Experiment Log - FAMILY_PROCESSING_DEMAND_SIGNALS

Hypothesis

French fries processing demand from BE/DE factories creates predictable pull-through effects on Dutch potato prices through fries-grade price signals and cross-border procurement patterns.

Data Sources (REAL DATA ONLY)

• Boerderij API: BE.157.2083 (Belgian fries), DE.157.2083 (German fries), NL prices
• CBS API: Trade statistics (if available)
• All using REAL international potato price data now accessible

Variants

• Variant A: Direct fries price signals (BE/DE lead-lag relationships)
• Variant B: Processing margin pressure (margin compression drives procurement)
• Variant C: Multi-market processing integration (combined signals)

Experimental Runs

Run 1: [Pending]

Date: TBD Variant: TBD Status: Not started

Configuration: - Target horizons: 30-day, 60-day - SESOI: 15% - Granger causality tests for lead-lag validation - Baseline models: persistent, seasonal_naive, ar2, historical_mean

Results: - Awaiting implementation

Verdicts

Overall Assessment: PENDING

Awaiting experimental validation.

HE Notes

• Created 2025-08-18
• Builds on FAMILY_CROSS_MARKET_COUPLING success (86-87% improvement)
• Focuses specifically on processing grade dynamics vs general cross-market
• Industry evidence: BE processors source 2.1M tons from NL (45% of needs)
• Academic support: Adjemian et al. (2016) 0.65 price transmission elasticity in potato-fries chain
• Uses REAL BE/DE fries price data from Boerderij API (438 BE records, 190 DE records)
• NO synthetic data - all from repository interfaces

Decision Log

[To be completed after experiments]

Codex Validation — 2025-11-10

Files Reviewed

• run_experiment.py
• hypothesis.yml
• experiment.md

Findings

1. Implementation exists but unexecuted. The runner outlines Boerderij BE/DE fries data ingestion, yet experiment.md still says “Awaiting experimental validation.”
2. Real-data usage unverified. Without a recorded run, we cannot confirm the BE/DE feeds were actually retrieved.
3. No baseline comparison. There are no MAE/DM statistics proving that processing-demand features beat the price-only baselines.

Verdict

NOT VALIDATED – This family remains untested until real data is processed end-to-end and the results demonstrate improvements over the mandatory baselines.