How Detectiv works

Three detection layers, each specialized. All running in parallel so the total latency stays under 50 milliseconds regardless of which signals fire.

Real-Time Scoring Engine

Every transaction hits our scoring engine the moment the authorization request is submitted. The engine evaluates 400+ signals simultaneously - device, network, behavioral, and historical - and returns a risk score before the payment processor responds.

Scores are returned as a number from 0 to 1000 with a plain-language explanation of the top contributing factors. Your team sees exactly why a transaction scored high, not just that it did.

  • 400+ risk signals evaluated per transaction
  • Score + top 5 contributing factors returned
  • Threshold rules configurable per product and geography
  • Feedback loop retrains models from dispute outcomes
SCORE RESPONSE - TXN-447821 - 38ms
risk_score: 847 / 1000
recommended_action: BLOCK
contributing_signals:
→ device_id seen in 4 prior chargebacks (30d)
→ shipping_country != billing_country
→ velocity: 6 attempts on BIN in 90min
→ email domain registered 8 days ago

Behavioral Biometrics

How someone moves through your checkout is as distinctive as their fingerprint. Keystroke dynamics, mouse movement patterns, and interaction timing create a behavioral profile for each user. When a session deviates from that profile, it flags immediately.

This is the layer that catches account takeovers even when the credentials are valid. The bot pasting a password moves differently than the human who set it.

Behavioral session heatmap

Identity Graph Analysis

Synthetic identities are built by mixing real data fragments - a valid SSN, a real zip code, a working phone number - in combinations that individually pass checks but collectively reveal fabrication. Our graph maps every identity element across the network to surface these patterns.

A phone number shared across 12 application attempts across different merchants and a two-week period is a signal no individual fraud system would catch. The graph catches it by design.

Entity Resolution

Links identities across shared data points - phone, email, device, IP, bank account - across your entire transaction network.

Temporal Analysis

Tracks the lifecycle of each identity element. New identities acting like established ones trigger immediate investigation flags.

Ring Detection

Identifies coordinated fraud rings by mapping shared infrastructure across seemingly unrelated accounts.

Real-Time Alerts

Graph anomalies trigger alerts within seconds. Your team sees the full context, not just the flagged account.

Case management for your fraud team

Automated Case Creation

High-score transactions automatically generate investigation cases with pre-populated evidence: scoring details, behavioral replay, and graph context.

Dispute Evidence Package

One click generates a formatted dispute response package with transaction timeline, device evidence, and behavioral data for card network submissions.

Fraud Analytics Dashboard

Real-time view of fraud rate, detection coverage, chargeback ratio, and model performance. Exportable to your BI tool or compliance reporting.

Run Detectiv on your transaction data

We'll analyze a sample and show you exactly what we would have caught - and what your current system missed.

Book a Technical Demo