Detection engines
The engines are what put work into the alert queue. The Business Rules Engine evaluates configurable rules against transaction and customer data; the Analytical Engine applies statistical and behavioural models for deeper pattern detection.
Pattern Recognition
The Analytical Engine applies statistical and behavioural models to transaction and customer data to find patterns a single rule would not describe.
It works alongside the Business Rules Engine rather than instead of it: rules catch what you already know to look for, and the models are what surface a shape nobody has written a rule for yet.
Behavioural Analysis
Behaviour is assessed over time rather than per transaction. A baseline is built per customer, so a payment is judged against what that customer normally does instead of against a global threshold.
This is what lets monitoring detect unusual patterns across a sequence of transactions, including layering, where no single payment is remarkable on its own.
Profiling and Scoring
Each customer carries a risk score and a risk tier, recalculated as new information arrives from screening, monitoring and the customer record rather than on a review cycle.
The rules and events that moved a score are retained alongside it, which is the difference between a number and a number that can be explained to a supervisor.
Clustering and Modelling
In preparation
This engine is named in the platform and its section is reserved here so the anchor resolves, but there is no verified description of it to publish yet. It will be written from documentation rather than inferred.
Natural Language Processing
Language processing reads unstructured sources: news, media and public registers. Each document is scored for risk and sentiment from keyword frequency, sentiment analysis and topic relevance, and sources in multiple languages are supported.
Text analysis, search, classification and data extraction are proprietary methods, used together with a cloud-hosted large language model.
Evolutionary Learning
In preparation
This engine is named in the platform and its section is reserved here so the anchor resolves, but there is no verified description of it to publish yet. It will be written from documentation rather than inferred.
See the engines against your own data
A demo runs against a scenario you choose, so the alert volume and the match quality are yours rather than ours.
Request a demo