Cohort centroid
PAM scoring
Top-N candidates
Suppression filter
Per-cohort rules (e.g. responsible-gambling lists) Per-household opt-outs (GDPR right-to-erasure, dietary restrictions) Regulated-product flags (alcohol, tobacco, gambling, age-restricted)
Delivered recommendations

Loyalty clubs

Responsible-gambling self-exclusion lists, member-tier eligibility, venue access restrictions

Airlines

Alcohol promos for under-21 routes, dietary-restriction member flags, frequency-cap rules

Hotels & premium retail

Religious-observance preferences, opt-outs, GDPR right-to-erasure, regulated-product flags

Suppression is not where Kornerway adds intelligence — it is where Kornerway earns deployment trust. The client sets the rules. Kornerway enforces them.

Every recommendation passes through the suppression layer before it leaves the system, so non-compliant promotions are never output. Suppressed slots are automatically backfilled with the next-best compliant recommendation — the member still receives their full allocation, just without the restricted content.

Because every promotion has already been decomposed into taste dimensions, suppression can go deeper than category tags. A promotion tagged ‘sports’ but containing a dining component can still be caught for members who opted out of dining — the engine reads the content, not just the label.

In this POC dataset, the suppression scaffold exists but is empty — the synthetic data carries no contraindication metadata. In production, partner compliance teams populate the rule set via configuration.

Internal — Suppression scaffold

{
  "0": [],
  "1": [],
  "2": [],
  "3": [],
  "4": [],
  "5": [],
  "6": [],
  "7": []
}

Reference: pluggable rule pattern — labeling_rules/ architecture applies to suppression rules as well.

→ Back to Resonance overview