Four new official databases join Radish: UK, Germany, Canada, Australia
In short
On August 30, 2026, Radish integrated four public food-composition databases: CoFID (United Kingdom, 2,820 foods), BLS (Germany, 6,225 foods), CNF (Canada, 2,409 foods) and AUSNUT 2023 (Australia, 3,215 foods) — 14,669 additional foods, on top of the CIQUAL (France) and USDA (United States) data already in use. Each table is under a verified open license, and every food was checked against the tables already in place before being added, so the same item never shows up twice.
We usually write about our corrections. This one is an addition: four national databases, public and openly licensed, that widen coverage of whole foods outside France and the United States. No new app feature here — a foundational data update.
What was added, table by table
Each table stayed in its own file instead of being merged into the existing CIQUAL/USDA data — that way a whole batch can be pulled out if a problem shows up, without touching the others.
- CoFID (United Kingdom) — 2,820 foods with complete macronutrients (calories, protein, fat, carbs all populated). Source: Public Health England / DHSC, McCance and Widdowson's Composition of Foods Integrated Dataset, 2021 edition, under the Open Government Licence v3.0.
- BLS — Bundeslebensmittelschlüssel (Germany) — 6,225 foods with complete macronutrients. Source: Max Rubner-Institut (2025), BLS version 4.0, CC BY 4.0 (DOI: 10.25826/Data20251217-134202-0).
- CNF — Canadian Nutrient File (Canada) — 2,409 foods, after excluding 3,529 entries flagged as unmodified copies of USDA data (source code "0" in the original file) and 55 further name duplicates. Source: Health Canada, 2026 edition, under the Open Government Licence – Canada.
- AUSNUT 2023 (Australia) — 3,215 foods with complete macronutrients. Source: Food Standards Australia New Zealand, AUSNUT 2023, under CC BY 4.0.
All four join the same search index as CIQUAL and USDA foods: they're findable like any other item in the database, not in a separate tab.
Deduplication: how we avoid showing the same food twice
Adding a national database to a database that's already international creates a direct problem: a given food — a chicken breast, an apple — very often already exists in CIQUAL or the USDA under an equivalent English name. Without a check, search would return "chicken breast, raw" twice, with two slightly different values and nothing to explain the gap.
The approach: deduplication happens on the way in, not afterward. Before adding a food from a new table, its normalized English name is checked against CIQUAL, the USDA, and every table already integrated that day — CoFID before BLS, CoFID + BLS before CNF, and so on. A name already present isn't added a second time; the food stays served by one table only. Rough scale of the rejections on this batch: 374 duplicates excluded on the BLS side (against CIQUAL, USDA, CoFID and the other embedded files), and 3,584 exclusions on the CNF side (3,529 USDA copies flagged by the source itself, plus 55 name duplicates).
What stays deliberately cautious
Three choices, documented in each table's source code, are worth stating publicly:
- Food category not decoded for BLS, CNF and AUSNUT. These three sources don't provide a food group in a directly reusable format; rather than invent one, these foods carry a generic category that doesn't trigger any category-specific Nutri-Score scale (the general scale applies).
- BLS micronutrients not yet displayed. The values exist in the data but aren't wired to the on-screen reference values yet, pending further review — the same cautious stance applied to CoFID when it first arrived.
- Four micronutrients deliberately absent from AUSNUT entries. The Australian source imputes any unmeasured value to zero at the source (no blank cells in the workbook). For vitamin A, vitamin C and vitamin B12, the zero rate in the data is high enough to be judged biologically implausible; those three fields, plus vitamin D (unverified), are omitted rather than shown as a value that might be false. Macronutrients and minerals, whose zero rate stays in a plausible range, are kept as-is.
In both cases — missing data or a zero we chose not to keep — the entry shows the gap rather than an estimate: the same rule already applied to CIQUAL and USDA data.
Search a food in the databaseCIQUAL, USDA, CoFID, BLS, CNF, AUSNUT — every entry shows its source.Get Radish →Why we're documenting this publicly
A food-composition database is never neutral: every country measures, classifies and publishes its data by its own conventions (what counts as a carbohydrate, what a zero means, which nutrients get measured at all). Merging four more sources without saying so would create a false sense of consistency. The rule we apply: every entry keeps a record of its source table, visible on its page (a source badge), and whatever was excluded or left out during integration is documented rather than silent.
General information about how the app's food database works, not nutrition advice. Accurate as of August 30, 2026.
Frequently asked questions
How many foods were added on August 30, 2026?
14,669 in total: 2,820 from CoFID (United Kingdom), 6,225 from BLS (Germany), 2,409 from CNF (Canada) and 3,215 from AUSNUT 2023 (Australia).
Are these datasets under an open license?
Yes. CoFID and CNF are under open government licenses (UK and Canada); BLS and AUSNUT are under CC BY 4.0. All four require attribution, present in the source code and on each food's entry.
How does Radish avoid showing a food twice?
Every new table is checked, food by food, by normalized English name, against CIQUAL, the USDA and every table already integrated, before being added. A food already present isn't added again; it stays served by a single table.
Why are some micronutrients missing on Australian or German foods?
On AUSNUT, four micronutrients (vitamins A, C, B12, D) are deliberately omitted because the source imputes any unmeasured value to zero, and the observed zero rate on these four fields is judged implausible. On BLS, micronutrients exist in the data but aren't wired to the display yet, as an extra precaution. In both cases, the gap is shown as a gap rather than an invented value.
The takeaway: 14,669 more foods, four verified open licenses, and deduplication that happens on the way in rather than after the fact. Every entry keeps its source, visible in the app.
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 →