Decisioning engine

The component that says approve, decline, or review.

After all the checks have run, something has to look at the results and decide what happens to the customer. The decisioning engine is that something.

The runtime that evaluates signals, models, and policy to produce an auditable outcome per case. A signal is a single piece of evidence, a model is a trained system that scores risk, and a policy is the set of business rules that say what to do with the results. Auditable means the system records how it reached each decision.

A decisioning engine typically works like this:

1. It receives the results of the checks, such as document, face match, liveness, device, and data results.
2. It applies models and rules to those inputs.
3. It compares the outcome with the thresholds the business has set.
4. It returns a decision, usually approve, decline, or send for manual review, along with the reasons.

The engine usually combines several inputs:

• Model scores, such as a trust score that summarizes many signals.
• Rules, such as an explicit condition that always declines a certain case.
• Policy settings, such as how strict to be for a given product or region.
• Customer context, such as risk level or history.

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

• Rules engine applies explicit if-then logic, and it is often one part of the decisioning engine.
• Trust score is a number from a model, which the engine can use as an input.
• Orchestration runs the flow of checks, while the decisioning engine makes the call at the end.