Product Ontology

Classify → describe → measure → infer → connect → validate

Level 1 · Taxonomy

How products are classified. IIH is the golden source.

Department→
Class→
Sub-Class→
Product type→
Product
Snacks→
Chocolate→
Chocolate bar→
Dark chocolate→
Lindt 85%
↓
Level 2a · Structured Attributes

Ground truth. No AI inference.

PIA · BrandBank · DQ
IdentityPhysicalCompositionSafetyNutritionPackagingSupplierAlcohol
Level 2b · Operational & Commercial

How the product performs in market.

Snowflake · analytics
Pricing · HTT, KVI, volatilityBehavioural · velocity, co-purchase, daypartOperations · MFC, lifecycle, OOS
Level 2c · AI Inferred

Semantic understanding. Probabilistic.

Claude LLM
FlavourOccasionsAudiencePositioning
↓ computed from levels 1–2
Level 3 · Relationships & Variants

Scored connections between products. 16 scoring components, 5 safety block types.

← scoring engine v3.2
Substitutes · Same PT, attribute-scoredComplements · Co-purchase, cross-catUpgrades · Premium, price floorSafety blocks · Allergen, dietary, ABV
↓
Level 4 · Three engines — consume the ontology

Substitution Engine

“What replaces this?”

Built · v3.2 · 16 components

Recommendation Engine

“What else would you like?”

Built · v1 · co-purchase + guardrails

Basket Completion Engine

“What’s missing?”

Future

↓
Level 5 · Governance
← auto-generated
Source trackingCompleteness scoringValidation flags

5 evaluation agents — validate engine output

Safety auditor
Allergens, dietary, age
Category expert
Customer acceptance
Commercial analyst
Price, brand fit
Mission judge
Basket coherence
Redundancy
Wasted slots