Skip to main content

95% fewer false positives, zero missed SARs

WIDTH scores every transaction against 12 typologies — structuring, TBML, mule networks — and pre-drafts FinCEN-format SARs within the 30-day filing window. Analysts triage signal, not a 40M-alert backlog.

Book a Demo → See the Platform
WIDTH AML monitoring dashboard
95%
False positive reduction
Benchmarked across 40M+ alerts; validated against FFIEC BSA/AML Manual thresholds.
24/7
Continuous screening
Re-screens against OFAC, UN, EU, and MAS lists within seconds of each list delta.
60%
Analyst time reclaimed
Alert-to-analyst ratio cut from 8:1 to 3:1 across production deployments.
12
Typologies out of the box
Structuring, layering, TBML, APP fraud, velocity, peer-group anomaly — each mapped to FATF Rec 20 or AMLD6 Article 6.
Rules vs behaviour-aware scoring

Fixed thresholds miss layering, flag salary

Peer-group normalisation cuts 95% of false positives while holding recall above 98% on structuring and TBML typologies.

Rule-based
Too loud, too late.
  • $10K CTR threshold flags 80% legitimate payroll
  • TBML and APP fraud invisible to amount-only rules
  • No peer-group context — sole trader flagged like a casino
  • Threshold tuning queued behind engineering sprints
  • Analysts reconstruct narrative from raw ledger data
→
AI-native (WIDTH)
Targeted, timely, explained.
  • Peer-group scoring: sole trader vs sole trader, not vs casino
  • Graph detection surfaces structuring rings across 40M+ alerts
  • Onboarding, fraud, and AML risk fused into one score
  • Threshold changes deploy same day, versioned and audit-logged
  • FinCEN-format SAR narrative pre-drafted on every flagged case
12 typologies mapped to FATF Rec 20 & AMLD6

Each model versioned, cited, committee-reviewable

MAS Notice 626 §9, FCA SYSC 6.3, and AUSTRAC AML/CTF policies encoded as executable overlays — not PDFs in a drawer.

Structuring & layering

Map structuring rings across institutions

Graph engine traces deposit structuring below the US $5K SAR threshold, funnel accounts, and cross-institution layering — per Wolfsberg TBML Principles and FATF Rec 20 typologies.

Explore detection →
Transaction graph
98%
Network recall
Graph
Native model
Sub-sec
Propagation
Sanctions & PEP

Re-screen on every list delta

OFAC, UN, EU, MAS, and HMT lists monitored continuously — each match timestamped and audit-logged against the list version that triggered it.

See coverage →
Trade-based ML

Flag invoice anomalies, phantom shipments

Detects over/under-invoicing and fictitious trade flows per Wolfsberg TBML Principles — price deviation scored against 130+ live commodity benchmarks.

See TBML →
Crypto bridging

Fiat-to-chain in one risk score

Chainalysis and Elliptic risk signals fused with your core ledger — FATF Rec 15 travel-rule data attached to every flagged chain transaction.

Explore crypto →
SAR drafting

SAR narrative drafted at alert creation

FinCEN-format narrative pre-drafted with cited transactions, typology classification, and MAS Notice 626 §9 / FCA SYSC 6.3 policy references — filed within the 30-day window.

How cases work →
Threshold tuning

Cut false positives, keep recall

Daily backtests replay new alerts against last quarter's case outcomes — surfacing threshold drift before your MLRO does. False-positive reduction averages 40–60% in 90 days while staying inside MAS Notice 626 §6.5 and FATF Rec 10 risk-based-approach guardrails.

See tuning →

Cut your alert backlog by 95% in 30 days

30-minute session: we replay 90 days of your alerts, score them against 12 FATF typologies, and show the exact false-positive reduction.

Book a Demo → Explore Platform