About the Client
An international company that builds, integrates and maintains banking software for financial institutions. Their customers are banks and payment companies, which meant our work had to meet the standards those banks apply to their own suppliers.
Challenges
Banks lose money to fraud in digital channels: stolen cards, hijacked accounts, mobile wallets used by somebody other than their owner. Rules written by hand catch yesterday’s patterns and annoy honest customers.
- Decisions had to be made in real time, while a transaction was still being authorised.
- The system had to run inside the client’s infrastructure, not in our cloud: a hard requirement for banking data.
- Risk analysts, not developers, had to be able to change the rules and the forms they work with.
- A false positive is also a failure: blocking a genuine customer costs the bank a relationship.
Our Approach
We built the product as a single-page web application over a server component installed in the client’s environment. It connects to the systems that already hold the signals, such as card authorisation, digital banking back ends and mobile wallets, and receives activity as it happens.
Each event is scored against behavioural patterns rather than a fixed rule list. When something looks wrong, the system acts immediately: it can decline a card authorisation or suspend a specific function in digital banking while leaving the rest of the customer’s access working.
The part the client valued most was the configuration layer. Analysts build and save their own forms and settings, so a new fraud pattern becomes a change in the interface rather than a release.
Results
- Fraudulent activity is detected and stopped while the transaction is still in flight.
- Losses from chargebacks and account takeover are reduced, because the block happens before the money moves.
- Honest customers keep working: the response is targeted at one function, not the whole account.
- New fraud patterns are configured by analysts in hours instead of waiting for a development cycle.
Technology Stack
React single-page application, server component deployed on-premise, integrations with card authorisation systems, digital banking back ends and mobile wallet platforms.
Business Impact
Fraud prevention is one of the few areas where software value is directly measurable: every blocked transaction is money not lost, and every false positive avoided is a customer not annoyed. Because the system runs inside the bank’s perimeter, it also met the data residency rules that rule out most hosted alternatives.