Karina Su - Portfolio

Binance fraud detection

Interactive data visualization enabled intelligent tracing of suspicious transactions.

Risk analysts traced stolen funds through a dense table, one account at a time. I turned it into an interactive money-flow map, and daily case reviews doubled.

Role

Product designer

Timeline

3 months

Team

1 PM, 4 engineers

Analysts traced stolen funds by hand.

As crypto fraud increased, Binance’s risk team needed to trace suspicious money flows to investigate fraud and protect user trust. However, the existing investigation tool was data-heavy and fragmented.

Analysts had to manually search, compare, and connect transaction data across multiple modules, completing only 10 cases per 8-hour shift despite a daily workload of 15. The resulting backlog forced analysts to work overtime and delayed fraud investigations, putting customer trust at risk.

Rank the risk, then map the money.

I simplified the case-tracing flow in two parts.

A dashboard that ranks cases by risk. The old dashboard showed every metric, chart and row at once, and nothing was ranked, so analysts scanned hundreds of rows to find the few cases that mattered. Now the cases that need attention come first.

An interactive money-flow diagram. Before, analysts sorted the data by hand and struggled to see how transactions related. The diagram shows each transaction’s direction and counterparty with arrows, and clicking a counterparty expands the next layer, so analysts can see the whole picture in one place.

Twice the cases. Less fraud.

With the money trail on one map, each analyst’s daily case reviews doubled, from 10 to 20, beating the 1.5× goal and moving above the daily workload of 15. Fraud losses fell by 20% as analysts kept up with the queue, and customer NPS rose from 50 to 90 because people reporting fraud got answers sooner.

Overview