First, the key terms.

Criterion
A dimension used to evaluate an alternative.
Normalization
Transforming measures to a comparison scale under stated rules.
Compensation
Allowing a strong score on one criterion to offset a weak score on another.

The idea, explained

A weighted score combines judgments under a defined model, not a universally objective truth. Criteria should represent distinct concerns and use clearly defined scales. Weights reflect trade-offs over specified ranges, so a weight is not meaningful independently of how scores are constructed. Pairwise comparisons can help people articulate priorities, but inconsistent judgments and workload grow with the number of elements. Explain whose values are represented, how disagreement is handled and which requirements cannot be traded away.

STEP 1

Define criteria and constraints

Distinguish desirable features from mandatory thresholds. A system that fails a required privacy condition may need exclusion rather than a small penalty balanced by low cost. Avoid counting the same benefit twice under correlated criteria.

STEP 2

Elicit scores and trade-offs

For direct rating, define anchored meanings and which direction is better. In pairwise comparison, record how strongly one criterion is preferred relative to another under a specified method. Review inconsistent comparisons rather than letting software conceal them.

STEP 3

Assess robustness and explain

Compare rankings under plausible weights and scores. If a tiny change reverses the result, report that fragility. A group decision may require presenting disagreement rather than averaging away a conflict about values.

Follow a worked example.

Two fictional software options receive benefit scores on a common 0–100 scale. A has usability 80 and reliability 60; B has usability 60 and reliability 90. With weights 0.4 and 0.6, A scores 68 and B scores 78.

Walk it through, one step at a time

  1. Calculate A: 0.4 × 80 + 0.6 × 60 = 68. Calculate B: 0.4 × 60 + 0.6 × 90 = 78.
  2. Change usability weight to 0.8 and reliability to 0.2. A becomes 76 and B becomes 66, reversing the ranking.
  3. Discuss whether the trade-offs and scores are credible and whether either option fails a mandatory requirement. The changing result shows the role of values, not a mathematical error.

Where people get stuck

Weights and normalized scores can embed subjective choices even when the final calculation is fully reproducible.

Now try it yourself.

Pause before opening each answer. Explain your reasoning in your own words.

1. Check the core idea

Why should an MCDA report show sensitivity to weights?

Reveal the answer & reasoning

It reveals whether the preferred option is robust or depends narrowly on uncertain or disputed judgments.

2. Apply what you learned

In pairwise judgments, A strongly outranks B and B strongly outranks C, but C strongly outranks A. What needs attention?

Reveal the answer & reasoning

The judgments may be inconsistent under the chosen framework. Review meanings and evidence rather than presenting an unexplained ranking as objective.