A weighted college score can look objective because its final result is a number. Yet the selected criteria, scales and weights contain judgments. A small lead may disappear when reasonable priorities change. Instead of treating the top score as a verdict, calculate where the ordering switches and decide whether the underlying preference is stable enough to act on.

The OECD and European Commission's handbook on composite indicators discusses construction choices and sensitivity analysis. This article applies that general methodological concern to an invented personal shortlist. The scores below are fictional ratings with no connection to actual universities or published league tables.

Make the scale assumptions explicit

Suppose two feasible colleges are rated on affordability and curriculum fit using a common zero-to-one-hundred scale, where higher is better. Assume the scale intervals are meaningful enough for a weighted sum. College A scores 90 on affordability and 60 on fit; B scores 60 and 90. Both already meet the student's non-negotiable requirements.

Weight on affordability A weighted score B weighted score
30% 69 81
50% 75 75
70% 81 69

Let w be the affordability weight and one minus w the fit weight. A's score is 60 plus 30w; B's is 90 minus 30w. Equating them gives 60w equal to 30, so the switch occurs at w equal to 0.5. A leads only when affordability receives more than half the weight.

That result is not a flaw. It reveals that the decision is genuinely preference-dependent under the stated ratings. A student who cannot justify whether affordability deserves 45% or 55% should not present a narrow weighted lead as robust. The next step may be to investigate the criteria more carefully rather than calculate extra decimal places.

Check whether the scores can be added

Raw tuition and a five-star fit rating cannot be multiplied by weights and added without a justified transformation. The chosen normalisation can change the result. A min-max scale built from the current shortlist can also change when another college is added, even if A and B themselves are unchanged.

Examine double counting. If annual price and four-year net cost are both heavily weighted, affordability may effectively receive more influence than the displayed percentage suggests. Similarly, several reputation measures can encode nearly the same information. A model with ten criteria is not necessarily more balanced than one with three.

Keep hard constraints outside the compensatory score. An unaffordable funding gap, an unavailable required subject or an unresolved eligibility condition should not disappear because another criterion scores highly. Filter for feasibility first, then compare preferences among the remaining options.

Stress uncertainty in the ratings

Weights are not the only uncertain inputs. A curriculum-fit score of ninety may depend on an optional module that is not guaranteed. Replace uncertain inputs with plausible ranges supported by evidence and repeat the calculation. If the preferred college changes under small defensible adjustments, label the result sensitive.

Record the reason behind each rating and the source that could change it. A course handbook, written fee quote or confirmed accommodation offer is more useful than an unexplained number. The model should direct attention toward evidence gaps, not conceal them.

The final decision note can be concise: “A leads when affordability carries more than half the weight, while B leads below that threshold, given these invented ratings.” That sentence is more informative than announcing a winner. It tells the student which judgment actually drives the choice and allows a family or adviser to disagree with the assumption without having to reject the arithmetic.