This study develops SIGNAL (Sources, Indicators,
Graduated response, Narrowed scope, Audit trail, Learning loop), a framework
that converts the continuous monitoring of large taxpayers by an assigned
account manager into risk-triggered audit selection in jurisdictions that have
adopted electronic invoicing. It addresses a paradox observable in Ukraine and
increasingly relevant to countries implementing mandatory e-invoicing: tax
administrations now receive transaction-level data in near real time, yet the decision
to audit a large taxpayer is still driven largely by annual plans, while the
richest structured accounting data are requested only after an audit has begun.
Methods. A design-science approach was combined with
documentary analysis, structured practitioner reflection and simulation. Phase
one analysed a corpus of twenty-eight Ukrainian legal instruments, oversight
reports, official publications, foreign regime documents and international
guidance covering 2015–2026, together with five foreign e-invoicing and
reporting regimes. Phase two constructed the framework, drawing on the author’s
practice as the assigned account manager for about ten large taxpayers in the
Large Taxpayers Office of the State Tax Service of Ukraine and subsequent work
in tax debt collection and in an anti-corruption unit. Phase three evaluated
alternative trigger rules through a Monte Carlo simulation of a synthetic
portfolio of 1,000 large taxpayers observed over 24 months, with sensitivity
analysis across six scenarios.
Scientific
novelty.
The study identifies a gate–insight gap: in Ukraine, e-invoice data are used
intensively at the gate, to suspend the registration of individual invoices
(7.7 million invoices carrying UAH 144.1 billion of VAT were suspended in 2021
– H1 2024, while 49.4% of complaints against regional suspension decisions were
upheld), but only weakly for insight, that is, for the selection and scoping of
audits. Since 1 January 2026 the standard audit file SAF-T UA has been
requested from large taxpayers only during a documentary audit. In the
simulation, a trigger that requires a weighted signal score to persist for two
consecutive months raised the precision of audit selection from 10.0% for a
single-signal rule to 90.5%, while retaining 99.7% of non-compliant taxpayers,
at the cost of a mean detection delay of about three months; the advantage
held, with reduced magnitude, under correlated signals, gradual onset of
non-compliance and lower prevalence.
Contributions. The study extends the compliance risk
management literature by specifying how monitoring signals should be converted
into audit decisions, rather than assuming that better data automatically
produce better selection, and links that literature to research on continuous
auditing, where alert overload is a recognised failure mode. It contributes to
the literature on corruption in tax administration by showing that a formal
signal log reduces the discretion of the account manager, the officer with the
longest and closest relationship with a large taxpayer.
For countries introducing mandatory e-invoicing,
including Poland since February 2026 and the European Union under the VAT in
the Digital Age package, the study offers transferable design rules for linking
real-time data to proportionate, documented enforcement. It is, to the author’s
knowledge, the first framework that treats the account manager’s monitoring and
the audit selection decision as a single, auditable process rather than as
separate administrative functions.