See USDA's next move months early
Prices move when USDA changes its numbers. Our nowcasts tell you which way those numbers are going, origin by origin, long before USDA gets there. Here is how often, how early, and what that adds to the information everyone already has.
Lead
We point the right way before USDA moves
In 82 crops since 2011, USDA changed a major exporter's production by 3% or more between its first number and its last. We were already on the right side of USDA in 63 of them, typically about 3 months before USDA had made half of the change. A free benchmark (the trend against USDA's number) gets 45.
Called in advance, by origin
Russia 2021: USDA started at 85 Mt
Our production was below USDA from mid-May; USDA cut in August. The final crop was 75.2 Mt.
More examples
| Crop | USDA's first number | Our call | USDA half way | Final |
|---|---|---|---|---|
| Russia 2021 | 85.0 Mt | 15 May: 80.0 Mt | 12 Aug | 75.2 Mt |
| Russia 2017 | 67.0 Mt | 01 Jun: 73.8 Mt | 10 Aug | 85.2 Mt |
| Ukraine 2023 | 16.5 Mt | 15 May: 21.3 Mt | 11 Aug | 23.0 Mt |
| Kazakhstan 2025 | 14.5 Mt | 15 Jun: 17.3 Mt | 14 Nov | 19.3 Mt |
| Canada 2021 | 32.0 Mt | 01 Jul: 30.4 Mt | 12 Aug | 22.4 Mt |
| United States (HRW) 2023 | 13.4 Mt | 15 May: 15.9 Mt | 11 Aug | 15.6 Mt |
Direction
Two calls in three on the right side of USDA
At every forecast date: is our production above or below USDA's number in force, and does USDA then move that way? The fair comparison is not a coin flip but the best fixed guess, such as "USDA always raises Russia". We are strongest where public information is thinnest: Kazakhstan, Russia, Ukraine and Australia. Argentina is our weak spot, and we show it.
| Origin | Seasons | Right on the rest of the season | Best fixed guess | Right on the next report | Best fixed guess |
|---|---|---|---|---|---|
| Kazakhstan | 15 | 83% | 54% | 84% | 58% |
| Russia | 14 | 81% | 63% | 74% | 57% |
| Canada | 13 | 78% | 75% | 64% | 59% |
| Ukraine | 13 | 77% | 81% | 81% | 50% |
| Australia | 14 | 65% | 51% | 75% | 55% |
| United States (HRW) | 14 | 66% | 62% | 65% | 56% |
| Argentina | 13 | 51% | 53% | 20% | 95% |
| All origins | 96 | 70% | 60% | 66% | 55% |
Value added
What it adds to what everyone already knows
Anyone can forecast USDA's revisions from free information: USDA's own habits, the long-run trend, its last move. We tested that, season by season, using only what was known at the time. Free information alone does no better than assuming USDA is right. Adding our nowcast cuts the error by 11% (between 7% and 15% across resampled seasons), with the largest gain from May to July, while USDA is still on its early numbers.
All origins, out of sample
| Forecast of USDA's coming revision | Average error (points) | Direction right |
|---|---|---|
| USDA's number is right | 14.0 | – |
| Free information (USDA's past habits, trend, last revision) | 14.8 | 61% |
| Free information + Wheat Nowcast | 13.2 | 66% |
By origin
Markets
From information to positions
USDA's numbers move the market
Report days are about 40% more volatile than other days, and the price moves against the supply revision: more wheat, lower price.
Being early is the edge
In our tests, the part of USDA's revisions we see coming still moved prices on report day: the market did not appear to have it before USDA published.
We claim trades only when they are proven
One trading rule passed our tests on history, so we sealed it before the season and score it live from 2027. Everything else we sell as information, not as a signal.
How we measure
No hindsight
Every number on this page is rebuilt as it would have stood on the day: the nowcast uses only data published by then, USDA's number is the one in force that day, and the final crop is used only to score. Forecasts of USDA's revisions are refitted each season on earlier seasons only. Results are from 2011-2025 history; since November 2026 every live forecast is time-stamped before the report it anticipates, on the track record.
Know where USDA is going
Two weeks of full Pro access, no commitment.