Client profile
ChainVault Exchange, a Singapore digital asset exchange with 220 staff.
Challenge and scope
Two analysts face about 11,000 alerts a month. With 89% classified as false positives in the baseline, routine reviews compete with escalated investigations.
Several tools can produce alerts for the same wallet or transaction. If reviewers handle each one independently, records lose the connection between a routine screening hit and activity that warrants investigation. The queue also needs a way to distinguish a duplicate signal from a genuinely separate event involving the same wallet.
Delivery workflow
Triage 11,200 monthly alerts and prepare evidence for 780 enhanced due diligence (EDD) escalations. Check wallet and transaction context in the approved tools and route decisions on reporting or wallet blocking to the exchange.
Associate related alerts with the triggering transaction and record the source, address, network, and evidence timestamp. Use the exchange’s approved categories for common false positives and separate cases that need deeper review. Give the investigator a short summary of the risk signal, reviewed evidence, and unresolved questions.
Review a sample of closed alerts and escalate disagreements to the AML lead. Calibrate the disposition rules against previously adjudicated records using the exchange’s approved reference set. Record any decision to change a rule or remove access with the relevant owner in the exchange’s systems.
Delivery team
Poland, about 15 FTE. 1 AML lead, 2 senior analysts and 12 alert triage associates.
Systems and tools
Chainalysis, Elliptic, ComplyAdvantage, internal case console.
Client decision rights
The exchange owns risk appetite, suspicious activity reporting, and wallet-blocking decisions.
Early milestone
The false-positive rate reaches 41% in week 4.
Outcomes
Company-reported results, with the baseline, measurement window and supporting record described for each measure.
| Measure | Baseline | Result | Window | Supporting record |
|---|---|---|---|---|
| False-positive rate | 89% | 8% | 5 months | Alert case export |
| Alert triage time | 26 min | 3 min | 5 months | Alert case export |
| EDD turnaround | 8 days | 2 days | 5 months | Alert case export |
| Confirmed missed true positives | 3 / quarter | 0 / quarter | 5 months | Internal audit record |
| Near-misses caught in QA | No baseline | 2 / quarter | 5 months | Internal audit record |
Interpreting the results
The decrease in false positives and shorter triage times depend on the alert population and rules used in each period. Record which events reached manual review and how repeated alerts were counted. Interpret missed positives and near-misses with their review sample and quarter, alongside the speed and volume of triage.
Recommendations for new clients
For exchanges reviewing more than 5,000 alerts a month, compare how false positives are defined in both periods. Sample cleared alerts and track confirmed misses separately from near-misses found in quality review.
Choose a sample spanning typical networks, repeated-wallet alerts, and cases that previously escalated. Provide the exchange’s decision policy and a set of reviewed examples. Make a senior reviewer available for disputed classifications, and inspect the resulting case records and escalation handoffs before expanding any automated routing.