Research & Analysis

A CONAB “national” coffee total that carried a single state. A CFTC column name off by one underscore that read as empty, became zero, and flipped the sign of a published positioning signal. A WASDE trade bloc that lost a member without changing its name, moving 15.6 million tons in a marketing year that had already closed. All three passed every check we had, because in each case the data was internally fine — the defect lived in what the file did not say. This is what they were, why the obvious tests miss them, and the MIT-licensed tool we published in response.

A grower's husbandry follows margin, not price, and coffee is perennial — so the consequence of a skipped fertilizer application lands two seasons later. We built 22 years of Brazilian cost structure from CONAB's own planilhas to test that. Three tests failed, and the reason turned out to be the instrument rather than the idea: a cost share is mostly a fertilizer price index in disguise, and once the price is removed there is no grower-margin signal left. An open research note, published with its dead ends intact — including the one result that reversed after we found a double-count.

Colombia replanted most of its coffee between 2007 and 2019 and it worked: the young share of the park rose from 56.7% to 85.9%, and yield hit the best in the area record. It has fallen every year since, and aged area is up 43%. The counterintuitive part is where — the aging is concentrated in the south, the region that expanded, not in the shrinking Eje Cafetero, because one planting wave is crossing out of the young band together. We also tested whether an aging park predicts falling yield, and report the null: the level correlation looks strong at a two-year lag, and does not survive first-differencing.

We read the entire Fundação Procafé archive — 529 monthly plant-health bulletins from 2010 to 2026 — and turned it into a coffee leaf-rust time series no one else has. June 2026 rust was genuinely low for June (rank 3 of 17), not just moderate; our ERA5 temperature matches Procafé's own field gauge to r=0.98 over 196 months (rainfall is noisier, and blind to dew); warmer regions carry more rust on average but the year-to-year ranking reshuffles; and leaf retention — how much canopy a tree keeps after a rust year — is a leading indicator of the 2027 crop that balance sheets can't see. Here's what it changed in our signal.

Every Thursday the USDA export-sales number moves the ag desks — and it's the most misread report on the screen. Commitments front-load (half to 85% of the soybean year is booked by early December), overshoot the final total, then wash back out on cancellations. Across soybeans, corn and cotton, you can't extrapolate that pace to beat WASDE — it misleads exactly when it looks most exciting. Here's what the weekly report is actually good for.

A living, cited digest of what the USDA's own reports — WASDE-cycle circulars and FAS attaché posts — say about China's grain demand, kept separate from our own read. The August cycle cut China's corn book a second month — MY2026/27 to 5 MMT, MY2025/26 to 4 — with zero U.S. corn on it. The newer signal is sorghum, the one U.S. feed-grain line into China that has actually been shipping: imports cut 7.5 → 6.6 MMT, U.S. sorghum exports cut 5.2 → 4.3 in the same table. Its soybean book stays huge and growing (~115 MMT, ~61% of world trade) but the tonnage routes to Brazil and Argentina. Attaché figures are flagged as staff research, not official; every claim carries its source; refreshed each WASDE cycle.

Everyone watches South American production; far fewer watch what happens to the grain before it reaches a port. Argentina crushes over 40 million tonnes of soybeans a year and Brazil now diverts about a sixth of its corn crop into ethanol — both growing faster than the crops that feed them. How we measured it, including the marketing year we had to identify from the data, the series we had to derive because the regulator discontinued it, and the one number we still cannot source.

Most sugar commentary starts with a weather map. But the single most price-relevant lever in the world sugar balance is a fortnightly decision inside Brazilian mills: make sugar, or make ethanol. This is how that decision is priced — sugar-ethanol parity, the CONAB Center-South mill-mix, and the 70% pump rule — and why it belongs on a softs trader's screen next to the No.11 net position.

A third of Brazil's grain no longer sails from São Paulo — it rides barges down the Amazon and its tributaries, the Arco Norte. Whether those rivers are deep enough to float a loaded barge in October is set, months earlier, by how warm a patch of the tropical North Atlantic was back in April. The geography of that teleconnection, the two-leg truck-and-barge chain that carries the crop, Brazil's thin logistics backbone, and how we turned it into a validated, months-ahead navigability forecast.

Once a month, one USDA report resets the terms of debate in every grain, oilseed and cotton market at once. Its job is to take everything known about how much of a crop exists and how much will be used, and reconcile it into a single balance sheet — so you don't have to assemble trade, crush, feed and stocks yourself. What the WASDE is, how the balance sheet works, why traders read it, and how ending stocks becomes the stocks-to-use ratio the market actually trades.

Fertilizer is the quiet hinge between the energy market and the food market — the nitrogen in a bag of urea was, months earlier, natural gas. A guide to our four Fertilizer Supply-Risk reports — Brazil, India, the US and the global producers — what each tracks, why each region sits on the chain differently, and how to read the Tightening / Neutral / Easing signal at the top of every one.

They sit side by side on the board and both say "cattle," but they are not two prices for one thing — they are the same animal at two stages of its life, joined by corn and time. How the cattle crush works, why feeder cattle behaves as a leveraged bet on live cattle minus corn, and why the spread between them sits near the widest on record — with the herd-rebuild twist that keeps feeders tight precisely when you'd expect relief.

Most people watch El Niño and assume that is the ocean-climate story. The Indian Ocean Dipole is a separate driver that frequently runs alongside ENSO, sometimes runs entirely without it (2019), and peaks two months earlier — which is why we publish two ocean indices rather than one. Forty-four years of data on why one index cannot stand in for the other.

When ENSO and the Indian Ocean Dipole line up, their rainfall effects can reinforce into floods or drought over East African coffee — or, for the Indian monsoon, quietly cancel. The phase-combination matrix, how rare a true alignment actually is, and an honest account of where the record goes silent.

Frost is the most violent weather risk in coffee, and the nights that do the damage are calm, clear and dry — the exact inverse of the wet nights that wake leaf rust. The physics of a killing frost in Brazil's arabica belt, why we model an estimated leaf temperature rather than the thermometer reading, and where the signal falls short.

Why we monitor environmental conditions for coffee leaf rust (Hemileia vastatrix) across seven origins — the two-stage thermal biology behind the signal, why we also track input-affordability, the model's known limitations, and a defense of its value as a leading, transparent indicator.

El Niño and La Niña are the most powerful systematic supply risk factors in commodity markets — and the most underused. A practical guide to reading ENSO signals across coffee, grains, softs, and energy.

NOAA has officially transitioned from ONI to RONI — the Relative Oceanic Niño Index. For commodity traders, this is not a technical footnote. It changes how you read ENSO signals, validate historical analogs, and manage supply risk.