Indirect tax accountability as record agreement: inventory accounting and auditing under digital technologies and big data

An industrial enterprise keeps three accounts of the same goods. Operations hold a physical record of what moved, the enterprise system holds a transactional record of what was documented and priced, and the tax function holds a fiscal record of what was declared. Indirect taxation works to the extent that the three records agree. The standard evasion schemes in manufacturing, namely unrecorded output, invented purchasing, disguised disposal and elastic valuation, all drive the records apart while leaving each one internally plausible. Serialized marking, sensor networks, enterprise event logs and clearance-model electronic invoicing have changed what can be tested, because the three records now exist at population scale and in machine-readable form. Drawing on tax compliance economics, big data research in auditing, continuous assurance and organizational capability theory, this article reconstructs indirect tax accountability as a property of agreement among the three records rather than of any one of them. It first states the conditions under which agreement is evidence at all. A reconciliation carries weight only where its two sides descend from causally separate processes, so the records must contain measurement anchors generated outside the enterprise transaction-processing layer. Integrated systems violate that condition by default. Agreement is then formalized as a set of reconciliation residuals, and a detection condition is derived from stated assumptions. A scheme survives only while its divergence stays inside the noise floor of the tightest independent check that sees it, so the sustainable tax gap is the sum of those noise floors rather than a function of the number of controls in place. A sensitivity analysis shows how the bound moves when the assumptions of separability, exogenous tolerances and zero detection risk are relaxed. Four leakage channels are traced through the inventory cycle, matched to the reconciliation that exposes each, and condensed into four propositions with refuting observations and candidate measures. Uzbekistan is the instantiating setting, where clearance invoicing, serialized labeling and, from January 2026, automated invoice risk scoring have outrun enterprise capability, and where documented administrative reconciliation practice between 2021 and 2023 supplies a live episode and its governance failure mode.