NODUS
Geo-Intelligent Media Performance
🚗 National Auto Group
847 Active FSAs
LIVE
Demo Mode · MVP v2.0
⚙ Demo Customizer · Nodus internal use only
Audience Preset
Visible Tabs
🗺 Geo Intelligence
Province map · FSA scores · intent stages
⚡ Campaign Optimizer
Budget allocation · ROAS · reallocation log
🤖 Agent Console
7 agents · Redis bus · Insight Agent
🧠 Model Internals
Gating network · learner weights · scoring trace
Client / Prospect Name
Updates the header pill and all demo references
Display Options
Show demo hint banner
The blue script bar below the header
Show KPI bar
Daily spend · ROAS · bid count metrics
Shareable URL
nodus_demo.html
💡 Demo script: Geo Intelligence → click a province → Campaign Optimizer → click ⚡ Trigger Reallocation → Agent Console → Generate Insight & Ask Agent
Daily Spend
$4,700
↑ On target
Blended ROAS
3.6×
↑ +0.2 vs last week
Active FSAs
847
12 Scenario A · 835 Scenario B
Bid Adjustments Today
1,284
Across 6 channels · real-time
Reallocation Events
3
Auto · no human intervention
Canada Geo Score Map
1,600+ FSAs scored
YT NT NU NL BC AB SK MB ON QC NB NS PE
Low
High
Purchase Imminent
Consideration
Awareness
Intent Stage Distribution
847 FSAs
Scenario Distribution
Gating Network
A 12 FSAs — 3-Tier Stacking (DMS data)
B 835 FSAs — Proxy Ensemble
Next Gating eval: Mar 31
⬆ Graduation candidate: M5V → Scenario A
Top FSAs by Geo Score — National Auto Group
Live · Refreshed every 5 min
FSAMarketProvinceScenario Geo ScoreBid MultiplierIntent StageΔ 24h
Budget Allocation — Daily CAD
$4,700 / day
ROAS by Channel
Blended: 3.6×
Reallocation Event Log
3 events today
Channel KPI Summary
15-min window · live
ChannelBudgetROASCPMConv.Status
Message Bus — Redis Pub/Sub
Live stream
Insight & Reporting Agent
claude-sonnet-4-6
📊Weekly Performance Narrative
Click "Generate Insight" to run the Insight Agent via Claude API.
Gating Network — Dual-Track Scenario Router
12 Scenario A · 835 Scenario B
FSA Features
geo · demo · intent · proxy
Confidence
Scorer
Decision
Gate
▶
Scenario A
XGB + CatBoost + LSTM
Ridge
Meta-Learner
▶
Scenario B
K-Means + Proxy
Synthetic
Value Grader
FSA Score
bid multiplier [0.5–3.0]
Confidence Threshold (Scenario A gate): 0.65 ↑ Higher = fewer FSAs reach Scenario A
Live Routing Decisions
Scenario A — Base Learner Weights
Ridge Meta-Learner
Meta-Learner re-weights base models after each training cycle. Weights shift with data volume and feature drift.
XGBoost
38%
CatBoost
29%
LSTM
33%
Meta-Learner Output — FSA M5V
0.921
propensity score
Bid multiplier: 2.8× applied to Google Search
Cold Start — Scenario B Proxies
No Sales Data Mode
When a client has no geo sales data, the Synthetic Value Grader bootstraps FSA scoring from day one using these proxy signals.
📍 Store proximity index
82%
👥 CRM geo-density
64%
📊 Vertical benchmark (auto)
91%
🏷️ Pixel engagement (48h)
38%
🌐 Web traffic by region
55%
✅ Client uploaded: 47 store locations · 12,400 CRM records
⏳ Pixel live: 2 days — collecting engagement signal
📈 Expected Scenario A graduation: 4–6 weeks
FSA Data Maturity
B → A Graduation
FSAs graduate from Scenario B (proxy) to Scenario A (full ensemble) when conversion history ≥ 200 events. Threshold marker shown in orange.
🟢 Scenario A: 5 FSAs ⚪ Scenario B: 842 FSAs
Live FSA Scoring Pipeline — Step-by-Step Trace
Execution trace