Transaction Monitoring
Automated screening of payments against rules and models to flag suspicious activity for human review.
Industry & DomainTransaction monitoring runs every payment through rules and models and raises an alert when something looks wrong: unusual amounts, new counterparties, structuring patterns, sanctioned jurisdictions. Almost all alerts are false positives — clear rates above 90% are normal — which makes the review queue, not the detection engine, the part that determines whether the system works.
That makes it a data table problem as much as a modelling one. The analyst needs the triggered rule, the customer's normal behaviour, the counterparty history and the decision trail on one screen, not across five tabs. Every tab is time, and time under volume pressure is what converts genuine review into reflex clearing.
Tuning is a product decision with a cost on each side
Loosen the rules and genuine cases pass; tighten them and analyst volume rises past what real review allows. Both failures are invisible in the dashboard, which will show a healthy clear rate either way.
In practice
A remittance provider added a rule for first-time corridors and alert volume tripled overnight. The team hired two analysts rather than tuning it. Six months later an audit found the clear rate had risen from 89% to 97% — the extra alerts were being cleared without meaningful review, so the rule had produced cost and no detection.
Where teams get it wrong
- Measuring the detection engine and not the review queue, where the decisions actually happen.
- Alert volumes above what the team can genuinely review, which produces reflex clearing.
- Evidence spread across tabs, so context costs time the analyst does not have.
- No audit trail of why an alert was cleared, which fails the next regulatory review.
- Treating a high clear rate as proof the system is accurate.
Learn more
You may ask
Frequently Asked Questions
Why do transaction monitoring systems produce so many false positives?
Rules are deliberately conservative, because missing a genuine case is a regulatory failure and clearing a false one is only a cost. Clear rates above 90% are normal, which is why review-queue design matters more than detection accuracy.
What makes a good alert review interface?
The triggered rule, the customer's baseline behaviour, the counterparty history and the prior decision trail on one screen — plus volumes matched to the attention each case needs.
Related terms
All terms- Anti-Money Laundering (AML)The controls a regulated business must run to detect and prevent money laundering — screening, monitoring, and reporting what they find.
- Data TableA grid of rows and columns for scanning, comparing and acting on records — the screen most B2B users spend their day inside.
- Human in the LoopKeeping a person in the decision path of an automated system — reviewing, approving or correcting before an action takes effect.
- KYC OnboardingThe identity verification a regulated product must complete before a customer can transact — and the step where most fintech signups are lost.
Defined by Mara Last reviewed .
Let's talk about your product.
Happy to look at what you're building and say where design would move the needle.
Contact Us