A single failing criterion can veto an otherwise near-perfect whole.

Evidence

Swiss decision-making about their nuclear bomb shelters provides a nice little illustration of how noncompensatory decision-making works in practice (Schärlig 1985). Switzerland evaluated each feature of a nuclear shelter in a point system as follows: very good = 1, good = 2, usable = 5, and bad = 40. Defense officials’ overall evaluation of shelters used the sum of ten criteria coded using this system. The decision rule was to use the sum of the ten criteria (so larger means worse) with the ordinal categorization scheme 10–11: very good, 12–20: good, 21–50: useable, and > 50: bad. As one can see, bad on any dimension basically means the shelter is virtually unusable. If one dimension is bad, i.e., 40, and the other nine are all very good, the total is 49. So if one dimension is bad the shelter is just barely “usable.” — Gary Goertz, Social Science Concepts and Measurement, 970

Categorization

• CANDY: The Swiss scored bomb shelters so that one ‘bad’ dimension (40 points) outweighs nine ‘very good’ ones, leaving the shelter barely usable.
• XREF: Connects to vetocracy and veto-player arguments (Tsebelis) and to reliability engineering, where a single point of failure caps system performance.