table.

Thus instead of writing:

"SMART found the best alternative"

the report should read:

"With these criterion ranges and these weights, the alternative with the highest utility score is this one; the score is a summary of the given weights and ranges"

Classical SMART works with crisp data. Alongside the base method, DecisionMind offers two extensions (FUZZY-SMART, SMART-WEIGHT); there are three SMART members in total. You need alternatives in rows, criteria in columns, one number per cell, no empty cells; direction information for every criterion; and importance scores given directly by the decision-maker. SMART converts these into weights; it does not generate the importance scores themselves, which are a subjective input. A minimum of two alternatives and two criteria is required; three to twelve criteria work comfortably. As the number of criteria grows, it becomes harder for the decision-maker to give each one a consistent ratio score.

SMART is a suitable choice if the decision-maker can score the criteria's importance directly and quickly, without taking on AHP's pairwise-comparison burden, and the criteria are numerical with a fully populated table. It is practical for decisions with few criteria (three to six), under time pressure, resting on the judgement of a single decision-maker or a small group.

There are two situations where it should not be used. The first follows from its philosophy: if no compromise is acceptable on one criterion, SMART is unsuitable. The second follows from scale: if there are many criteria, direct ratio scoring becomes inconsistent; in that case, AHP's pairwise comparison or BWM's limited comparison burden gives more reliable weights.

Few criteria, fast direct scoring, compensation accepted → SMART

Many criteria, consistency checking through pairwise comparison wanted → AHP

Low comparison burden, comparison only against the best/worst criterion → BWM

Comparing ranking methods rather than weights is wanted → TOPSIS, SAW, VIKOR

No compromise on one criterion, sub-threshold alternatives must be screened out → screening first, then ranking

SMART's greatest strength is its simplicity and transparency: both the scaling (direct 0-to-1 placement) and the weighting (direct ratio scoring) are operations the decision-maker can easily understand and check; neither AHP's pairwise comparison matrix nor its consistency check is needed. It is applied quickly with a small number of criteria, and the result can be followed step by step on the table. Edwards and Barron's (1994) later work argued that this simplicity can, in practice, produce decisions similar to those from more complex utility functions.

Its limitations stem from its simplicity. First, direct ratio scoring is subjective and carries no consistency check; unlike AHP, the method does not itself catch whether the decision-maker has given contradictory scores. Second, Zanakis, Solomon, Wishart and Dublish's (1998) comparative simulation study produced an important finding: simple weighted-sum methods (SMART included) can produce different rankings from more complex methods such as AHP under certain data conditions. This shows that the choice of method can affect the result. Third, as the number of criteria grows, the consistency of ratio scoring falls. Fourth, it carries the same full-compensation assumption as TOPSIS. Fifth, the linear-scaling assumption is not always realistic; on some criteria, utility does not rise linearly, in which case variants such as SMARTS/SMARTER or a non-linear scaling are needed.

The most common mistake is giving importance scores without justification or at random; SMART's simplicity makes it easy to set weights carelessly, and this turns into a weakness. A second mistake is marking a criterion's direction (higher/lower is better) incorrectly; the scaling is then built the wrong way round and the ranking is reversed. A third is reading the utility score as a percentage or an absolute measure of success and comparing scores across different analyses. A fourth is trying to give direct ratio scores to a large number of criteria; at that point one should switch to AHP or BWM. A fifth is assuming linear scaling holds for every criterion; on a criterion with diminishing marginal utility, this assumption gives a misleading result.

The governing principle is this:

A SMART score is a direct summary of the criterion ranges and importance scores the decision-maker supplied; if these inputs are subjective and contested, the score is contested too, and the report must show this.