Twice
the same customer investigated
Two teams open two cases on one person and neither knows.
By use case
One score, two disciplines
Fraud teams and AML teams investigate the same customer from opposite ends. Fusion puts both signal sets on one record — and one decision.
What you switch on
What FRAML needs switched on
Fusion is the point; the feeds either side of it are the inputs.
Why this matters here
Two teams, two tools, one customer, two answers.
Fraud sees behaviour and no list context. AML sees lists and no behaviour. Each closes a case the other would have escalated, and the disagreement only surfaces when a regulator reads both files.
Blind halves
signals never combined
Behaviour without list exposure, and lists without behaviour.
No shared record
so no shared decision
The escalation path forks and the audit trail forks with it.
1
Score per customer
2
Disciplines, one record
−44%
Duplicated investigations
100%
Signals traceable
Workflow · Two sides
Scroll to advance
The workflow, step by step
-
Step 01
One entity, both feeds
Transaction behaviour and screening results attach to the same resolved customer.
-
Step 02
Signals scored together
Velocity, device and network signals weigh alongside PEP status and list exposure.
-
Step 03
The disagreement is the finding
Where the two disciplines diverge, the case is raised rather than closed by whichever ran first.
-
Step 04
One case, one trail
Fraud and AML work the same record, and the export contains both perspectives.
complead / fusion / entity
- Entities1,241,336
- Feeds attached2
- Conflicts flagged18
complead / fusion / score
- Behaviour signals9
- List signals3
- Fusion score82 review
complead / fusion / conflict
- Divergent cases18
- Escalated11
- Cleared with reason7
complead / cases / #4102
- Contributors2 teams
- Evidence items16
- Exportcombined
Entity
Score
Conflict
Case
What you answer to
What supervisors expect from a combined model
Combining disciplines is encouraged. Combining them into something you cannot explain is not.
Read the compliance guides
Supervisor
Explainability of any combined score
Continuous
Regulator
Model governance and version history
Per change
Auditor
Decision log with contributing signals
Per case
FIU
Reporting unaffected by internal structure
On detection
Case study · Payments
Read the case study
Two teams stopped investigating the same person twice
Fraud and AML each ran their own queue on the same customer base. One shared entity record and one score removed the duplication and, more usefully, surfaced the cases where the two views disagreed.
−44%
Duplicated cases
18
Divergences found in month one
1
Shared queue
Cross-family links
Where teams like yours go next
FAQ
Before you ask us
Does one score hide the detail?
No. Each score decomposes into the contributing signals, their source and their weight — that is the point of it.
Do we need both modules?
Fusion needs at least two feeds to be worth anything. Most teams start with screening and add fraud signals.
Who owns the combined case?
Whichever team the escalation rules name, with the other able to contribute to the same record.
Will our auditor accept a combined model?
They accept models they can interrogate. Version, basis, decision log and override path are all retained.
Find the cases your two teams disagree about
Send a month of fraud and AML outcomes on the same customers. The divergences are usually the interesting part.
3,000+
Data sources checked
220+
Countries covered
15 min
Always real-time data