Lifecycle Pillar
Catch changes before they become problems
Monitor merchant web presence on policy schedules and surface material drift for analyst review. Websites change constantly. You need to notice before your regulator does.
How it works
Crawl merchant website
AURA navigates the merchant's website, captures full-page screenshots, extracts text, and records metadata across all discoverable pages.
Classify against rules
Content is classified using keyword rules first, then AI for ambiguous pages. Results are compared against the merchant's baseline profile.
Surface drift for review
Material changes are scored by severity and routed to the analyst queue. Minor changes are logged; category pivots trigger alerts.
Scans
Scan history and status
| Scan | Merchant | Status | Result |
|---|---|---|---|
| a3f8c1d2 | Warung Seribu | completed | Violation |
| 7b2e4d91 | Bunga Emas Digital | completed | Violation |
| e9d1f0ab | Toko Cahaya | completed | Clean |
| 1c4a8b33 | Dao Siam Crafts | completed | Clean |
| d5f2a7c8 | Lakan Bay Co | fetching | — |
| f8b3e619 | Pho Minh Trading | queued | — |
Category pivot detection
Each merchant starts with a baseline content profile from their initial scan. Subsequent scans compare current content against this baseline to identify meaningful shifts: a fashion retailer adding pharmaceutical products, a restaurant pivoting to gambling content, an e-commerce store introducing counterfeit goods. The drift score quantifies how far a merchant has moved from their approved profile.
Tiered content classification
Tier 1 applies deterministic keyword and regex rules for known BRAM violations: fast and zero cost. Pages without clear rule matches go to Tier 2 AI classification. Ambiguous cases go to Tier 3 deep analysis with full-page screenshots. Costs stay proportional to classification difficulty.
Multilingual rule sets
Rule sets support English, Bahasa Malaysia, and Chinese (simplified), the primary commercial languages across Malaysian and ASEAN merchant websites. Rules can be configured per merchant segment with different detection sensitivity for high-risk versus standard categories. Custom rules can be added at the organizational level.
Evidence-grade capture
Each scan produces timestamped screenshots, extracted text, classification results, and content hashes. These artifacts form the evidence chain if a violation is later investigated or reported to a card scheme.
Frequently asked questions
- Can monitoring run on different schedules by merchant segment?
- Yes. Teams can define weekly, monthly, and risk-triggered monitoring by portfolio policy. High-risk segments can be scanned daily while standard merchants are checked weekly or monthly.
- Is web drift enough to make final decisions?
- No. Drift is one signal layer alongside investigation workflows, reputation monitoring, and other evidence. Significant drift triggers an investigation, not automatic enforcement.
- What languages are supported for content classification?
- English, Bahasa Malaysia, and Chinese (simplified). The keyword rules and AI classifiers are trained on content in all three languages.
- How are false positives handled?
- Analysts can review, validate, or dismiss any classification result. Dismissed results are logged with rationale for audit purposes.