Methodology

Technical documentation on TradeStance's pricing models, data pipeline, and analytical framework.

Total Landed Cost Model

P_L = (P_F + C_L + C_S + C_I) x (1 + T) + D + F
P_LTotal Landed Cost — the all-in cost of goods delivered to destination
P_FFactory Price (FOB) — sourced from multi-tier hierarchical price discovery
C_LLogistics Cost — freight, port handling, inland transport
C_SStorage & Handling — warehousing, cold chain, fumigation
C_IInsurance — marine cargo, CIF uplift, war risk premium
TTax Rate — VAT, GST, or equivalent consumption tax
DDuties — MFN tariff, preferential rates, anti-dumping
FFees — brokerage, customs clearance, regulatory compliance

Multi-Tier Price Discovery

TradeStance employs a hierarchical waterfall strategy to find the most accurate price for any commodity:

1
Commodity Exchanges

Direct feeds from major global commodity exchanges (LME, ICE, CBOT) for real-time spot and futures pricing.

2
Government Statistics

Official government commodity boards and statistical offices providing reference prices and export data.

3
International Organizations

UN, IMF, and FAO datasets providing trade statistics, commodity price indices, and food price monitoring data.

4
Market Data Aggregators

Licensed data providers supplying financial market data, economic indicators, and commodity-specific pricing.

5
Verified Web Sources

Structured extraction from exchange-specific websites and regional commodity boards with multi-source cross-validation.

6
AI Consensus Engine

Proprietary AI model cross-references multiple sources to establish consensus pricing with statistical outlier rejection.

Data Guardian Pipeline

Every data point passes through a multi-layer validation gate before entering the system:

Floor Validation

Rejects prices below physically plausible minimums based on commodity-specific thresholds.

Ceiling Validation

Flags prices exceeding historical percentile bounds as potential errors or manipulation.

Staleness Detection

Purges entries older than configurable time windows from active price surfaces.

Source Reliability

Dynamic scoring of each data source based on accuracy history, freshness, and consistency.

Anomaly Forensics

Statistical analysis comparing new prices against rolling historical windows for deviation detection.

Consensus Gate

Multi-source agreement requirement: multiple sources must converge within tolerance for high-confidence pricing.

DIA-LSTM Forecasting Model

Architecture: Dual-stream Integrated LSTM with Attention
  |-- Temporal Stream
  |   |-- Input:  Price sequences with configurable lookback
  |   |-- Features: log_return, volatility, moving average convergence
  |   +-- LSTM with multi-head attention mechanism
  |-- Semantic Stream
  |   |-- Input:  News sentiment embeddings
  |   +-- Dense layers for feature compression
  +-- Fusion Layer
      |-- Concatenation of temporal and semantic outputs
      |-- Dense layers with dropout regularization
      +-- Output: Forward price forecast with confidence intervals
Sequence Length30 days
Attention HeadsMulti-head
Training MethodWalk-forward
Early StoppingPatience-based
Forecast Horizon30 days
Confidence95% CI
RegularizationDropout + L2
DeploymentCloud AI Platform

Data Sources (35+)

Commodity Exchanges

  • London Metal Exchange (LME)
  • ICE Futures
  • CBOT / CME Group
  • TuRIB (Turkey)
  • Agmarknet (India)

Government / Intl Orgs

  • UN Comtrade
  • FAO / FAOSTAT
  • IMF Commodity Prices
  • TurkStat / TMO
  • US Federal Reserve (FRED)

Market Data Providers

  • Financial Market APIs
  • Metal Price Services
  • Currency Exchange Data
  • Freight Rate Indices
  • News & Sentiment Feeds

Autonomous News Engine

The TradeStance news pipeline generates market intelligence at regular intervals without human intervention:

Scheduled Trigger
  |
  |-- 1. Ingest Raw Data
  |   |-- Commodity exchange prices
  |   |-- Government reference prices
  |   +-- Spot prices from major exchanges
  |
  |-- 2. News Aggregation
  |   |-- Global trade headline collection
  |   |-- Commodity-specific coverage filtering
  |   +-- Deduplication and relevance scoring
  |
  |-- 3. AI Processing
  |   |-- Generate concise Trade Briefs
  |   |-- Extract: country, commodity, sentiment
  |   +-- Professional financial reporting tone
  |
  |-- 4. Sentiment Scoring
  |   |-- NLP classification: BULLISH / NEUTRAL / BEARISH
  |   +-- Confidence-weighted impact scores
  |
  +-- 5. Storage & Distribution
      |-- Database persistence
      |-- Real-time broadcast to connected clients
      +-- System health report generation