| Tacit contribution | Important cues and judgments may be absent from records, while elicitation can distort them. |
| Heterogeneity | Several performers and settings improve discrimination, while indiscriminate sampling raises cost. |
| Contradiction | Variants may be real, while some disagreement is error or incomplete evidence. |
| Tool visibility | Event logs are precise about recorded fields, while invisible coordination and intent remain outside them. |
| Further discrimination | An unseen case can test prediction or transfer, while an unavailable or decision-irrelevant case adds no useful assurance. Retrofitting after inspection loses the intended discrimination. |
| Specialist burden | Consequential decisions may justify a programme, while ordinary recovery needs an affordable stop. |