CIQUAL vs USDA: why Radish relies on two databases rather than one
In short
CIQUAL (ANSES, France) and USDA FoodData Central (United States) measure food under different conventions — what counts as a carbohydrate, how a missing nutrient is recorded, which edition applies. Radish doesn't blend them into an average: every entry keeps its source table. Since August 30, 2026, four more databases (CoFID, BLS, CNF, AUSNUT) follow the same principle.
A question that comes up often: why does the same "chicken" entry show a slightly different value depending on whether it came from the French table or the American one? The answer isn't a bug — it's the nature of a food-composition database. Here's what Radish's two founding tables measure, how, and why they're never blended into a single number.
CIQUAL: the French reference, maintained by ANSES
CIQUAL is the food-composition table published by ANSES (the French agency for food, environmental and occupational health safety). Radish uses the CIQUAL 2025 edition, updated November 19, 2025 (DOI: 10.57745/RDMHWY), under the Etalab Open Licence 2.0. Every CIQUAL code shown on a Radish entry points to a specific record in that table — verifiable on ANSES's own site in thirty seconds. Matches between a Radish food and its CIQUAL code were decided one by one, not by an algorithm alone: an automated name-matcher's error rate proved too high to publish without human review.
USDA FoodData Central: the American reference
USDA FoodData Central is the equivalent database maintained by the US Department of Agriculture. It covers both whole foods and branded products. On Radish, a USDA identifier shown on an entry is always an explicit, verified fdcId: a bare "USDA" badge with no identifier wouldn't be traceability, just an empty claim — the rule applied since August 1, 2026 is that such a bare badge is removed and the entry reverts to "estimated" rather than display a confidence it hasn't earned.
Why not blend them into a single value?
Two tables measuring the same food — an apple, a chicken breast — rarely land on the exact same figure to the gram. The causes are documented and aren't an error by either side:
- Carbohydrate calculation convention. One table may count carbs "by difference" (100 minus everything else) while another measures directly; whether fiber is included in that total also varies by table.
- Different varieties and farming practices. A dairy product or a cut of meat sold in France doesn't necessarily match its North American counterpart in average composition — breeds, animal feed, regulation.
- What a "zero" means. Some tables record 0 for "measured and truly nil," others for "not measured, assumed nil," or "below the lab's detection threshold." Those aren't the same information, even when the displayed figure is identical.
- Edition date. Each table updates on its own schedule; a CIQUAL 2025 value and a USDA value pulled on another date don't necessarily describe the same production context.
Averaging the two would produce a number that matches neither source and that couldn't be checked anywhere. The rule applied on Radish runs the other way: every entry shows the table it came from, with its code, so it stays verifiable. That's what the Data sources page documents, and what you can check directly against the app's CIQUAL data.
Now six databases, not two
The "one entry, one table, one verifiable code" rule didn't change while Radish only covered France and the United States. It was put to the test on August 30, 2026, when four more national databases joined: CoFID (United Kingdom, Public Health England/DHSC), BLS (Germany, Max Rubner-Institut), CNF (Canada, Health Canada) and AUSNUT 2023 (Australia, Food Standards Australia New Zealand) — 14,669 foods in total, each under a verified open license. The full breakdown, table by table and license by license, is in four new official databases join Radish. Here too, no silent merge: each table keeps its own file, its own source badge, and a food already present in one table isn't re-added by another.
See a sourced entry in the appCIQUAL, USDA and four more tables — every entry shows its source.Get Radish →What this means for you, in practice
If an entry you're looking at comes from CIQUAL, it describes food as measured in France. If it comes from the USDA, it describes an American measurement. Both are legitimate, and a small gap between them isn't an error to fix — it's two real measurements from two different contexts. What matters is knowing which one you're looking at: that's the role of the source badge shown on every entry, covered in the guide how to read a Radish food entry.
General information about how the app's food database works, not nutrition advice.
Frequently asked questions
Why does the same "apple" entry show different values depending on the source?
Because CIQUAL and the USDA measure real food in different contexts (varieties, farming practices, carbohydrate calculation conventions), not the same "theoretical apple." Radish doesn't average them: every entry keeps the value from its source table, with its code.
Does Radish default to one table over another?
No: each food is served by one table only, whichever it was matched to first. The source table is shown on the entry via the source badge.
How many databases does Radish use today?
Six: CIQUAL (France), USDA FoodData Central (United States), CoFID (United Kingdom), BLS (Germany), CNF (Canada) and AUSNUT 2023 (Australia).
Can a value change if the source table gets updated?
Yes, once that new edition is integrated into Radish. The edition date in use is documented on the Data sources page.
The takeaway: two ways of measuring a food aren't two errors to fix, but two different realities. Radish keeps each table separate, with its verifiable code, rather than fabricating an average that would exist nowhere else.
General information about how the app's food database works, not nutrition advice.
Ready to try Radish?Private calorie counter, official Nutri-Score, no ads. Free.Get Radish →Prefer the app installed? Get Radish on iPhone or Android →