Back to case studies

Backend Engineer

Real-Time Fraud Risk Scoring Across Transaction Lifecycle

A sanitized view of integrating Accertify and Feedzai into transaction lifecycle risk scoring with merchant-configurable controls.

Details are intentionally generalized to respect confidentiality.

AccertifyFeedzaiRisk ScoringPoliciesRefund RiskFeedback Loops
Timeline
2025
Domain
Risk scoring, provider integration, merchant policy controls
Impact
Integrated fraud-risk providers across pre-transaction, post-transaction, and pre-refund stages.

Architecture

Sanitized System Flow

The diagram shows the major responsibility boundaries without exposing proprietary internal implementation.

Fraud Risk Pipeline

Integrated Accertify and Feedzai flows, risk policy controls, country/currency lists, lifecycle scoring, and feedback handling.

1Transaction Intent
2Policy Controls
3Risk Provider
4Decision
5Payment Flow
6Feedback Signals

Context

Fraud-risk systems sit on the critical path for some transactions and on operational feedback paths for others.

The architecture described here is generalized and avoids confidential provider or platform internals.

Problem

The platform needed risk checks before transactions, after transactions, and before refunds, with provider-specific integrations and merchant-configurable controls.

Risk decisions needed to influence flow behavior while remaining observable and explainable.

Constraints

Risk calls could not make the payment path fragile.

Provider outages needed safe handling and clear policy behavior.

Country and currency blacklist/whitelist controls had to remain merchant-configurable.

Refund and chargeback feedback needed to improve downstream operational insight.

My Role

Built backend integrations and lifecycle scoring paths for Accertify and Feedzai use cases.

Implemented merchant-configurable controls and feedback flows across transaction and refund scenarios.

Technical Design

Separated policy evaluation from provider integration so merchant controls and provider decisions could be reasoned about independently.

Supported pre-transaction, post-transaction, and pre-refund scoring with stage-specific request shape and response mapping.

Added feedback handling for failure outcomes, refund signals, and chargeback-related updates.

Made country and currency blacklist/whitelist behavior explicit in the risk decision path.

Tradeoffs

Running risk checks in more lifecycle stages improved coverage, but required careful latency and failure handling.

Merchant configurability increased surface area, but gave operators the control needed for real-world risk policies.

Impact

Expanded fraud-risk coverage across transaction and refund lifecycle stages.

Improved merchant control over risk posture through configurable country and currency policies.

What I learned

Risk integrations are only useful when their decisions are explainable and their failures are bounded.

Policy design matters as much as provider connectivity.