Free · No Integration · No PII · 48-Hour Turnaround

Send us 10 of your historical fraud cases. Get back decisions that explain themselves.

Within 48 hours, you receive — for every case — the decision our engine would have made, the exact signals that fired, the matched fraud pattern, and an investigator-ready narrative. Most fraud teams use it to pressure-test whether "explainable" vendors actually explain anything.

Email Your Cases or write to hello@zarelva.com Download CSV Template JSON Example
₹0for the scored sample
48 hrsfrom data to full readout
0integration work required
No PIIanonymized tokens only

How it works · three steps

STEP 1

Fill the template

Up to 10 historical events in a simple CSV or JSON — ideally a mix of confirmed fraud, confirmed clean, and the disputed cases your team argued about. Blank fields are fine.

STEP 2

Anonymize & send

Replace real identifiers with tokens (ACC-A, DEV-1). No names, numbers, PAN/Aadhaar, VPAs, or IPs — the exercise stays cleanly outside DPDP personal-data scope. Email it to hello@zarelva.com.

STEP 3

Get the readout

Per case: decision · risk score · fired signals with evidence · matched named pattern · investigator narrative. Plus a summary of where we agreed with your known outcomes — and where we diverged, and why.

Why the divergences matter most. Where the engine disagrees with your recorded outcome is where the conversation gets interesting — either we missed something your team saw, or your controls missed something the signals caught. Both are worth knowing, and both are free.

The template · every field maps to a real signal

Nothing decorative — each column below corresponds to a signal the engine actually evaluates. Event types: login · session_start · password_reset · payout_change · beneficiary_add · withdrawal

FieldWhat it answers
is_new_deviceDevice never seen on this account before (≈7 days)?
emulator_detected / hooking_detectedEmulator, VM, or Frida/Xposed-style instrumentation present?
remote_control_appAnyDesk / TeamViewer-class app on the device?
datacenter_asn / proxy_vpn / impossible_travelHosting IP, VPN/proxy, or physically impossible location jump?
recent_password_reset / mfa_or_recovery_changedCredentials or recovery paths changed shortly before the event?
recent_payout_changePayout method or beneficiary changed just before a withdrawal?
seconds_since_login / rushed_sensitive_actionHow fast did they race from login to the sensitive action?
velocity_spike / time_to_cashout_secondsWithdrawal bursts, or funds in-and-out within minutes?
shared_payout_cluster / device_account_countShared payout destinations, or one device across many accounts?
known_outcomeYour ground truth — fraud, clean, or disputed — so we can score ourselves against it.
⬇ CSV Template (with 3 example rows) ⬇ JSON Example (for engineering teams)

Want the full picture? · the paid pilot

ZSIG Historical Replay Assessment

₹1,50,000 + GST · fixed fee · four weeks · international clients: $1,800 fixed
  • Up to 500 historical events replayed through the engine — no integration, no production access, anonymized data only
  • Decision Accuracy Report — exact-match and divergence rates against your known outcomes, by fraud typology
  • Per-decision narratives — the full explainability output your analysts and auditors can judge for themselves
  • Signal Gap Map — the fraud patterns in your data that current controls missed, and what closing each gap takes
  • Regulatory readiness view — how the decision trail maps to RBI audit and dispute-defence expectations under the 2026 framework
  • One-hour executive readout, delivered by the practitioner who ran the analysis

Founding pilot partners: the full pilot fee is credited against a first-year deployment if you proceed. The deliverables are yours either way.

Book a 30-Minute Scoping Call About the Engine
Who runs this. Zarelva is a Bengaluru fraud intelligence practice built on 6+ years of frontline fraud operations at Flipkart, Google, and Amazon. The methodology is published openly — ZAIS, our account-takeover standard with RBI mapping, and ZEOL™, our emergency outflow-lock standard. We work under NDA from the first conversation. MSME UDYAM-KR-03-0675917.