Global shock. Local transmission. Auditable evidence.

Each product starts from a structural baseline. Live observations are normalized to roughly −1 to +1 and multiplied by product- and country-specific sensitivities. Positive contributions raise pressure; negative contributions cushion it.

1. Upstream fundamentals

Global commodity prices and official supply/demand revisions capture changes in cost and availability for oil, gas, coffee, cocoa, sugar, grains, feed and vegetable oils.

2. Physical stress

Weather, drought, crop condition and hydrology measure physical production or logistics stress. Higher-resolution official national sources are preferred where available.

3. Farm and wholesale transmission

Farm-gate, agricultural input and wholesale prices help identify whether an upstream shock is moving down the price chain.

4. Energy and logistics transmission

Energy and transport signals are applied where a product or market has a defensible exposure. Rhine pressure, for example, matters most for Rhine-connected fuel markets.

5. Retail and official confirmation

Consumer inflation and official outlooks are low-weight corroboration. They help confirm pass-through without replacing the model’s upstream causal signals.

6. Avoiding double counting

Transmission and confirmation signals receive smaller weights than core commodity, weather, crop and currency drivers. More sources are not automatically better when they measure the same shock.

7. Freshness and failure policy

Every signal has quality and age. Recent cached observations receive a penalty; older fallbacks become stale. If no defensible observation exists, its contribution is zero and confidence falls.

8. Country differences

The same global move can produce different scores because currency, weather, domestic production, logistics and pass-through differ by market. Language changes presentation; country changes the economic exposure.

9. Historical accountability

Every completed run stores country, product, score, confidence, timestamp, inputs, explanation and model version. Methodology changes never rewrite old forecasts, and alerts compare same-model history.

10. What this is not

The platform does not guarantee shelf prices, provide investment advice or claim to observe private retailer contracts, hedges, taxes, margins and inventories. Its value should be judged by its preserved forecasting record.

Source selection

The pipeline prefers official, machine-readable, reuse-compatible sources that add genuinely different information. Data licensing →