Abstract | A crude oil refinery is employed as a benchmark to test a fundamental assumption in the seismic risk assessment of asset portfolios, i.e., that a fragility function can characterize an asset (or class thereof) with negligible loss of fidelity. Although this is often taken as granted, it also implies that one should not care about breaking the correlation in response that should exist between similar assets subjected to the same ground motion. After all, this is a direct consequence of the summarization of multiple structural analysis results into a single fragility curve that is only parameterized by the (typically scalar) intensity measure. The alternative is to separately consider individual ground motion records and only aggregate per-record results at the final level of impact metrics. Stacking the deck against the conventional approach, a refinery offers an ideal testbed of interconnected assets and multiple refining processes. The results highlight that wherever multiple similar or identical assets are involved in a system disruption, breaking their record-to-record response correlation can severely bias the assessment results.