Rules engine

Explicit if-this-then-that logic on top of the models.

A rule might say that any application from a certain country needs extra review, or that a failed liveness check always results in a decline.

A deterministic evaluator applying declarative conditions to enriched case data. Deterministic means the same inputs always produce the same result, declarative means the rules state what should happen instead of how to compute it, and enriched means the case data has been supplemented with signals from other checks.

A rules engine typically works like this:

1. It receives a case with its data and signals.
2. It checks each rule's conditions against the data.
3. Rules whose conditions are met fire and set an outcome or a flag.
4. The engine returns the combined result, often to a decisioning engine.

Rules usually take a simple shape:

• A condition, such as a failed liveness check or a document from a certain country.
• A threshold, such as a score below a set value.
• An action, such as decline, review, or request a stronger check.

Rules and models serve different strengths:

• Rules are fast, explainable, and easy to change when a new pattern appears.
• Models can find patterns that are too subtle or complex to write down as rules.

A rules engine sits next to terms that are easy to mix up:

• Decisioning engine is the larger component that can combine rules with model scores to reach an outcome.
• Velocity checks are a kind of rule that counts how often something happens.
• Orchestration coordinates the whole flow, in which rules are one part.