Abstract
<title>Abstract</title> <p>University quality evaluation increasingly relies on heterogeneous evidence: institutional documents, expert judgement, student experience, digital traces and performance indicators. Such evidence is rarely complete or unambiguous, yet evaluation systems often transform it into a single score or category. This article reframes quality evaluation as a problem of evidence governance rather than only measurement. Using 29 anonymised university cases from a domain-specific university education quality evaluation context, it develops a robust boundary-diagnosis framework that combines uncertainty-preserving evidence representation, order-weight robustness checks and case-level diagnostic interpretation. The analysis shows that the three-category structure is stable under the main specification, but the more important finding is not the final classification itself. Boundary cases such as A5 and A9 reveal how similar aggregate scores may conceal different institutional weaknesses. The framework therefore turns a ranking exercise into a governance conversation: which cases are stable, which are threshold-sensitive, which dimensions constrain improvement, and which modelling assumptions matter. The article contributes to debates on responsible quantification, AI-era evidence use and quality governance in higher education by showing how computational tools can support interpretation without replacing institutional judgement.</p>