What sets TrustRelay apart
The one-line answer: Most KYB/AML tools sell you data or a score. TrustRelay Atlas gives you a resolved, provenance-tracked knowledge graph and a reproducible investigation behind every decision — and it's built natively for Benelux + Romania.
When someone asks "what makes TrustRelay different?", these are the five answers, each grounded in how the system actually works.
1. Ontology-first knowledge graph — not a data dump
Competitors typically return records: you query a company and get that vendor's row. Atlas
resolves every source into a single canonical entity in a versioned
ontology (v3.5), with relationships (Ownership, Directorship,
RegisteredAt, …) in a synced property graph. That means you can ask
graph questions — walk the ownership chain to the UBO; find every company at this address; show the
common connections between these two people — that a flat data feed simply cannot answer.
Proof in the code: entity resolution uses blocking + multi-algorithm similarity (Jaro-Winkler, Levenshtein, token-set) with explicit decision bands (merge ≥ 0.95, review 0.60–0.80, create < 0.40) and a dual PostgreSQL + Neo4j model. See Entity resolution.
2. Claims + survivorship — provenance other tools discard
When two sources disagree, most platforms silently pick one (or show you both and make you decide). Atlas keeps every claim — value, source, trust, timestamp — and applies explicit survivorship to choose the winner, recording the decision as a mutation. The canonical value ships with its alternative claims attached.
This is the single biggest differentiator for defensibility: when a regulator asks "why does your record say this?", the answer — winning source, trust weight, the rule that decided, and the alternatives — is attached to the value itself. And protected fields (PEP/sanctions) can never be silently overwritten by a registry refresh.
Proof in the code: the claims-first read path returns the preferred value plus
alternative_claims; enforce_protected_fields blocks unauthorized overwrites of is_pep /
is_sanctioned; every change is a mutation with before/after + source.
3. Configurable risk matrices — your policy, not ours
Risk in Atlas is a configurable matrix engine, not a black-box vendor score. Compliance authors dimensions, factors, scoring methods (reference-lookup, boolean, threshold-ranges), aggregation, and one-way escalation rules — and every evaluation is deterministic and hashed for audit, with live preview and portfolio re-scoring before commit.
Why it wins: a vendor risk score is opaque and un-auditable; an Atlas evaluation is reproducible (same input + snapshot → same score, verified by SHA-256 fingerprints) and encodes your policy. See Risk scoring and ADR-019.
4. Durable agentic OSINT investigations — repeatable, not bespoke
A single investigation runs seven OSINT modules as a durable Temporal workflow — it survives restarts, tolerates partial failure, retries activities, and keeps full history. The same subject investigated twice yields a comparable result.
Why it wins: competitors that bolt OSINT on as a manual analyst task can't reproduce or audit it; Atlas makes the investigation itself a first-class, replayable artifact. See Temporal workflows.
5. Strict multi-tenant isolation — enforced in the database
Tenant isolation is enforced by PostgreSQL Row-Level Security through a restricted atlas_app
role (fail-closed), and provider credentials are encrypted per tenant with AES-256-GCM + HKDF
per-tenant subkeys — so even a cross-tenant decryption attempt fails cryptographically.
Why it wins: many platforms isolate tenants only in application code. Atlas isolates at the data layer, which matters for regulated customers with strict segregation requirements. See Security & multi-tenancy.
6. Built for Benelux + Romania
TrustRelay treats NL/BE/LU/RO as its home market, not cells in a global grid — native KVK integration, jurisdiction-aware survivorship and UBO thresholds, and a graph that makes cross-border Benelux↔Romania structures legible. See Benelux & Romania.
The elevator pitch
| Ask | Most competitors | TrustRelay Atlas |
|---|---|---|
| "Show me the company" | A data record | A resolved entity in a knowledge graph |
| "Why this value?" | Trust us | Winning claim + alternatives + mutation history |
| "How risky?" | A black-box score | A reproducible, hashed matrix evaluation you configured |
| "Can you reproduce it?" | Manual re-check | A durable, replayable workflow |
| "Who really owns it?" | Shareholder list | Computed UBO with the ownership chain |
| "Is my data isolated?" | App-level | Database RLS + per-tenant encryption |
For the side-by-side against named competitors, see the competitive landscape.