Capability 03 — Compliance
Monitor
A merchant is approved on the day it is read and changes the next. Monitoring is the part that keeps reading — surfaces, products, buyers and transactions — and records what it found in a form that survives being questioned.
Transaction monitoring
Threshold and pattern breaks raised as alerts for review. Alerts never auto-approve and never auto-clear.
Behavioral analytics
Velocity, reuse and clustering signals surfaced against a baseline, with the baseline shown alongside.
Alerts and investigations
Every alert resolves to the rule, the capture and the policy version behind it, so it can be examined.
Case management
Findings, remediation and rescan tracked to a deadline, with escalation past it automatic.
Dashboards and insights
Portfolio state as it is, including what is unconfigured, failing closed or awaiting approval.
What is actually watched
Monitoring that stops at the public homepage misses the case the rule exists for. Merchants routinely keep the storefront clean and put the prohibited claim on a blog, in an email campaign, or on an affiliate's page.
What a finding carries
A finding is only useful if it survives someone disagreeing with it six months later. Every one records the captured evidence and the machinery that produced it, so a disputed finding can be re-examined rather than argued about.
Findings have no close operation
There is no dismiss button, no severity override and no administrative resolve — not on the screen, and not held by any role including the ones that can halt a program.
A finding changes state because a rescan passed. The merchant submits what it changed, the same classifier re-reads the same surface, and a pass moves the state while a fail restarts the clock. The absence of the button is the control, and adding one for convenience would quietly undo the product.
Recording the work is not endorsing it
Remediation tracking records what a merchant says it did and what the rescan found. It never records an opinion that the remediation was adequate. That judgement belongs to the rescan result and to the institution, not to an operator sitting between them.
What monitor does not do
A control layer is only useful to an institution if its limits are as clear as its capabilities.
The classifier matches against wording the institution and its counsel wrote. It never invents the category it is matching against.
No role can close one. Only a passing rescan changes state.
Card-brand monitoring duties and supervisory expectations stay with the institution. This is instrumentation for meeting them.
See it running
Monitor runs in the control plane alongside the other three capabilities, on representative data.
Or start a pilot conversationThirty-three screens. Nothing in it is a real merchant or a real transaction.