Observed, estimated, revised and forecast

Open any agricultural statistics table and the numbers look the same. Same column, same units, same number of decimal places. Some of them are records of what happened. Some are the agency's best guess at what happened. Some are guesses at what will happen. Nothing about the typography tells you which.

The agencies do tell you. They tell you in a flag, a footnote or a letter beside the year, and that marker is routinely the first thing lost when a number is copied into a chart, a headline or a knowledge base.

What the flags say

FAO puts a single letter on every value in its land use tables: A for an official value the country reported, E for a value FAO estimated, I for a value FAO imputed, X for a value taken from another organisation.

Of the 7,074 irrigation figures AgricultureID holds from that source, 2,676 — 38% — carry the A flag. The majority are the agency's own work, published in the same column, formatted identically.

This is not a criticism of FAO. Estimating and imputing is what a global statistical agency is for; the alternative to an imputed value is usually a gap that makes the whole series unusable. The failure would be to reproduce those values without the flag, at which point a reader has no way to tell a reported national statistic from a modelled one.

The letter that changes everything

The USDA Economic Research Service publishes cost-of-production estimates by crop. AgricultureID holds 306 of them. Every single one is a forecast: the source labels the years 2026F and 2027F, and the F is the whole of what distinguishes this dataset from an accounting of what farms actually spent.

Drop the F and the sentence "growing an acre of maize costs $935.79" becomes a statement about the world rather than a projection for a year that has not finished. It reads as a measurement. It is not one.

Revisions are their own category

A revised figure is neither the original observation nor a new one. Agencies revise as late returns arrive, and a series that looks stable in one release can move materially in the next. A corpus that overwrites the old value silently loses the fact that the number changed — which is often the most interesting thing about it.

AgricultureID keeps immutable snapshots with checksums, so the previous release is still there, and change is derived by comparing them rather than asserted.

What this costs us

Carrying the kind of every number means the platform publishes fewer clean statements than it could. It means a chart sometimes needs a footnote. It means we say "the source forecasts" where a competitor says "costs are".

The alternative is a number that is real, sourced, correctly transcribed, and about something other than what the reader thinks.