Skip to main content

8-Scenario Mapping

Published 2026-05-14Updated 2026-06-307 min read

The 8 core scenarios most directly tied to marketing innovation in this data-silo + external-signal fusion PoC. Every scenario specifies a data mix combining first-party data with the 4 external types (social · weather · economy · competitor).

1. 8-Scenario Overview​

#ScenarioFirst-Party DataExternal DataURLPrimary Persona
S1Natural-language semantic searchSKU · Member · First-party reviewsOlive Young · X reviews (social)/semantic-searchP1, P5
S25-department persona chatbotAll first-party dataAll 4 external types (tool autonomy)/chatAll
S3Category · BU insight cardsOrderTransaction · ChannelSellThroughSearch trend + Weather + Competitor/insightsP3
S4Lifestyle persona matching + clusteringMember RFM · Category affinitySocial personas (Instagram · Olive Young)/personasP2
S5Omnichannel campaign ROAS simulationPast campaigns · Channel sell-throughSearch trends · SNS responses/campaign-roasP1, P5
S6External signal fusionFirst-party GMV (BU · Category · Store)All 4 types (trend · weather · economy · competitor)/signalsP2, P3
S7Omnichannel member journeyOwned mall · CartEvent · SNS ad responses · channel transactions(none, prioritize first-party data integration)/journeyP4
S8Marketing consent · PII · minor guardrailsMember consent · Compliance(none, Bedrock Guardrails)/complianceP4

2. 14 Patterns → Compressed Mapping to 8 Scenarios​

Among the 14-scenario pool that frequently appears in ontology + Agentic AI PoC patterns, the 8 most directly relevant to LG H&H marketing innovation have been selected and redefined.

Original PatternLG H&H MappingNotes
Hybrid searchS1 Natural-language semantic searchStrengthened by combining external reviews
Persona chatbotS2 5-department persona chatbotAs-is
Insight cardsS3 Category · BU insight cardsStrengthened by combining 4 external types
Persona matchingS4 (merged)Merged with clustering
ClusteringS4 (merged)Merged with persona matching
Look-alike expansion(excluded)Deferred to PoC v2
Campaign ROIS5 Omnichannel campaign ROAS simulationAs-is
Regional competition map(excluded)Deferred to PoC v2
Terms · GuardrailsS8 Marketing consent · PII · minor guardrailsAdded minor cosmetics guard
External signal fusionS6 External signal fusionExpanded to 4 types
Anomaly detection(excluded)PoC v2
Payment · Membership matrix(excluded)PoC v2
Unified customer journeyS7 Omnichannel member journeyEmphasizes behavior crossing BUs
Weather × SalesS6 integratedAs a subset of external signals

3. Per-Scenario Data Mix Details​

S1 — Natural-Language Semantic Search​

DataSourceUse
First-party SKUOpenSearch idx_productBM25 + KNN
Member profileOpenSearch idx_customerKNN
First-party reviewsOpenSearch idx_reviewBM25 + KNN
Olive Young reviewsExternal crawl → idx_reviewKNN (social SOV)
X / Instagram reviewsExternal → idx_social_trendKNN
KG 1-hopNeptuneCustomer × Product × Brand × BU

S2 — 5-Department Persona Chatbot​

12 tools invoked autonomously (see S2 design spec). All scenarios' tools are unified-invoked through this chatbot.

S3 — Category · BU Insight Cards​

DataSourceUse
GMV (BU × Category × Month)Snowflake / Neptuneline/bar chart
Search trendsNaver · Google Trends → SocialSignalDual axis
WeatherKMA → WeatherSignalDual axis
Competitor new productsPublic appearances → CompetitorSignalannotation

S4 — Persona Matching + Clustering​

DataSourceUse
RFM featuresNeptuneKMeans input
Category affinityNeptune (Customer × Category frequency)KMeans input
Social personasInstagram · Olive Young review keywords → SocialSignalLLM labeling reinforcement

S5 — Omnichannel Campaign ROAS​

DataSourceUse
Past campaignsSnowflake CAMPAIGN_MARTBayesian prior
Channel sell-throughChannelSellThroughPer-channel response rate
SNS responsesX · Instagram ad clicks · responses → SocialSignalPosterior adjustment
Search trendsNaver · Google → SocialSignalSeparate non-campaign effects

S6 — External Signal Fusion​

DataSourceUse
First-party GMVSnowflake / Neptuney-axis
Social trendsSocialSignalx-axis / lag analysis
WeatherWeatherSignalx-axis (temp · precipitation)
EconomyEconomicSignal (FX · CPI)x-axis
CompetitorCompetitorSignalannotation / comparison

S7 — Omnichannel Member Journey​

DataSourceUse
Owned mall SearchEvent / CartEvent / OrderTransactionNeptuneTimeline
Touchpoint (SMS/KakaoTalk/SNS ad)NeptuneTimeline
ChannelSellThrough (Mart/H&B/Convenience)NeptuneTimeline
Coupon / ReviewRatingNeptuneTimeline

S8 — Guardrails​

DataSourceUse
Bedrock GuardrailsAWSInput · output guards
Member consentCompliance · Membership.opted_in_marketingBlock send tools
Estimated minorCustomer.age_bandBlock cosmetics campaigns

4. Persona × Scenario Matrix​

DepartmentS1S2S3S4S5S6S7S8
P1 Brand Marketerstarstarstar
P2 Insightsstarstarstar
P3 D&A · MarTechstarstarstar
P4 CRM · LG Membersstarstarstar
P5 MD · Channel Salesstarstarstar

5. Top-5 Demo Path (30 minutes)​

[S1] Natural-language semantic search — 4 min
   "Women in 30s with sensitive scalp + organic shampoo in 20K KRW range + Olive Young rating 4.5+"
   → SKU(Elastine) + Member + first-party / Olive Young reviews simultaneously on cards

[S2] Persona chatbot — 5 min (switcher demo)
   "Top 5 channels where Perioe sales dropped this month and the cause?"
   → Answer in P5(MD) → switch to P1(Brand Marketer) → change in answer tone

[S3] Insight cards — 4 min
   "Beauty BU monthly GMV + search trends + weather combined card"
   → matplotlib PNG + LLM comments (per persona)

[S5] Campaign ROAS simulation — 4 min
   "Next week Su:m37 new product launch / VIP 50K / 100M KRW"
   → Bayesian channel mix + ROAS distribution

[S7] Omnichannel member journey — 3 min
   "Member cust_001234 90-day journey"
   → Owned mall → SNS → Olive Young → Mart → Repurchase single timeline

6. PoC Schedule (8 weeks)​

WeekWorkScenarios
1-2Data ingestion (first-party N=500~5K + synthetic 49.5K + 4 external types)Infrastructure
3-4S1 + S2 + S3 implementationS1, S2, S3
5-6S5 + S7 + persona switcherS5, S7, P1~P5
7S4 + S6 + S8S4, S6, S8
8Demo rehearsalAll