Glossary
Plain explanations of terms used in BLOOMCAST screens and articles.
| Term | Meaning |
|---|---|
| MAPE | Mean absolute percentage error: how far forecasts missed the actual price on average, in %. Lower is better. |
| No change | The simplest forecast: the price stays where it is. BLOOMCAST’s baseline. |
| Skill | 1 − model MAPE ÷ no-change MAPE. Above zero means better than the baseline. |
| Direction hit rate | Share of moves beyond ±3% where up or down was called correctly. |
| 80% forecast range | A range from past error distribution, built so roughly eight in ten actual prices fall inside. |
| Rolling backtest | Re-forecasting at many past dates with only the data known then, and scoring the result. |
| Demand calendar | Demand events such as graduations, Parents’ Day, kimchi-making season and Christmas, entered as upcoming and just-passed events. |
| Sentinel-2 L2A | Atmospherically corrected images from ESA’s Sentinel-2 optical satellites: 10 m resolution, about every 5 days. |
| Crop supply index | A BLOOMCAST indicator of green crop area in main growing regions, measured monthly from Sentinel-2. |
| ESA WorldCover | A global 10 m land-cover map, used to keep cropland only. |
| NDVI (MODIS MOD13Q1) | Vegetation index showing how green and dense plants are; 250 m, 16-day composite. |
| ERA5 | ECMWF reanalysis weather data, used for temperature, rainfall and sunlight anomalies. |
| UN Comtrade | UN international trade statistics. Import value divided by net weight gives the unit value. |
| World Bank Pink Sheet | Monthly commodity benchmark prices from the World Bank. |
| KAMIS | Korea Agricultural Marketing Information Service, providing wholesale prices including the Garak market. |
Get started
Tell us the items you trade.
We reply with a trial forecast for those items, shown next to the simple “price stays the same” baseline, so you can judge it on your own numbers.