Two-thirds of the state, one directorate
This is the fourteenth paper in the African Revenue Intelligence Research Series, instantiating ARIM (Paper 1) for Senegal. It models large-taxpayer-first SENTAX rollout familiar from Nigeria (Paper 4), a legislated Minimum Corporate Tax floor, and strict zero-threshold digital VAT distinct from Uganda's sector mandate (Paper 8) and Botswana's threshold (Paper 11).
The Direction Générale des Impôts et des Domaines (DGID) mobilised 2,915 billion FCFA in fiscal and domanial revenue in 2025 — a 12% increase on 2024 and, distinctively, equivalent to 65% of total state revenue. Few administrations in the series carry so much of public finance on their own shoulders.
DGID administers VAT at 18% (10% for approved tourist accommodation and restaurant services) and Corporate Income Tax at 30% under the General Tax Code, with the Direction des Grandes Entreprises (DGE) handling large taxpayers separately. Where one directorate carries this much, even modest gains in audit selection matter.
Large taxpayers first
The new SENTAX platform, replacing ETAX, becomes mandatory for DGE taxpayers from 1 September 2026, with paper declarations no longer accepted for that segment. No timeline has yet been announced for general taxpayers.
ARIM therefore scores in two tiers. Where a taxpayer is DGE and SENTAX is mandatory, the transaction domain draws on platform data and carries high confidence. Everyone else is scored on a proxy built from returns, and carries low confidence until e-invoicing reaches them. The hypothesis is that DGE taxpayers will show materially more complete transaction data from September 2026 onward.
The floor that does the work
Senegal applies a turnover-based Minimum Corporate Tax whenever ordinary Corporate Income Tax would come out lower. Most countries have to approximate profit-shifting signals from sector margins alone; here the law itself supplies a floor to test against.
Model SN1 measures how far a taxpayer’s ordinary tax sits above that floor. A small positive gap is flagged, because a company engineering its declared profit toward the floor is implicitly declaring the lowest tax-efficient profit the law permits. Since the Minimum Tax is already computed from returns DGID collects, the model can run on existing data with no new infrastructure.
That caution is built into the framework through a false-positive case: an exporter whose tax sits close to the floor but whose documented export-sector costs support the declared margin is not flagged.
No threshold, no grace period
Non-resident digital-service providers must register for VAT from their very first taxable Senegal transaction, with no minimum-turnover threshold — the strictest registration standard documented in the series. A simplified remote regime lets them register without a local fiscal representative or bank account.
Because there is no threshold to cross, Model SN3 cannot wait for one. It flags any provider with observable Senegal activity, such as app-store presence or local payment processing, that is not registered. Broad observable-presence monitoring replaces the threshold-crossing check, and the response begins with a registration compliance notice rather than an assessment.
Six domains, one explainable score
Each Senegalese taxpayer is scored as a domain-weighted sum across the six ARIM domains. Taxpayers are matched deterministically on the NINEA national identifier and probabilistically across registry, director and shareholder evidence, with particular attention to related parties in privileged-tax-regime or non-cooperative jurisdictions.
Tax behaviour
Minimum-tax-floor proximity and VAT-to-sales ratio deviation.
Transaction behaviour
SENTAX data for DGE taxpayers from September 2026; a proxy for the general population.
Network behaviour
Director and shareholder centrality across linked entities.
Sector behaviour
Export Free Enterprise qualification checks and sector-adjusted margin deviation.
Historical compliance
Prior audit-adjustment count and magnitude.
External data
Digital-service presence checks and UEMOA/ECOWAS customs mirror-statistics gaps.
A flag is a signal, never a finding
On a fully synthetic dataset of 30 taxpayers across eight sectors, the seeded floor-engineering case ranked third and both digital-services cases ranked in the top four — each surfaced through independent domain evidence, confirming that the models address separate typologies. This shows mechanical coherence, not real detection performance, and the framework claims no detection or recovery figure until DGID validates it on its own data.
Every case file carries domain-tagged reason codes an auditor can read. Because DGID has occasionally taken the position that OECD Transfer Pricing Guidelines are not strictly binding, any transfer-pricing-adjacent reason code should state the domestic legal basis relied upon (Articles 17–18 of the General Tax Code) rather than implying conformity with guidelines DGID has not fully adopted.