Evals for Tookitaki

Scenario-based detection coverage

Whether known typologies from the scenario repository are matched against transaction activity with correct triggering conditions.

Behavioral / unknown-pattern detection

Hybrid detection beyond static rules, covering patterns not expressed as an explicit scenario.

Simulation before deployment

Testing a scenario or threshold change against historical data and reporting expected alert impact prior to activation.

Detection explainability

Stating which scenario, thresholds, and transaction features drove an alert in terms an investigator can act on.

Multi-attribute entity matching

Matching on combined attributes rather than name string alone, and distinguishing true matches from coincidental overlap.

Language and script handling

Matching names transliterated or written in non-Latin scripts without losing true matches.

False-positive discrimination

Correctly clearing near-miss candidates that are not the listed party, with a stated reason.

Real-time payments screening decisions

Hold, release, or escalate decisions on in-flight payments with the basis for the decision recorded.

Rule-based score computation

Applying configured risk rules and indicator categories to produce a rating consistent with the stated policy.

Event-driven rescoring

Recalculating risk when transactions, profile changes, or periodic review events occur.

Score explainability

Producing a 'why this score' account that names the contributing indicators, including for compound rules.

Customer 360 context assembly

Pulling historical scores, events, and prior alert activity into a single view used for a decision.

Device and behavioral signal use

Incorporating device profiling and behavioral biometrics into a fraud decision.

Typology-based fraud detection

Recognizing patterns such as account takeover, mule activity, or multi-device access from typology definitions.

Fraud scenario segmentation

Applying different detection logic and thresholds per customer or product segment.

Contextual alert output

Returning alerts carrying enough context that a fraud analyst can triage without re-querying source systems.

Alert ranking and triage

Ordering alerts by materiality so high-risk items surface ahead of noise.

Alert-to-case consolidation

Grouping alerts from monitoring, screening, fraud, and scoring into a single customer-centric case.

Assignment, escalation, and roles

Routing cases to the right investigator and honoring role-based access and escalation paths.

Audit trail and SAR/STR support

Logging every action and producing the narrative and evidence needed for a regulatory filing.

Typology retrieval and interpretation

Finding the relevant typology for an observed pattern and stating its detection logic accurately.

Scenario authoring

Translating a described laundering or fraud pattern into a well-formed, testable scenario definition.

Scenario testing in the studio

Running a drafted scenario and reporting what it would and would not catch.

Currency of typology coverage

Reflecting updated or newly published typologies rather than stale definitions.