Brand Retail Knowledge Graph Schema

for a global multibrand retailer with LLM-powered content

0. Design Overview

Graph model: Property graph

Nodes have
label (type)
id (internal primary key)
Properties (scalars, lists, or JSON blobs)
Edges have
type (relationship name)
Direction (always defined)
Optional properties (weights, timestamps, etc.)

Core domains

Commercial & Product – brands, products, GTIN/GPC, internal categories, vendors, supply chain, geography
Brand Guidelines & Safety – visual/verbal/audio tokens, rules, risk domains, cultural context
Audience & Cultural Intelligence – topics, signals, trends, need states
Marketing & Campaigns – campaigns, themes, channel activations, calendar moments
AI Context & Feedback – context, model configs, content generation, metrics, feedback, optimisation
Customers & Archetypes – real customers, segments, synthetic Archetypes linked to real customers

Conventions

Node labels: UpperCamelCase (e.g. Brand, Product, Archetype)
Property names: snake_case (e.g. brand_type, created_at)
Edge types: UPPER_SNAKE_CASE (e.g. OWNS_BRAND, HAS_RULE)
Recommended on all nodes: id, created_at, updated_at

1. Node Types (Labels & Properties)

1.1 Brand & Product Domain

1.1.1 `Brand`

Represents master brand, private label, hosted brand, and co-brand.

Label: Brand

Property	Type	Description
`id`	string	Internal unique ID
`external_id`	string	Optional brand code (ERP / MDM)
`name`	string	Display name
`brand_type`	enum	`MASTER`, `OWNED`, `HOSTED`, `CO_BRAND`
`description`	string	Short description
`tagline_default`	string	Default global tagline
`is_active`	boolean	Active/inactive
`primary_language`	string	e.g. `en`, `fr`, `es`
`localized_names`	map/json	`lang_code -> name`
`localized_taglines`	map/json	`lang_code -> tagline`
`metadata`	map/json	Arbitrary brand metadata

1.1.2 `Vendor`

Suppliers / manufacturers / brand owners.

Label: Vendor

Property	Type	Description
`id`	string	Internal unique ID
`external_id`	string	Vendor code
`name`	string	Legal or trading name
`vendor_type`	enum	`MANUFACTURER`, `DISTRIBUTOR`, `OWNER`, …
`gln`	string	Optional GS1 Global Location Number
`country_code`	string	ISO-3166 country
`is_active`	boolean	Active/inactive
`metadata`	map/json	Extra attributes

1.1.3 `Product`

Individual sellable item (SKU/GTIN).

Label: Product

Property	Type	Description
`id`	string	Internal product ID (SKU ID)
`sku`	string	Internal SKU code (if distinct)
`gtin`	string	GTIN / EAN / UPC
`name`	string	Product name
`description`	string	Long description
`brand_name_on_pack`	string	Text as printed on pack
`status`	enum	`ACTIVE`, `DISCONTINUED`, `COMING_SOON`
`launch_date`	datetime	Optional
`attributes`	map/json	Key/value attributes (size, color, etc.)
`media_refs`	list	URIs / IDs of images, videos
`metadata`	map/json	Extra

1.1.4 `InternalCategory` (ICS)

Retailer’s internal category node.

Label: InternalCategory

Property	Type	Description
`id`	string	Internal category ID
`code`	string	Category code
`name`	string	Category name
`level`	string	e.g. `DEPARTMENT`, `CATEGORY`, `SUBCAT`
`path`	string	Denormalized path string
`is_leaf`	boolean	Leaf flag
`metadata`	map/json	Attributes (online_only, margin_band…)

1.1.5 `GpcCategory`

GS1 Global Product Classification node.

Label: GpcCategory

Property	Type	Description
`id`	string	Internal ID
`gpc_code`	string	GS1 GPC code
`name`	string	GPC category name
`level`	string	`SEGMENT`, `FAMILY`, `CLASS`, `BRICK`
`definition`	string	Official description
`metadata`	map/json	Extra

1.1.6 `BrandedCategory` (optional)

Brand-specific category naming.

Label: BrandedCategory

Property	Type	Description
`id`	string	ID
`name`	string	Brand’s own term
`description`	string	Optional
`metadata`	map/json	Other attributes

1.2 Operational & Geographic Domain

1.2.1 `SupplyChainNode`

Stores, warehouses, DCs, etc.

Label: SupplyChainNode

Property	Type	Description
`id`	string	Internal ID
`external_id`	string	Store/DC code
`name`	string	Name (“Store 1234 – London”)
`node_type`	enum	`STORE`, `WAREHOUSE`, `DC`, …
`gln`	string	Optional GLN
`address`	string	Free-form or structured
`lat`	float	Latitude
`lon`	float	Longitude
`is_active`	boolean	Active/inactive
`metadata`	map/json	Extra fields

1.2.2 `GeographicRegion`

Global → region → country → state → city etc.

Label: GeographicRegion

Property	Type	Description
`id`	string	Internal ID
`code`	string	e.g. `GLOBAL`, `EMEA`, `US`, `US-CA`, `LON`
`name`	string	Human-readable name
`region_level`	enum	`GLOBAL`, `REGION`, `COUNTRY`, `STATE`, `CITY`
`metadata`	map/json	Localization, currency, legal notes

1.2.3 `CustomerSegment`

Audience segments.

Label: CustomerSegment

Property	Type	Description
`id`	string	Segment ID
`name`	string	“Gen Z Value Seekers”
`description`	string	Short narrative
`criteria`	map/json	Targeting rules (age, behaviour…)
`metadata`	map/json	Extra

1.2.4 `Customer`

Real customers (pseudonymised).

Label: Customer

Property	Type	Description
`id`	string	Internal customer ID (pseudonymised)
`external_id`	string	Optional hashed ID from CRM/CDP
`origin_system`	string	Source system (`CDP`, `LOYALTY`, `ECOM`, …)
`is_active`	boolean	Active customer flag
`created_at`	datetime	First seen
`last_seen_at`	datetime	Last activity
`demographic_bucket`	map/json	Non-PII buckets (age_band, income_band, household_size, etc.)
`behaviour_profile`	map/json	Aggregated traits (e.g. “high frequency”, “deal seeker”, “luxury skew”)
`metadata`	map/json	Non-PII extras (e.g. loyalty_tier, channel_pref)

1.3 Archetypes (Synthetic Personas)

1.3.1 `Archetype`

Data-driven or curated archetypes representing groups of customers.

Label: Archetype

Property	Type	Description
`id`	string	Archetype ID
`name`	string	“Eco-Conscious Gen Z Foodie”, “Busy Value Parent”
`description`	string	Narrative description
`archetype_type`	enum	`SYNTHETIC`, `CURATED`, `HYBRID`
`generation_method`	string	`CLUSTERING`, `RULE_BASED`, `MANUAL`, `LLM_ASSISTED`, …
`embedding_ref`	string	Optional: ID for vector/embedding in feature store
`feature_profile`	map/json	Behavioural + attitudinal profile (e.g. price_sensitivity, channel_pref)
`language_codes`	list	Languages typically used (e.g. [`en-GB`,`es-ES`])
`is_active`	boolean	Whether archetype is currently in use
`metadata`	map/json	Additional fields (model_version, author, notes, etc.)

1.4 Brand Guidelines & Safety

1.4.1 `VisualIdentityToken`

Label: VisualIdentityToken

Property	Type	Description
`id`	string	ID
`name`	string	“Core Visual System – Brand X”
`primary_colors`	list	Hex codes or color names
`secondary_colors`	list	Optional
`logo_refs`	list	URIs/asset IDs
`imagery_style`	string	e.g. “bright, lifestyle photography”
`iconography`	string	Guidance text
`metadata`	map/json	Extra

1.4.2 `VerbalIdentityToken`

Label: VerbalIdentityToken

Property	Type	Description
`id`	string	ID
`name`	string	“Verbal Identity – Brand X Global”
`tone_of_voice`	string	e.g. “playful, supportive, concise”
`style_guidelines`	string	Prose rules for copy
`required_elements`	list	e.g. “clear CTA”, “benefit-led opening”
`approved_keywords`	list	Words/phrases to use
`prohibited_keywords`	list	Words/phrases to avoid
`example_snippets`	list	Short exemplar copy strings
`metadata`	map/json	Extra

1.4.3 `AudioIdentityToken`

Label: AudioIdentityToken

Property	Type	Description
`id`	string	ID
`name`	string	“Audio Identity – Brand X”
`audio_logo_ref`	string	URI/asset ID
`music_style`	string	e.g. “upbeat, electronic”
`voice_guidelines`	string	TTS / voiceover style notes
`metadata`	map/json	Extra

1.4.4 `ContentGuideline`

Must-do / must-not-do rule or general guideline.

Label: ContentGuideline

Property	Type	Description
`id`	string	ID
`rule_type`	enum	`MUST_DO`, `MUST_NOT_DO`, `ADVISORY`
`name`	string	“No profanity”, “Always include CTA”
`description`	string	Detailed rule text
`dimension`	string	`TONE`, `LEGAL`, `BRAND_SAFETY`, `FORMAT`, …
`severity`	integer	e.g. 1–5
`source`	string	`LEGAL`, `BRAND`, `EXPERIMENTAL`, …
`metadata`	map/json	Extra

1.4.5 `RiskDomain`

Label: RiskDomain

Property	Type	Description
`id`	string	ID
`name`	string	“Alcohol”, “Politics”, “Body Image”
`description`	string	What this risk covers
`severity`	integer	1–5
`regulatory`	boolean	True if regulatory-sensitive
`metadata`	map/json	Extra

1.4.6 `CulturalNorm`

Label: CulturalNorm

Property	Type	Description
`id`	string	ID
`name`	string	“Family-oriented”, “Price sensitivity”
`description`	string	Norm explanation
`valence`	enum	`POSITIVE`, `NEGATIVE`, `NEUTRAL`
`metadata`	map/json	Extra

1.4.7 `CulturalSignal`

Label: CulturalSignal

Property	Type	Description
`id`	string	ID
`name`	string	“Ramadan 2026 in MENA”, “Pride Month”
`signal_type`	enum	`HOLIDAY`, `MOVEMENT`, `EVENT`, `TABOO`, …
`description`	string	What it is, why it matters
`start_date`	datetime
`end_date`	datetime
`sensitivity`	enum	`LOW`, `MEDIUM`, `HIGH`
`metadata`	map/json	Extra

1.4.8 `CalendarMoment`

Retail/cultural calendar block (often drives campaigns).

Label: CalendarMoment

Property	Type	Description
`id`	string	ID
`name`	string	“Back to School 2026”, “Black Friday”
`moment_type`	enum	`RETAIL`, `CULTURAL`, `SPORT`, `RELIGIOUS`
`start_date`	datetime
`end_date`	datetime
`metadata`	map/json	Extra

1.5 Audience & Trends

1.5.1 `Topic`

Semantic concept (for search, social, etc.).

Label: Topic

Property	Type	Description
`id`	string	ID
`name`	string	“sustainability”, “gaming”
`description`	string	Optional
`aliases`	list	Synonyms / related labels
`metadata`	map/json	Extra

1.5.2 `AudienceSignal`

Raw observation: search volume, social mention, etc.

Label: AudienceSignal

Property	Type	Description
`id`	string	ID
`timestamp`	datetime	When observed
`channel`	string	`SEARCH`, `SOCIAL`, `SITE`, `CRM`, …
`metric_type`	string	`SEARCH_VOLUME`, `MENTION_COUNT`, `SENTIMENT_SCORE`, …
`value`	float	Numeric value
`unit`	string	`INDEX`, `%`, `COUNT`
`source_system`	string	Where it came from
`metadata`	map/json	Extra

1.5.3 `AudienceTrend`

Aggregated / interpreted trend.

Label: AudienceTrend

Property	Type	Description
`id`	string	ID
`name`	string	“Eco-conscious Gen Z”
`description`	string	Trend narrative
`start_date`	datetime	When first detected
`end_date`	datetime	Optional
`strength_score`	float	0–1
`confidence`	float	0–1
`source`	string	Analytics system / methodology
`metadata`	map/json	Extra

1.5.4 `CustomerNeedState`

Job-to-be-done / need.

Label: CustomerNeedState

Property	Type	Description
`id`	string	ID
`name`	string	“Quick mid-week dinner”
`description`	string	Need narrative
`intensity_score`	float	0–1
`metadata`	map/json	Extra

1.6 Marketing & Campaigns

1.6.1 `Campaign`

Label: Campaign

Property	Type	Description
`id`	string	Campaign ID
`name`	string	“Back to School 2026 – Brand X”
`description`	string	Summary
`campaign_type`	string	`BRAND`, `CATEGORY`, `SEASONAL`, etc.
`start_date`	datetime
`end_date`	datetime
`status`	enum	`PLANNED`, `LIVE`, `ENDED`
`budget`	float	Optional
`kpi`	map/json	e.g. target CTR, revenue, etc.
`metadata`	map/json	Extra

1.6.2 `ChannelActivation`

Per-channel execution.

Label: ChannelActivation

Property	Type	Description
`id`	string	ID
`channel`	string	`EMAIL`, `INSTAGRAM`, `TV`, etc.
`content_type`	string	`POST`, `AD`, `DISPLAY`, etc.
`scheduled_at`	datetime	When it goes live
`status`	string	`PLANNED`, `LIVE`, `ENDED`
`metadata`	map/json	Extra (placement, format…)

1.6.3 `ContentTheme`

Label: ContentTheme

Property	Type	Description
`id`	string	ID
`name`	string	“Eco-Friendly”, “Holiday”
`description`	string
`theme_type`	string	`SEASONAL`, `VALUE_PROP`, `EMOTIONAL`
`metadata`	map/json

1.7 AI Context & Feedback Loop

1.7.1 `Context`

Composite context for LLM (brand + channel + region + segment + archetype, etc.).

Label: Context

Property	Type	Description
`id`	string	ID
`name`	string	“Brand X – UK – Social – Archetype A”
`description`	string
`metadata`	map/json	Arbitrary context metadata

1.7.2 `ModelConfig`

LLM configuration used.

Label: ModelConfig

Property	Type	Description
`id`	string	ID
`model_name`	string	e.g. `gpt-5.1`
`model_version`	string	Optional
`temperature`	float
`top_p`	float
`max_tokens`	integer
`system_prompt_ref`	string	ID or URI for system prompt
`created_at`	datetime
`metadata`	map/json

1.7.3 `ContentGenerationEvent`

One generation call / batch.

Label: ContentGenerationEvent

Property	Type	Description
`id`	string	ID
`timestamp`	datetime	When generated
`request_id`	string	Request trace ID
`prompt_hash`	string	Hash of raw prompt
`input_type`	string	`PROMPT`, `TEMPLATE_FILL`, etc.
`metadata`	map/json	Traces, params, etc.

1.7.4 `ContentVariant`

A generated content variant.

Label: ContentVariant

Property	Type	Description
`id`	string	ID
`content_type`	string	`COPY`, `SUBJECT_LINE`, `CAPTION`, …
`language`	string	`en-GB`, `fr-FR`, etc.
`body_text`	string	Text content (or pointer)
`version_tag`	string	`A`, `B`, etc.
`is_published`	boolean	Whether used externally
`metadata`	map/json	Extra

1.7.5 `EngagementMetric`

KPI result for content, campaign, etc.

Label: EngagementMetric

Property	Type	Description
`id`	string	ID
`metric_type`	string	`CTR`, `CVR`, `ATC_RATE`, `VIEW_TIME`, …
`value`	float
`time_window`	string	e.g. `2025-12-01/2025-12-07`
`channel`	string	Optional
`region_code`	string	Optional
`segment_id_ref`	string	Optional
`metadata`	map/json

1.7.6 `UserFeedbackEvent`

Explicit feedback on content.

Label: UserFeedbackEvent

Property	Type	Description
`id`	string	ID
`timestamp`	datetime	When given
`feedback_type`	string	`THUMBS_UP`, `THUMBS_DOWN`, `RATING`, `COMMENT`, `ESCALATION`
`score`	float	Rating or +/-1 as needed
`comment`	string	Free text
`source`	string	`CONSUMER`, `AGENT`, `INTERNAL_REVIEW`
`metadata`	map/json

1.7.7 `AIQualityEvaluation`

Structured evaluation (human/automated).

Label: AIQualityEvaluation

Property	Type	Description
`id`	string	ID
`timestamp`	datetime
`evaluator_type`	string	`HUMAN_RATER`, `AUTO_CHECKER`
`scores`	map/json	e.g. `{brand_fit:0.9, clarity:0.8}`
`pass_fail_flags`	map/json	e.g. `{legal: true, tone: false}`
`metadata`	map/json

1.7.8 `OptimisationRecommendation`

Learning / suggestion node.

Label: OptimisationRecommendation

Property	Type	Description
`id`	string	ID
`timestamp`	datetime	When generated
`scope_type`	string	`BRAND`, `CATEGORY`, `CAMPAIGN`, `CONTEXT`, `ARCHETYPE`
`scope_ref`	string	ID-of-scope (if not inferred from edges)
`change_type`	string	`TONE_ADJUSTMENT`, `RULE_UPDATE`, `TEMPLATE_CHANGE`, `TARGETING_CHANGE`
`description`	string	Proposed change text
`confidence`	float	0–1
`metadata`	map/json

2. Relationship Types (Edges)

Notation: (SOURCE_LABEL)-[:EDGE_TYPE {props}]->(TARGET_LABEL)

2.1 Brand & Product Relationships

(Brand)-[:OWNS_BRAND]->(Brand)
Master brand → owned/private label
Props: since (datetime)
(Brand)-[:HOSTS_BRAND]->(Brand)
Retailer hosts external brand
(Brand)-[:BRAND_OWNER_IS]->(Vendor)
For hosted brands where vendor is corporate owner
(Brand)-[:CO_BRANDED_WITH]->(Brand)
Co-branding partnership (usually symmetric)
(Vendor)-[:SUPPLIES_PRODUCT]->(Product)
(Brand)-[:HAS_PRODUCT]->(Product)
(Product)-[:IN_INTERNAL_CATEGORY]->(InternalCategory)
Props: is_primary (boolean)
(Product)-[:HAS_GPC_CATEGORY]->(GpcCategory)
(InternalCategory)-[:CHILD_OF]->(InternalCategory)
(GpcCategory)-[:GPC_CHILD_OF]->(GpcCategory)
(InternalCategory)-[:MAPPED_TO_GPC]->(GpcCategory)
Props: confidence (float 0–1)
(BrandedCategory)-[:BELONGS_TO_BRAND]->(Brand)
(BrandedCategory)-[:MAPPED_TO_INTERNAL_CATEGORY]->(InternalCategory)
(BrandedCategory)-[:MAPPED_TO_GPC]->(GpcCategory)

2.2 Operational & Geographic

(SupplyChainNode)-[:LOCATED_IN]->(GeographicRegion)
(GeographicRegion)-[:REGION_CHILD_OF]->(GeographicRegion)
(CustomerSegment)-[:PRIMARY_REGION]->(GeographicRegion) (optional)
(CustomerSegment)-[:RELATED_TO_NEED_STATE]->(CustomerNeedState)
(Customer)-[:IN_SEGMENT]->(CustomerSegment)
Props: since (datetime), membership_source, confidence (float)

2.3 Archetypes & Customers

(Archetype)-[:BASED_ON_SEGMENT]->(CustomerSegment)
Props: weight (float 0–1)
(Archetype)-[:INFLUENCED_BY_TREND]->(AudienceTrend)
Props: impact_strength (float 0–1)
(Archetype)-[:FOCUSED_ON_NEED_STATE]->(CustomerNeedState)
(Archetype)-[:REPRESENTS_CUSTOMER]->(Customer)
Props:
- match_score (float 0–1)
- primary (boolean) – if this is the customer’s primary archetype
(Campaign)-[:TARGETS_ARCHETYPE]->(Archetype)
(Context)-[:CONTEXT_FOR_ARCHETYPE]->(Archetype)
(ContentGenerationEvent)-[:FOR_ARCHETYPE]->(Archetype)
(ContentVariant)-[:PERFORMS_FOR_ARCHETYPE]->(Archetype)
(EngagementMetric)-[:ATTRIBUTED_TO_ARCHETYPE]->(Archetype)
(UserFeedbackEvent)-[:FROM_ARCHETYPE_BUCKET]->(Archetype) (optional)
(OptimisationRecommendation)-[:APPLIES_TO_ARCHETYPE]->(Archetype)

2.4 Brand Guidelines & Safety

(Brand)-[:USES_VISUAL_IDENTITY]->(VisualIdentityToken)
(Brand)-[:USES_VERBAL_IDENTITY]->(VerbalIdentityToken)
(Brand)-[:USES_AUDIO_IDENTITY]->(AudioIdentityToken)
(Brand)-[:HAS_CONTENT_GUIDELINE]->(ContentGuideline)
Props: applicability (e.g. GLOBAL, REGION_SPECIFIC)
(InternalCategory)-[:HAS_CONTENT_GUIDELINE]->(ContentGuideline)
(GpcCategory)-[:HAS_CONTENT_GUIDELINE]->(ContentGuideline)
(RiskDomain)-[:GOVERNED_BY_RULE]->(ContentGuideline)
(CulturalSignal)-[:RELATES_TO_RISK_DOMAIN]->(RiskDomain)
(CulturalSignal)-[:EXPRESSES_NORM]->(CulturalNorm)
(CulturalSignal)-[:OCCURS_IN_REGION]->(GeographicRegion)
(CalendarMoment)-[:ASSOCIATED_WITH_CULTURAL_SIGNAL]->(CulturalSignal)
(Brand)-[:ALIGNS_WITH_SIGNAL]->(CulturalSignal)
(Brand)-[:MUST_AVOID_SIGNAL]->(CulturalSignal)
(Campaign)-[:ATTACHED_TO_CALENDAR_MOMENT]->(CalendarMoment)
(Campaign)-[:MUST_NAVIGATE_SIGNAL]->(CulturalSignal)

2.5 Audience & Trends

(AudienceSignal)-[:ABOUT_TOPIC]->(Topic)
(AudienceSignal)-[:OBSERVED_IN_SEGMENT]->(CustomerSegment)
(AudienceSignal)-[:OBSERVED_IN_REGION]->(GeographicRegion)
(AudienceTrend)-[:DERIVED_FROM_SIGNAL]->(AudienceSignal)
Props: weight (float)
(AudienceTrend)-[:RELEVANT_FOR_SEGMENT]->(CustomerSegment)
(AudienceTrend)-[:INFLUENCES_BRAND]->(Brand)
(AudienceTrend)-[:INFLUENCES_CATEGORY]->(InternalCategory)
(AudienceTrend)-[:INFLUENCES_GPC_CATEGORY]->(GpcCategory)
(AudienceTrend)-[:INFORMS_CAMPAIGN]->(Campaign)
(CustomerNeedState)-[:EXPRESSED_AS_TREND]->(AudienceTrend)

2.6 Marketing & Campaigns

(Brand)-[:HAS_CAMPAIGN]->(Campaign)
(Campaign)-[:PROMOTES_BRAND]->(Brand)
(Campaign)-[:PROMOTES_CATEGORY]->(InternalCategory)
(Campaign)-[:PROMOTES_GPC_CATEGORY]->(GpcCategory)
(Campaign)-[:TARGETS_SEGMENT]->(CustomerSegment)
(Campaign)-[:TARGETS_ARCHETYPE]->(Archetype)
(Campaign)-[:HAS_THEME]->(ContentTheme)
(Campaign)-[:RUNS_IN_REGION]->(GeographicRegion)
(Campaign)-[:HAS_ACTIVATION]->(ChannelActivation)
(ChannelActivation)-[:IN_REGION]->(GeographicRegion)
(ChannelActivation)-[:USES_THEME]->(ContentTheme) (optional)

2.7 AI Context & Feedback

2.7.1 Context wiring

(Context)-[:CONTEXT_FOR_BRAND]->(Brand)
(Context)-[:CONTEXT_FOR_REGION]->(GeographicRegion)
(Context)-[:CONTEXT_FOR_SEGMENT]->(CustomerSegment)
(Context)-[:CONTEXT_FOR_ARCHETYPE]->(Archetype)
(Context)-[:CONTEXT_FOR_CAMPAIGN]->(Campaign) (optional)
(Context)-[:USES_VERBAL_IDENTITY]->(VerbalIdentityToken)
(Context)-[:USES_VISUAL_IDENTITY]->(VisualIdentityToken)
(Context)-[:USES_AUDIO_IDENTITY]->(AudioIdentityToken)
(ContentGuideline)-[:OVERRIDES_IN_CONTEXT]->(Context)

2.7.2 Generation & evaluation

(ContentGenerationEvent)-[:USES_MODEL_CONFIG]->(ModelConfig)
(ContentGenerationEvent)-[:FOR_BRAND]->(Brand)
(ContentGenerationEvent)-[:FOR_CAMPAIGN]->(Campaign)
(ContentGenerationEvent)-[:FOR_CATEGORY]->(InternalCategory)
(ContentGenerationEvent)-[:FOR_SEGMENT]->(CustomerSegment)
(ContentGenerationEvent)-[:FOR_ARCHETYPE]->(Archetype)
(ContentGenerationEvent)-[:FOR_REGION]->(GeographicRegion)
(ContentGenerationEvent)-[:USES_CONTEXT]->(Context)
(ContentGenerationEvent)-[:PRODUCED_VARIANT]->(ContentVariant)
(ContentVariant)-[:CONFORMS_TO_RULE]->(ContentGuideline)
(ContentVariant)-[:VIOLATES_RULE]->(ContentGuideline)
(AIQualityEvaluation)-[:EVALUATES_VARIANT]->(ContentVariant)
(AIQualityEvaluation)-[:CHECKS_RULE]->(ContentGuideline)
(UserFeedbackEvent)-[:FEEDBACK_ON_VARIANT]->(ContentVariant)
(EngagementMetric)-[:MEASURES_VARIANT]->(ContentVariant)
(EngagementMetric)-[:ROLLS_UP_TO_CAMPAIGN]->(Campaign)

2.7.3 Learning loop

(EngagementMetric)-[:UPDATES_TREND]->(AudienceTrend)
(UserFeedbackEvent)-[:SUPPORTS_TREND_QUALITATIVELY]->(AudienceTrend)
(AIQualityEvaluation)-[:GENERATES_RECOMMENDATION]->(OptimisationRecommendation)
(EngagementMetric)-[:INSPIRES_RECOMMENDATION]->(OptimisationRecommendation)
(OptimisationRecommendation)-[:APPLIES_TO_BRAND]->(Brand)
(OptimisationRecommendation)-[:APPLIES_TO_CATEGORY]->(InternalCategory)
(OptimisationRecommendation)-[:APPLIES_TO_CAMPAIGN]->(Campaign)
(OptimisationRecommendation)-[:APPLIES_TO_CONTEXT]->(Context)
(OptimisationRecommendation)-[:APPLIES_TO_ARCHETYPE]->(Archetype)
(OptimisationRecommendation)-[:PROPOSES_RULE_UPDATE]->(ContentGuideline)
(OptimisationRecommendation)-[:PROPOSES_TOKEN_UPDATE]->(VerbalIdentityToken)
(ContentGuideline)-[:REFERENCED_BY_MODEL_CONFIG]->(ModelConfig)
(VerbalIdentityToken)-[:REFERENCED_BY_MODEL_CONFIG]->(ModelConfig)

3. Implementation Notes (for Data Engineers / KG Modellers)

Platform
Any property graph DB (Neo4j, JanusGraph, Neptune, TigerGraph, etc.) will work.
Use constraints / indexes on IDs and key business keys (GTIN, category codes, region codes).
Multi-language
Either store language variants in *_localized maps, or create a LocalizedText pattern if preferred.
Versioning
For guidelines and tokens, either version via properties (version, valid_from, valid_to) or via new nodes linked by PREVIOUS_VERSION_OF.
Customer privacy
Only pseudonymised IDs and bucketed attributes in Customer.
No raw PII (email, name, phone, exact address) in the graph.
Scale considerations
Archetype→Customer edges can be limited to primary archetype and/or sampled customers.
Detailed memberships can live in warehouse / feature store and be projected into the graph selectively.
LLM integration path
At generation time, the application resolves:
- Customer → Archetype → Segment → Brand / Category / Region / Trends / Guidelines / Context
- Then it assembles a prompt from those nodes (tokens, rules, examples) and logs a ContentGenerationEvent.

Brand Retail Knowledge Graph Schema

0. Design Overview

1. Node Types (Labels & Properties)

1.1 Brand & Product Domain

1.1.1 Brand

1.1.2 Vendor

1.1.3 Product

1.1.4 InternalCategory (ICS)

1.1.5 GpcCategory

1.1.6 BrandedCategory (optional)

1.2 Operational & Geographic Domain

1.2.1 SupplyChainNode

1.2.2 GeographicRegion

1.2.3 CustomerSegment

1.2.4 Customer

1.3 Archetypes (Synthetic Personas)

1.3.1 Archetype

1.4 Brand Guidelines & Safety

1.4.1 VisualIdentityToken

1.4.2 VerbalIdentityToken

1.4.3 AudioIdentityToken

1.4.4 ContentGuideline

1.4.5 RiskDomain

1.4.6 CulturalNorm

1.4.7 CulturalSignal

1.4.8 CalendarMoment