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.
The four-pillar moat — verified
Atlas's defensibility is architecture + locality, not data volume — and the bundle is what's hard to copy. In a verified scan of 25 competitors, no rival rates "strong" on more than one of Atlas's four load-bearing pillars; Atlas is the only platform strong on all four.
| Pillar | Atlas | Sayari | Quantexa | Moody's | D&B / Altares | ComplyAdvantage | GraydonCreditsafe |
|---|---|---|---|---|---|---|---|
| Native BE+RO registry & UBO depth | ●●● | ●●○ | ●○○ | ●●● | ●●● | ●○○ | ●●● |
| Claims / survivorship / provenance | ●●● | ●○○ | ●○○ | ●○○ | ●○○ | ●○○ | — |
| Configurable, reproducible risk | ●●● | ●○○ | ●●○ | ●○○ | ●●○ | ●●○ | ●○○ |
| Durable agentic OSINT | ●●● | ●●○ | ●○○ | ●●○ | ●●○ | ●●○ | ●○○ |
| Pillars rated "strong" | 4 | 0 | 0 | 1 | 1 | 0 | 1 |
Legend: ●●● strong · ●●○ moderate · ●○○ limited · — none. Two pillars are near-category gaps: per-attribute claims/survivorship/provenance (every named incumbent rates limited or none) and a buyer-configurable, deterministic, hashed risk engine (no competitor ships tunable and replayable scoring). Full ratings for all 25 vendors are in the competitive landscape matrix; the threat side — incumbents now racing toward agentic compliance — is treated honestly below.
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.
It would be wrong to say "nobody covers both regions." Competitor data products span both — GraydonCreditsafe, D&B/Altares, Coface and register passthroughs (Kyckr, Vespia) reach NL/BE/LU and Romania. And TrustRelay is a data aggregator, not a data provider: NorthData (which also spans both) is one of Atlas's own data sources, not a rival — so its coverage is an input to Atlas. The defensible difference is narrower and real: no platform combines native deep resolution + claim-level provenance + configurable reproducible risk + durable investigation across both regions at once. Coverage already exists; the resolved, auditable, investigable bundle does not. See the aggregator-vs-provider note.
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 |
What could erode this — and our answer
A differentiator is only worth as much as its durability. The honest threats, and how the architecture answers each:
| Threat | Why it's real | Our answer |
|---|---|---|
| Incumbents go agentic (D&B × Anthropic/Claude MCP, Moody's screening agents, Dow Jones Integrity Check, ComplyAdvantage "Cassie") | They can close the orchestration gap faster than Atlas can close the data gap | Orchestration is necessary, not sufficient — the durable pillars are claims/survivorship provenance + configurable reproducible risk + native registry depth, which agents alone don't provide. We consume World-Check/Grid/Dow Jones/kompany/Dilisense as claim sources, neutralising the data-scale gap while keeping the architecture moat |
| D&B–Altares entrenchment in Benelux (exclusive D&B partner, official KvK provider, owns Companyweb) | Powerful distribution Atlas lacks on home turf | Compete on resolution + provenance + investigation + Romania, not on Benelux credit-data distribution; win the greenfield verticals (trust offices, CASPs, CSPs) they don't serve |
| Regulatory access risk (RO UBO register may close; Benelux gatekeeper access restricted) | Narrows the native-depth moat | It narrows for everyone; Atlas infers UBO from shareholder/officer graphs when registers close, and moves now in Romania's still-open window |
| Build-vs-buy / "good enough" (stitch Kyckr + ComplyAdvantage + NorthData) | Well-funded buyers may glue point tools | Prove the integrated, auditable whole beats the stitched parts on defensibility and total cost — one replayable system, not five feeds and glue code |
| Heavyweights add BE/RO connectors (Quantexa, Sayari) | Vastly more capital and ER maturity | Locality + mid-market TCO + claim-level auditability; they are enterprise-priced, services-heavy, single-/dedicated-tenant, and thin on native Romania |
For the side-by-side against named competitors, see the competitive landscape; for the emerging, recently-funded challengers (incl. Belgian and AI-native players), see the competitor compendium.