Entity Patterns

Abbreviation expansion improves match rate by 23% — "Mgmt" → "Management", "Intl" → "International" account for 41% of fuzzy corrections.

Legal suffix stripping (LLC, Ltd, Corp, GmbH) reduces false negatives by 18% without impacting precision.

Token-order-invariant matching catches 12% more fund names where words appear in different order (e.g. "Total Return PIMCO" vs "PIMCO Total Return").

US-domiciled entities represent 58% of resolution volume. Auto-resolve rate is 12% higher for US entities due to richer GLEIF coverage.

Cayman Islands and Luxembourg fund registrations cluster around quarter-end dates, causing 3x spike in resolution requests.

APAC entity names with transliteration variants (Nomura/野村) require multi-script matching — current Jaro-Winkler underperforms by 15%.

Bimodal confidence distribution: 68% of resolutions score >90% (auto-resolve) or <40% (no match). The 40-90% review band contains only 32% of volume.

Adding LEI as input boosts composite score by avg 28 points. Only 23% of incoming records include LEI — opportunity to enrich upstream.

Country code match contributes disproportionate signal — wrong country drops confidence by avg 31 points even when name is 95%+ similar.

GLEIF cache staleness: 8.3% of cached entities have registration status changes within 90 days. Recommend 30-day refresh cycle for active counterparties.

Duplicate entity detection: 3.1% of resolved entities map to multiple LEIs (merged, transferred, or lapsed registrations). Ownership chain analysis resolves 78% of ambiguity.

Fund-to-manager clustering: top 20 asset managers control 74% of resolved fund entities. Manager LEI lookup resolves 89% of ambiguous fund names.

S&P Global PMID linkage available for 62% of resolved entities — enables downstream portfolio analytics and risk aggregation.