FAOSTAT Food and Agriculture Statistics

FAOSTAT Food and Agriculture Statistics is the shelf's answer to a question most retail datasets never reach: where do the goods actually come from, and what did they cost at the farm gate. It is the Food and Agriculture Organization's flagship statistical database, holding roughly 171 million observations across 69 statistical domains, covering 245+ countries and territories across all FAO regional groupings from 1961 onward.

Five domains do most of the work a food retailer asks of it. Production: Crops and livestock products records 278 products spanning primary and processed crops and livestock, with area harvested, yield and production quantities. Trade: Crops and livestock products carries export and import quantities and values under International Merchandise Trade Statistics methodology, stamped 2026-07-24 in the release verified this pass. The Detailed Trade Matrix resolves who sold to whom, bilaterally. Food Balances shows supply and utilisation per food item, processed commodities included. Producer Prices run annually from 1991 with monthly resolution from January 2010. Get a sample of this dataset cut to your categories before anything else.

One delivered row per country-product-element-year observation, typed identically whichever of the 69 domains it comes from. The shapes below are written field-for-field from the verified dictionary - each value is the documented example value carried on that field, with real rows populating once a scope names countries, commodities and years:

# one delivered row per area-item-element-year observation

row_shape  : full observation (annual)
area       : Afghanistan          area_code : 2      m49 : '004
item       : Almonds, in shell    item_code : 221    cpc : '01371
element    : Producer Price (LCU/tonne)              code : 5530
year       : 1993                 months : Annual value
unit       : LCU                  value  : 46000.000000
flag       : A (official figure)
--------------------------------------------------------------------
row_shape  : monthly price observation
months     : January through December, from 2010
unit       : LCU                  flag  : per-observation provenance code
--------------------------------------------------------------------
row_shape  : bilateral flow (detailed trade matrix)
exporter   : reporting country    importer : partner area
elements   : export/import quantity (t) and value (1000 USD)
--------------------------------------------------------------------
row_shape  : food balance sheet line
elements   : supply and utilisation per food item, processed included
unit       : kcal/capita/day among others
--------------------------------------------------------------------
domain census, five headline shelves of the 69:
QCL  Production: Crops and livestock products   4,209,110 obs   2025-12-31
TCL  Trade: Crops and livestock products       17,143,873 obs   2026-07-24
TM   Trade: Detailed trade matrix              52,410,630 obs   2025-12-23
FBS  Food Balances (2010-)                      4,820,497 obs   2025-10-28
PP   Prices: Producer Prices                    1,319,563 obs   2026-01-09

Two things in these shapes earn a slow read. First, the grain: nothing here is one table per topic - every domain collapses to the same five keys, so a six-decade archive loads as a single panel. Second, the flag column rides on every observation, separating official declarations from estimates and imputations; pipelines that respect it produce defensible numbers. Get a sample of this dataset and rows land in exactly this shape, filled with your scope.

Eight documented field groups define every delivered row, each definition verified against the source's own file structure during the August 2026 research pass rather than reconstructed from prose documentation. Three identify the observation - geography, commodity, and the measured element - with numeric codes plus M49 and Central Product Classification crosswalks, so joins bind on integers instead of spellings. One states the year, one states the reporting period, one states the unit, and Value carries the magnitude itself, stored upstream as text with fixed decimals and cast to numeric on load.

The quiet differentiator is the eighth surface: a per-observation provenance flag classifying each figure as official, estimated, imputed or externally sourced. Most compilations flatten that distinction away; here it travels on all hundred-seventy-million-odd rows, which means a model can weight an official customs declaration differently from an imputed estimate instead of treating them as equals. Everything beyond the spine folds under additional fields on request - pivots wide, decoded flag vocabularies, aggregates pre-separated from countries, category subsets named in commodity terms. Say which when you request the sample.

Geography - 245+ countries and territories spanning all FAO regional groupings, from major producers to the smallest territories. Regional grouping labels sit beside actual countries, which makes quick global totals convenient and naive sums dangerous: bind on the numeric area codes and filter the groupings deliberately before any panel work.

Temporal - 1961 to the most recent year available across most domains, which is longer than most commercial series in the category. Two exceptions matter for planning: producer prices begin annually in 1991 and go monthly from January 2010, and the current-generation Food Balances run from 2010 forward. Six decades of production and trade history sits in one uniform shape.

Granularity - country-year-product-element rows throughout, rising to bilateral exporter-importer-product-year rows inside the Detailed Trade Matrix. There is no store, SKU or shipment level anywhere in the catalogue; anyone promising shelf-level history from this source is selling a different dataset. What the country grain buys is completeness: every commodity, every year, every geography, one join key.

API, files, or your warehouse. Daily, weekly, or hourly.

Your cadence is your call regardless of the rhythm of the underlying releases - these are yearly and monthly observations, so even an hourly warehouse load simply re-serves the latest declared values while keeping dashboards, joins and caches warm. Every delivery ships the complete field dictionary above, the sample rows in exactly this shape and the coverage profile mapped to the countries, commodities and years you named, long-form or pivoted wide. Ask for one commodity or the full multi-hundred-million-row grid; the schema does not change with the breadth.

• Sourcing and assortment intelligence - which countries grow, raise and process a category, and how that map shifts decade over decade, computed from one ledger instead of stitched national statistics offices.
• Producer-price to shelf-price margin analysis - farm-gate prices in local currency units from 1991, monthly since 2010, joined onto your own retail pricing for margin-trend work.
• Supply-security and disruption monitoring - bilateral import dependence by commodity and partner from the Detailed Trade Matrix, the arithmetic behind contingency briefs for specialty categories.
• Demand-forecasting features - production, trade and balance series that give category forecasts exogenous supply-side signals rather than sales data alone; see demand forecasting.
• Market sizing and category entry - consumption and supply utilisation per food item to size a niche before committing shelf space; see market sizing.
• Citation-grade baselines - a UN-agency compilation with provenance flags gives reports and buyer decks numbers that survive scrutiny; see price monitoring.

Market researchers and consultants rank highest - relevance 3 of 3 - because category sizing, sourcing maps and price benchmarks all fall out of one normalized feed; see market researchers use cases. Data scientists and ML engineers, also 3 of 3, get roughly 171 million tidy rows that load once and join forever, with flags ready to become feature weights; see data scientists use cases. E-commerce operators score 2: assortment and sourcing signals behind specialty-food catalog decisions; see e commerce operators use cases. Journalists, academics and students quote UN-published figures with provenance attached; see journalists academics use cases. Competitive-intel teams sit at 1: the grain is country-commodity-year, so no firms and no transaction prices appear anywhere in it. Everything above ships daily, weekly, or hourly.

Within other specialty retail data, this record owns the supply side at global scale: production, bilateral trade, balances and producer prices for every agricultural commodity, six decades deep, one dictionary across all 69 domains. The neighbours own different jobs. Eurostat - European Statistical Office Data Portal reads European demand - retail turnover indices for 38 NACE classes - where this record reads the goods themselves. Kaggle - World Food Facts (Open Food Facts mirror) holds product-level attributes and ingredients for barcodes on shelves, with no country-time panel underneath. FDA Recalls, Market Withdrawals & Safety Alerts covers safety events in one market. Observatory of Economic Complexity - Meat & Seafood Preparations Trade Profiles presents trade profiles for one commodity family this record covers across all of them.

Head-to-head against the nearest demand-side counterpart: Eurostat Data Portal vs FAOSTAT Food & Ag Statistics. The practical pairing runs the other way too - pair the production and price domains here with a barcode-level product base and the macro trend explains what the shelf observes.

Three things worth settling upfront.

First, the grain is country-level by design. Store, SKU and shipment detail does not exist in any of the 69 domains; if your question lives below the country line, this record works as the exogenous frame around a finer source, not as the fine source itself.

Second, regional labels ride beside countries among the geography values. Summing names instead of binding on numeric codes counts continents alongside their members before the first chart renders - deliveries separate the two at load time on request.

Third, units vary by element, deliberately. Tonnes, hectares, local currency, thousand USD and kcal/capita/day all appear, declared per row. Deliveries ship the unit column resolved rather than silently harmonized, because the harmony you want depends on the question you asked. Name the countries, commodities and years that matter and the sample ships shaped to that scope.

Because the raw artifact is a sixty-nine-shelf library wearing one shared schema and several decades of inconsistent national reporting, and almost nobody's question needs all of it in that shape. Datadory normalizes types, decodes the provenance flags, separates aggregates from countries, resolves naming conventions into clean joins, and pairs the production and price panels with whatever demand-side or product-level source completes the picture - scoped to the countries, commodities and years you actually model. Start with a sample of this dataset, then browse the rest of the shelf on the other specialty retail data hub or the best other specialty retail datasets ranking.