Operations Blindness in Customer-Side Decisioning: Information Limits, a Data-versus-Formalism Decomposition, and Negative Results

Operations blindness is customer-side decisioning misattributing enterprise-caused failures to customers when it cannot see enterprise state. Proposition 4 shows that, under A-ALIAS-HIST — the entire customer-visible history having the same distribution under both causes — any customer-side-only rule has slice errors summing to one. Proposition 2 partially identifies the decision-relevant quantities, with bound width equal to the silent fraction cross-sectionally under calibration. An internally prespecified three-legged test of a state-conditioned artifact failed in three runs: leg 1 reached +14.85 points against a 15-point threshold, leg 2 raw-data recovery was 83.6% against a below-50% criterion, and the primary endpoint was null. The strongest supported finding (post-hoc, matched): a trained operations-aware model scored 0.371 six-cause accuracy, against 0.321 for an equally trained customer-only model and 0.255 for the hand-coded artifact. Field evidence is weaker: one suggestive reopening association with unknown causes, and one explanation-response pattern whose interval does not establish equivalence. Preprint — not peer reviewed. This is the author's version of a research manuscript prior to formal peer review, submitted to Information Systems Frontiers (Springer) on 28 October 2026 (manuscript ISFI-D-26-03294, under review). This record will be updated with the publication reference if the paper is published. The replication package for this paper is archived on Zenodo at https://doi.org/10.5281/zenodo.23721283 (concept DOI, always latest; exact version used for the results in this draft: https://doi.org/10.5281/zenodo.23721881, v4.9.2; latest package version at preprint posting: v4.9.3, https://doi.org/10.5281/zenodo.23722557). If this preprint is published in a journal, this record will be updated with the publication reference.