Every Thursday morning, USDA drops the Weekly Export Sales report and a number lights up the ag desks: net new sales of soybeans, corn, cotton. A big print feels bullish, a soft one feels bearish, and it's tempting to read the running total as a pace — are we ahead of USDA's forecast, or behind? Here's the uncomfortable part, drawn from seven crop years across soybeans, corn and cotton: you can't extrapolate that pace to beat WASDE, and the weekly number misleads exactly when it looks most exciting. This is why — and what the report is actually good for.

The Thursday misread The Setup

The instinct is reasonable. A weekly sales number is fresh, granular, and moves the market on release — surely a run of hot prints means demand is outstripping the forecast. The trouble is that the headline figure isn't a physical flow and it isn't a forecast. It's a book of forward contracts, and forward books behave nothing like the steady fill toward a season total that the "pace" mental model assumes. They front-load, they can be cancelled, and they routinely blow past the number they're supposedly racing toward. Read literally, the weekly print tells you less about where the year ends up than the monthly WASDE already does — a claim we'll put to a hard test below.

Two numbers wearing one name The Anatomy

The Export Sales report actually carries two different series, and conflating them is the root of the misread:

Commitments are promises. They're lumpy, they get made months ahead of shipment, and they can shrink — cancellations, washouts, contracts rolled into the next crop year. Shipments are physical: near-monotonic, and the thing WASDE is actually forecasting. The weekly headline everyone reacts to is the commitments line — a different animal from the physical flow that actually converges to the forecast.

Front-loaded, and prone to overshoot The Overshoot

Plot both series as a share of the eventual full-year total and the misread becomes obvious. Soybeans are the clearest case — buyers, China chief among them, lock in supply in the autumn:

Two facts fall out, and both trip up the pace reading:

• Commitments are front-loaded, not linear. Across the last seven soybean years, half to 85% of the entire marketing year — a median near 70% — was already on the books by early December. So when the cumulative-sales line flattens in spring, that isn't demand dying or a miss brewing. There was simply nothing left to book. The scary-looking slowdown is the most normal part of the year.
• Commitments overshoot, then wash out. In four of the last seven years, total commitments peaked above the final export number — 101% to 104% of it — before settling back as sales cancelled or rolled forward. Watch commitments cross 100% of the WASDE total in February and it looks like a record blowout. It's routine, and it self-corrects.

Where does it all end up? The shipments line lands within a few percent of the WASDE export forecast, every year. WASDE already prices in the front-loading and the washout. The weekly number is a pace read on that target — not a competing forecast.

The test: can the pace beat WASDE? The Test

That's the strong claim, so we tested it head-to-head. At a set of mid-year checkpoints — December through spring — we pitted two forecasts of the final export figure against each other: (1) the concurrent WASDE projection, and (2) a "pace extrapolation" that scales the current commitments up by how much of a typical year is usually booked by that week. To keep it honest, the seasonal-fill curve is computed leave-one-out — it never sees the year it's scoring. We measured each against the eventual settled figure, across soybeans, corn and cotton, seven crop years each.

WASDE wins everywhere — and the way it wins is instructive:

• Soybeans: the pace read is 2.5× worse than WASDE, and worst of all in December — precisely when an early edge would be worth something. In December 2020, with China front-loading hard, the pace extrapolation implied a record ~79 million tonnes; WASDE said ~60; the year settled near 62. The pace read was screaming blowout at the exact moment it was most wrong.
• Corn: a narrower gap, for an honest reason — corn exports are genuinely hard to forecast, so WASDE itself runs 8–10% off at mid-year. But the pace extrapolation is worse still. When the benchmark is shaky, the naive alternative isn't the answer.
• Cotton: WASDE wins at three of four checkpoints; the one exception is a dead heat in January, after which WASDE pulls back ahead. Robust, with an honest asterisk.

The through-line: the pace extrapolation is uniformly mediocre — around 13% error in December for both soybeans and corn — while WASDE's skill varies by crop. WASDE integrates crush, stocks, competitor supply, and analyst judgment; a sales curve extrapolated forward carries none of that. You can't out-forecast it by doing arithmetic on Thursday's number.

So what is the weekly number for? The Use

Plenty — just not the thing people reach for. The value of the Export Sales report was never the implied full-year total; WASDE has that covered. It's in the detail underneath the headline, and in the timing a monthly report can't give you:

• Who is buying, and who is walking away. The report breaks sales down by destination. A wave of cancellations, or one buyer's book quietly shrinking, is a genuine signal — and it shows up here weeks before it reaches a balance sheet.
• Shipment pace against commitments. The gap between what's sold and what's actually moving flags port and logistics stress — grain committed but not leaving.
• New-crop forward sales. Contracts booked for the next marketing year are an early read on demand for a crop still in the ground.
• Surprises against expectations. Not the level, but the deviation from what the trade penciled in for the week — that's the tradable event, and it's gone by the monthly report.