Snapshot
Free edition: live nowcasts and satellite data are shown with a 14-day delay (data to 07 Nov 2026).Free edition · 14-day delay Real-time access → ·
Glossary

The words on this site, in plain language

Every number we publish, explained for a grain buyer, a miller or a trader: what it means and why it matters to you.

Glossary

Forecasts

Nowcast

Our estimate of this season's final yield, updated as the season unfolds with the weather and satellite data available up to that day. It is a forecast of a number that already exists in the fields but that nobody has measured yet.

Why it matters: It tells you early where the crop is heading, weeks or months before the official numbers.

Yield: t/ha and bu/ac

How much grain a hectare (or an acre) produces. Outside the US we use tonnes per hectare (t/ha); for the US, bushels per acre (bu/ac), as USDA does. For wheat, 1 bu/ac = 0.0673 t/ha and 1 t/ha = 14.9 bu/ac: 50 bu/ac is about 3.36 t/ha.

Why it matters: Yield times area gives production, the tonnes that will reach the market.

Trend

The yield you would expect in a normal year: a straight line through past yields, which rise slowly with better seeds and farming. It uses only the seasons before the one being forecast.

Why it matters: It is the simplest benchmark. A forecast is worth paying for only if it beats the trend.

p50 (median)

The middle of our forecast: we think the final yield is as likely to come in above it as below it. It is the single number to use if you need only one.

Why it matters: Use it as your central case in a balance sheet or a price view.

80% interval (p10–p90)

The range we expect the final yield to fall in 8 times out of 10. p10 is the low end (1 chance in 10 of a lower yield), p90 the high end (1 in 10 of a higher one). A wide range means the season is still open.

Why it matters: It tells you how much to trust the p50 and gives you a ready-made bad case and good case.

Glossary

Accuracy

RMSE

Root mean square error: the typical size of a forecast's miss, in the same unit as the yield, with large misses counting more. An RMSE of 0.20 t/ha means the forecast was usually off by about 0.2 t/ha. Lower is better. We also show it as a % of the final yield, to compare origins.

Why it matters: Compare our RMSE with the trend's on the same days: the gap is what the model adds.

80% coverage

How often the final yield actually fell inside our 80% interval in the backtest. Close to 80% means the ranges are honest; well below means they were too narrow.

Why it matters: It checks that the bad case and good case we give you are real, not decoration.

Skill

How much smaller our error is than a simple benchmark's (the trend for yields, a no-change forecast for prices), in percent. +20% means 20% less error; a negative skill means the benchmark did better.

Why it matters: One number that says whether the forecast earns its keep.

Backtest

Running the model on past seasons as if we were living them, using only the data that existed on each day. It shows how the method would have done, season by season.

Why it matters: It is the evidence behind every accuracy number on the site.

Point in time and the ledger

Point in time means a forecast uses only what was known on its date, never later revisions. Each live forecast is written to a public ledger with a time stamp and a hash, a digital fingerprint that changes if anyone edits the entry afterwards.

Why it matters: You can check that we did not rewrite our forecasts after the fact.

Glossary

USDA and the market

WASDE

USDA's World Agricultural Supply and Demand Estimates, published around the 10th of each month: production, use, trade and stocks for every major country. They are the market's reference numbers and move prices on release day.

Why it matters: Our notes and alerts say before each report where we expect USDA to move.

USDA revision

A change USDA makes to its own estimate from one monthly report to the next. Our “USDA revision expected” alert fires when our nowcast is further from USDA's number than USDA usually moves in a month.

Why it matters: Revisions move prices; knowing their likely direction in advance is the edge.

Marketing year

The 12-month year in which a harvest is sold and used, starting at harvest: June–May for US wheat, July–June for the EU and the Black Sea. USDA balances and stocks are counted by marketing year.

Why it matters: When you read a balance, check which marketing year it is: two can overlap in one calendar year.

Stocks-to-use

Stocks left at the end of the marketing year divided by the year's use, in percent. A low ratio means a tight market with little buffer. Since 1995 wheat prices have moved roughly one for one, in percent, in the opposite direction to exporters' stocks-to-use.

Why it matters: It is the link from a crop number to a price: a smaller crop matters most when stocks are already low.

Cash price and futures

A cash price is what physical grain trades for at a named place, like Rouen or Bologna, in €/t. A futures price is for a standard contract on an exchange, such as Euronext milling wheat in €/t or Chicago wheat in US cents per bushel; 1 tonne of wheat is 36.74 bushels.

Why it matters: The gap between the two, the basis, is where local supply and demand show up.

Wheat classes

Wheat is not one product. Hard red winter (HRW) and spring wheat are high-protein bread wheats; soft red winter (SRW) and EU soft wheat go into biscuits, feed and blends; durum is the hard wheat for pasta, priced on its own.

Why it matters: A crop problem in one class can leave another untouched, and their prices can diverge.

Glossary

Satellite and alerts

VHI (Vegetation Health Index)

A weekly score from 0 to 100, from NOAA satellites, of how green and unstressed the vegetation is; it combines greenness and canopy temperature. Below about 40 means stress (drought or heat); above 60, good conditions.

Why it matters: It sees the whole growing area every week, long before any field survey.

Anomaly

The difference between this week's reading and the normal for the same week in past seasons. An anomaly of −10 means 10 points worse than usual for the time of year.

Why it matters: A VHI of 45 can be fine in July and bad in April: the anomaly takes the season out.

Percentile

Where a value ranks among past values for the same week: the 5th percentile means only 5% of earlier seasons were lower. Our alerts fire on rare percentiles, not on fixed levels.

Why it matters: It tells you how unusual a reading is, in plain odds.

Alert strength: notable, strong, major

Notable: rare for the time of year, about 1 season in 5 to 10. Strong: rarer still, about 1 in 20. Major, for USDA alerts: our gap with USDA is larger than 9 of USDA's monthly moves out of 10, and USDA's number sits outside our 80% interval.

Why it matters: Short of time? Read the strong and major alerts first.

14-day delay

The free site shows the same forecasts and alerts 14 days late. Pro shows them the day they are made, with downloads, alerts by e-mail and the assistant.

Why it matters: Markets price news within days: for decisions you need the live numbers.

Missing a word? Write to [email protected].

↑ ↓ to choose · Enter to open · Esc to close