> DECISION_MATRIX_v1.0
Architecture Comparison & Evaluation Matrix
Evaluate all 12 reference architectures side-by-side across latency targets, throughput capacity, and pipeline component trade-offs.
| ID & Architecture | Category | Latency | Throughput | Pipeline Components | Action |
|---|---|---|---|---|---|
FI-ARC-01Real-Time Streaming Transaction Monitoring Architecture | TRANSACTION_MONITORING | < 100ms | 10,000 TPS | 3-node Kafka cluster with Avro schema registry ▸ Distributed Flink on Kubernetes ▸ TimescaleDB hot-storage | View |
FI-ARC-02Low-Latency Multilingual Sanctions & PEP Screening Pipeline | SANCTIONS_SCREENING | < 50ms | 5,000 TPS | In-memory Redis search cluster ▸ Double Metaphone & Jaro-Winkler phonetic indexers ▸ Automated daily list synchronizer | View |
FI-ARC-03Graph Network UBO & Counterparty Clustering Architecture | GRAPH_ANALYTICS | < 500ms for 3-degree traversals | 1,000 queries/min | Clustered Neo4j graph database ▸ Automated batch ingest of corporate registries ▸ Cypher query REST API | View |
FI-ARC-04Explainable AI (XAI) Alert Triage & Hibernation Pipeline | MODEL_GOVERNANCE | < 100ms per alert | 2,000 alerts/min | LightGBM classifier with TreeSHAP ▸ FastAPI microservice ▸ Redis feature cache ▸ Weekly model retraining | View |
FI-ARC-05Virtual Asset On-Chain Forensics & Travel Rule Gateway | CRYPTO_COMPLIANCE | < 500ms | 250 TPS | Multi-hop address taint scoring engine ▸ TRISA / OpenVASP protocol node ▸ Automated deposit quarantine queue | View |
FI-ARC-06Trade-Based Money Laundering (TBML) & Dual-Use Inspection System | TRADE_INTEGRITY | < 2s per document | 500 documents/hour | Deep learning vision OCR pipeline ▸ Automated commodity price API connector ▸ Automated HS code embargo matching | View |
FI-ARC-07Air-Gapped Anti-Tipping-Off Case & SAR Filing Architecture | INVESTIGATION_GOVERNANCE | < 200ms | 500 cases/day | Dedicated PostgreSQL with transparent data encryption (TDE) ▸ Role-based access control with dual-factor authorization ▸ Automated FIU XML formatting | View |
FI-ARC-08Continuous Adverse Media & Public Registry NLP Surveillance | ONGOING_MONITORING | < 15 minutes from publication | 100,000 articles/day | Commercial news wire API ingestion ▸ Pre-trained transformer Named Entity Recognition (NER) ▸ Entity disambiguation against customer master data | View |
FI-ARC-09Unified FRAML (Fraud + AML) Cross-Channel Defense Architecture | FRAML_INTEGRATION | < 2 seconds cross-system notification | 15,000 events/sec | Shared Kafka event streaming bus ▸ Shared customer entity graph ▸ Cross-alert escalation trigger between fraud and compliance queues | View |
FI-ARC-10Privacy-Preserving Interbank Consortium & Zero-Knowledge Intelligence | CONSORTIUM_INTELLIGENCE | < 500ms | 2,000 queries/min | Private permissioned distributed ledger ▸ Homomorphic encryption query service ▸ Real-time consortium mule risk lookup | View |
FI-ARC-11Fair, Non-Discriminatory & Contestable De-Risking Governance Architecture | RIGHTS_GOVERNANCE | 5-day SLA | 500 appeals/month | Dedicated customer self-service contestation portal ▸ Automated adverse action notice generation ▸ Secondary human compliance review workflow | View |
FI-ARC-12Continuous Model Risk Management (MRM) & Emergency Drift Remediation | MODEL_GOVERNANCE | Weekly cadence | 10 models evaluated | Evidently AI telemetry pipeline ▸ Automated weekly Slack/Teams alert on drift ▸ Automated shadow model comparison testing | View |
Comparative decision matrix benchmarking all 12 financial integrity architectures across 36 maturity configurations. Evaluates p95 sub-second latency targets, streaming throughput (TPS), data pipeline architectures (Kafka, Flink, Neo4j, Milvus, Redis), compliance enforcement tiers, and total cost of ownership (TCO).
Comparison Matrix FAQ
What dimensions are compared across architectures in the matrix?
The matrix benchmarks p95 Latency SLA, Throughput (TPS), Primary Pipeline Components, Storage & Graph Engines, Target Compliance Tier, Complexity Index, and Estimated Monthly TCO across all 12 reference architectures.
Which architecture achieves the lowest latency in the matrix?
FI-ARCH-04 (Sanctions & PEP Pre-Auth Screening) achieves the lowest latency with a sub-30ms p95 SLA, leveraging in-memory vector index caches (Milvus/Qdrant) and pre-indexed phonetic hash lookups.
Which architecture provides the highest transaction throughput (TPS)?
FI-ARCH-01 (High-Velocity Streaming Monitoring) scales up to 50,000+ TPS in its Regulated Production configuration using distributed Kafka partitions and Flink stateful streaming clusters.
How are the Total Cost of Ownership (TCO) estimates derived?
TCO estimates represent typical managed cloud compute, storage, memory, and HSM cryptographic hardware costs across AWS/GCP/Azure standard enterprise pricing tiers, excluding human compliance personnel.
Can I filter the matrix by regulatory requirement or infrastructure engine?
Yes. The interactive client allows multi-column sorting and filtering by engine (Kafka, Flink, Neo4j, Milvus, Redis), compliance mandate (FATF, AMLR, BSA, MASAK), or maturity configuration.
Can the matrix data be ingested programmatically by autonomous agents?
Yes. Appending ?format=md provides structured tabular Markdown, and the underlying records are exposed via standard JSON-LD and the TinyCTO Streamable HTTP MCP runtime.