Fraud scoring & analysis

Every fraud score,
explained and traceable.

Verdict reads your transaction files, scores each one for fraud, and shows you the reasons behind the number — plus how suspicious transactions relate to each other. No black box.

No money movement · out of PCI scope CSV & XLSX in, scored verdicts out
Built for the people who move payments. ISOsISVsProcessorsAcquirers Plugs into the processors and decline taxonomies you already use.
The idea

A risk score is only useful if you can defend it.

Most fraud tools hand you a number and ask you to trust it. Verdict hands you the number and the reasoning — every time, on your own data, in language a reviewer and an auditor can both follow.

Flexibility in configuration

Reshape scoring without code

Weights, thresholds, rules, and denylists are editable data, not engineering tickets. Tune Verdict to your portfolio and your risk appetite, then change your mind tomorrow.

Power in analysis

Depth where it counts

Five scoring lenses, BIN and decline-code enrichment, and relationship analysis that links a single bad actor across merchants — the work that catches organized fraud, not just one-off declines.

Plain-language factors

Every verdict is explained

Each score returns its top contributing factors in words: “card seen across three merchants in two minutes,” not an opaque probability. Reviewers act faster when they can see the reasoning.

Versioned & reproducible

Traceable after the fact

Every run records the model and ruleset versions that produced it. Re-run later and you get the same verdict — the audit trail is built in, not bolted on.

Five lenses, one score

How Verdict looks at a transaction.

Each engine examines a different dimension of risk. Their signals are combined into one explained score, with each engine's contribution shown — so you always know which lens drove the call.

01

Rules

Your own logic, authored as editable data. The conditions you know matter for your portfolio.

02

Decline signal

Normalizes raw processor decline codes into a fraud taxonomy, so an issuer’s “do not honor” carries weight.

03

Graph / rings

Links transactions that share a card, email, or device across merchants — the fingerprint of organized fraud.

04

Merchant baseline

Learns what normal looks like per merchant, then flags the transactions that break the pattern.

05

Denylist floor

Known-bad entities set a minimum band. An analyst’s confirmed-fraud label becomes tomorrow’s automatic catch.

Combined in the aggregator → one score · one band · one explanation, with an optional ML signal when available.
From file to verdict

Four steps, fully hands-off.

Drop a file at the API, the portal, or a watched folder. Verdict does the rest and hands back scored, explained, exportable results.

1

Ingest

Upload a CSV or XLSX — by API, portal, or scheduled drop folder. Columns map to a canonical model automatically.

2

Enrich

Each row gets BIN reference data and a normalized decline class. No full card numbers stored.

3

Score

Five engines run, producing a score, a band, and the factors behind them — plus ring analysis across the batch.

4

Review

Triage in the portal, export to CSV/XLSX, and watch the portfolio trend build run over run.

Deliberately narrow

An analysis engine, not a processor.

Verdict scores and explains. It never touches money — and that design choice keeps it out of the heavy regulation that would otherwise apply, so you can put it to work without a compliance marathon.

  • No money movement. Out of money-transmission regulation by design.
  • No full PANs. Card numbers are truncated and emails hashed on ingest — out of cardholder-data scope.
  • Tenant-isolated. Every report is scoped to its owner; a run you don’t own simply doesn’t exist.
  • SSO & API keys. Analysts sign in with your identity provider; machines use scoped keys.
For builders

An API-first engine that drops into your stack.

The same API powers the portal, your integrations, and the drop-folder watcher. Wire it up once; feed it however you like.

One API, three channels

Push files over HTTP, let analysts upload in the portal, or have a processor drop them in a watched folder. Same pipeline, same verdicts.

Reports are the conduit

Immutable, per-run, tenant-scoped reports: summary, flagged, rings, declines, and a rolling portfolio trend.

Yours to configure

Scoring profiles, rulesets, and denylists are first-class API objects. Calibrate per tenant; reproduce any run on demand.

Bring your transactions. We’ll bring the verdict.

Verdict is rolling out to payments teams now. Tell us about your company and your stack, and we’ll set you up with a scoped trial on your own files.

We’ll reply within one business day.