Performance and scalability

Contractual SLAs

The SLAs below are contractual values, continuously monitored in production. Deviations generate alerts for the Guardline operations team.

Operation	Latency (SLA)
Transaction evaluation in the engine	P95 of 300ms
Restrictive list lookup	< 100ms
Automatic case creation	< 500ms
Scoring calculation	< 300ms

The values represent the 95th percentile or the upper bound observed under normal operating conditions. Monitoring is continuous, and latency data is available to the operations team.

Horizontal scalability

The architecture allows horizontal scalability per component. Each part of the platform can be sized independently according to the institution’s demand:

• Decision engine: additional replicas can be provisioned to absorb peaks in transaction volume, keeping latency within the defined SLAs.
• Module backends: each module (ONP, FPP, CMP) scales independently. An institution with a high onboarding volume and a low case volume, for example, can have more ONP replicas without changing the sizing of the other modules.
• Database: support for read replicas, distributing query load without impacting write operations.
• Cache: support for clustering, allowing read load to be distributed and in-memory storage capacity to be increased.

Sizing is adjusted according to the institution’s operational volume. The platform does not impose fixed limits on transactions, onboardings, or cases; capacity follows the allocated infrastructure.

Monitoring

Performance metrics are collected continuously and compared against the contractual SLAs. When a metric approaches or exceeds the defined limit, alerts are generated automatically for the operations team, which then handles resizing or investigates the cause.