Screening Accuracy-Latency Tradeoff Index (SALTI)

GMI-PAY-014Payments & Transaction ProcessingResearch-Forward

Definition

Estimated additional detection yield that would result from relaxing the real-time screening latency constraint: quantifying how much screening accuracy is sacrificed for speed.

The Measurement Gap

The Measurement Gap

No metric tells you what screening accuracy was sacrificed to meet real-time latency requirements. Every real-time sanctions screening system operates under a latency constraint (typically measured in milliseconds) because the screening must complete within the authorization window. To meet this constraint, the screening system loosens its matching thresholds, reduces the depth of entity resolution, and truncates complex multi-hop lookups that would otherwise identify sanctioned entities operating through intermediary structures.

Transaction throughput is tracked. Screening latency is tracked. Both are standard operational metrics that SRE teams monitor continuously. But nobody measures what was missed because the matching threshold was loosened to hit the latency target. The tradeoff between speed and thoroughness is made implicitly in the screening system's configuration and is never quantified.

The gap is structurally different from ERCD (which measures coverage erosion over time). SALTI measures the accuracy floor imposed by architecture: the permanent gap between what the screening model could detect given unlimited time and what it actually detects within the production latency budget. This gap varies by entity complexity: simple name matches complete well within any latency budget, while complex multi-hop entity resolution (shell company chains, beneficial ownership traversal) is the first capability cut when latency budgets tighten. The result is that screening systems are most likely to miss the most sophisticated evasion techniques: precisely the ones that represent the greatest compliance risk.

The tradeoff also varies by transaction volume. During peak processing periods, latency budgets tighten further (more transactions competing for screening capacity), which means accuracy is at its lowest when transaction volume is at its highest.

Decision Supported

How much sanctions screening accuracy is being sacrificed for speed, and would architectural investment in faster screening infrastructure materially improve detection?

Formula

Plain Language

Run the same screening model against the same test set at two latency levels: the production latency threshold and an unlimited-time batch threshold. Compare the detection rates. The difference, expressed as a percentage of the unlimited-time detection rate, is the accuracy sacrificed for speed.

Notation

SALTI = (DetectionRate_unlimited − DetectionRate_production) / DetectionRate_unlimited × 100

Marginal: SALTI_marginal = ΔDetectionRate / ΔLatency

Variables

Variable Definition Unit
DetectionRate_production Detection rate at the current production latency threshold %
DetectionRate_unlimited Detection rate with no latency constraint (5-second batch threshold as proxy) %
SALTI_marginal Additional detection yield per additional millisecond of latency budget % per ms
ΔDetectionRate Change in detection rate between two latency levels Percentage points
ΔLatency Difference in latency budget between two levels Milliseconds

Proposed formula, pending validation pilots.

At a Glance

Unit
Percentage
Trend Direction
Lower; less accuracy sacrificed
Reporting Frequency
Semi-annually (with dedicated test runs)
Category
Compliance Screening Coverage & Tradeoffs
← Entity Resolution Coverage Drift (ERCD)Dispute Auto-Resolution Fairness Index (DARFI) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment