Onboarding checks, sanctions screening and fraud decisions

Your risk checks run before a payment or after it, in systems the authorization path never calls.

Symptoms in your system

Customers drop out during identity verification and nobody can tell you which step lost them.
Business onboarding stalls because ownership documents are checked over email, and the same company gets verified twice.
Companies incorporated in one country are declined because a founder lives in another, and every document check asks for a domestic address.
Sanctions screening runs as an overnight batch, so a payment can settle hours before the name is checked.
Fraud rules live in a spreadsheet, and the person who wrote them has left.
Chargebacks arrive for payments your rules approved, and nobody can reconstruct which rule let them through.

What we build

Identity verification runs inside signup, and a customer who needs manual review is told during the flow.
Business onboarding collects ownership, control and document evidence once, and the file an examiner asks for is the one the system already holds.
Ownership and control are verified as separate tests, so a 25 percent shareholder in one country and a CEO in another each clear the check where they live.
Cases the model cannot represent reach a reviewer with the evidence attached, so an unusual ownership chain is decided under the same risk policy as everything else.
Sanctions screening covers the customer and the payment before the money moves, so a hit stops the transfer while it is still yours to stop.
Fraud rules and scores are versioned, and any approved or declined payment can be replayed against the rules that decided it.
Decisions happen inside the authorization path, within the time the scheme allows, so a bad payment is declined at authorization and never becomes a chargeback.
Analysts work one queue with the customer, the device and the payment history in front of them, and clearing a false positive stops being an investigation.

Failures that surface a year later

Screening thresholds left at the vendor default, so the alert queue grows with volume and hiring is the only lever left.
Fraud rules written for one attack and never retired, so the customers they block are increasingly the good ones.
No record of why a customer was approved, so a sample an examiner picks turns into a reconstruction exercise.
Checks run once at signup with nothing watching afterwards, so an account whose behavior changes is still treated as it was on day one.
Verification wired to one country of residence, so a legal structure the form cannot hold is declined by a field validation and never reaches a reviewer.
Sanctions list updates on a schedule nobody owns, so a name added this week is screened next month.

Standards and rails

KYC and KYB programs
AML transaction monitoring
OFAC, UN and EU sanctions lists
Beneficial ownership and control tests
FATF recommendations
EMV 3-D Secure
Scheme fraud and chargeback monitoring programs

Risk work we have done

We built the fraud platform inside a US startup payment facilitator, rules and machine learning together, deciding in the authorization path in under 50 milliseconds.

Risk operators write rules in a language of their own and ship them without waiting for a deploy.

KYC, sanctions and OFAC screening ran as components of the same platform, covering onboarding and payouts inside the same flow.