52% versus 23%. That's the detection gap between automated and manual transaction monitoring — and it's not even the most uncomfortable number in AUSTRAC's recent supervision work.
52% of entities using automated transaction monitoring recorded at least one detection 23% of entities relying on manual monitoring did the same |
That gap gets quoted a lot. The more uncomfortable finding is about what happens after an alert exists at all.
Rules that don't know the customer
AUSTRAC's supervision teams keep finding monitoring rules that are too limited, too generic, or simply not linked to the entity's own ML/TF risk assessment — built once and left alone. Specific indicators get missed too:
- Third-party deposits
- Unusual size or frequency of transactions
- Activity inconsistent with a customer's declared profile
A monitoring program that can't point to why its rules exist — in terms of a specific, documented risk — is running on inherited defaults.
The isolation problem
A recurring pattern: alerts reviewed in isolation. A customer holds several accounts, and the review looks at one of them. A pattern obvious across the full relationship — linked accounts, a third party appearing repeatedly, a shift in how the customer uses the service — stays invisible because nobody looked at the whole picture.
DID YOU KNOW? 42% of entities hadn't reviewed their transaction monitoring rules in the previous twelve months. A rule that was right for last year's customer base isn't automatically right for this year's. |
A review nobody can actually audit
AUSTRAC also flagged a quality-assurance gap that's easy to have without noticing: checking whether a review happened, rather than whether it was any good. The businesses affected tend to share the same three gaps:
- No documented escalation points
- No defined timeframes for alert investigation
- No service levels to check performance against
The result: reviews that can't be reconstructed later. No rationale captured, no evidence trail, no way to tell a thorough review from a rubber stamp after the fact.
The number that should worry every compliance committee
Put those two problems together and you get AUSTRAC's most pointed finding: only a small fraction of the alerts many entities' systems generate are actually reviewed. Not reviewed poorly — not reviewed at all.
The bottom line: The fix isn't more alerts. It's fewer, better-targeted ones, reviewed properly, with the rules themselves revisited on a real schedule.
PRAXANA
