Skip to main content

Block it in the milliseconds it takes to try.

Device, behaviour, identity, and network signals fuse into a single real-time decision. Every block, challenge, and allow is explainable — so your customers keep moving and your chargeback ratio drops.

WIDTH fraud decisioning dashboard
<100ms
Scoring latency
Median end-to-end decision time from event ingest to action.
7
Fraud types covered
Account takeover, synthetic identity, first-party, ATO mule, scam, card-not-present, API abuse.
99.9%
Decision availability
Multi-region failover with graceful fallback to signed policy defaults.
40%
Chargeback reduction
Measured on first-30-day CNP volume across six production launches.
Seven fraud types, one engine

Every attacker leaves a fingerprint.

The same decision engine watches every touchpoint — onboarding, login, payment, withdrawal — so attackers cannot hop channels to evade detection.

Account takeover

Stopped before the password change.

Device reputation, behavioural biometrics, and session continuity fuse to detect ATO in progress. The engine steps up challenges mid-session without breaking the honest user.

Explore ATO defence →
ATO defence
96%
ATO recall
0.3%
False challenge rate
2s
Step-up decision
Synthetic identity

Catch the composite.

Cross-check identity fragments against bureau, device, and velocity signals.

See synthetic →
Scam + mule

Protect the victim too.

Beneficiary reputation, coercion signals, and peer patterns flag authorised-push-payment scams.

See scam defence →
CNP + chargeback

Before the dispute.

Card-not-present scoring with issuer feedback loops to pre-empt chargebacks.

See CNP →
API abuse

Shape the attacker out.

Rate, pattern, and identity signals block credential stuffing and scraping at the edge.

See API defence →
Rules vs reasoning, again

Rule engines fight yesterday's attack.

Fraud rings iterate weekly. The only defensible posture is a decision engine that learns and explains.

Legacy rules
Rigid, reactive, noisy.
  • Rule tuning lags attack evolution by weeks
  • Honest customers blocked by over-broad thresholds
  • Explanations reconstructed by hand in disputes
  • No shared signal across channels
WIDTH Fraud
Adaptive, targeted, explained.
  • Models retrain on your feedback, daily
  • Peer-aware scoring protects honest customers
  • One-click decision reconstruction for disputes
  • Shared signal across onboarding, AML, cases
"

Our chargeback rate halved in the first quarter. More importantly, our false-decline rate dropped — so we stopped losing good customers to bad models.

PR
Head of Payments Risk
PCI-DSS L1 Acquirer · Asia-Pacific

Score every event. Block only the bad ones.

30 minutes. We'll replay a week of your production events through the engine and show you what you missed.