Recommendation labels
01
Best Overall
Balanced composite result.
02
Best Price
Highest return when selling; lowest cost when buying.
03
Lowest Risk
Strongest merchant risk profile.
04
Highest Liquidity
Best visible fit for the requested amount.
Decision system
Methodology
Recommendations stay deterministic, auditable and versioned so users can understand how each result was produced.
Evaluation pipeline
01
Normalize first
Exchange-native payloads are retained separately, then mapped into a stable quote contract.
02
Filter deterministically
Side, asset, fiat, amount bounds and payment method eligibility are hard filters before scoring.
03
Score transparently
Best Overall balances price (40%), risk (25%), liquidity fit (25%) and execution quality (10%).
04
Version every model
Risk and execution outputs include score version, confidence and human-readable factors.
05
AI explains, never ranks
The language model summarizes deterministic results. It cannot alter filters, scores or recommendations.