Method · 4 min read
How TaxSailor works
A one-page overview of the graph engine, from jurisdictions to a ranked route.
The problem
Cross-border capital passes through multiple tax regimes. A single transaction can be legal, optimal, and completely misunderstood - because the interaction between treaties, withholding, and residency rules is combinatorial.
The model
TaxSailor represents every jurisdiction as a node and every treaty relationship as a weighted edge. Weights encode effective rates after credits, exemptions, and anti-abuse tests. The optimizer runs a log-transformed Dijkstra so multiplicative retention becomes additive cost - mathematically identical, numerically stable.
Inputs
- Origin - where value is generated
- Destination - where the beneficial owner is resident
- Amount and instrument - dividend, interest, royalty, capital gains, or salary
Outputs
- The optimal legal route with intermediate structures and effective ETR
- A baseline direct route for comparison
- Citations - treaty article, domestic code, and effective date
- Deltas - savings vs. baseline, in currency and basis points
What it is not
TaxSailor is decision support, not advice. Every route carries citations so a tax practitioner can validate it. The engine is deterministic - the same inputs always return the same route with the same trace.