IFCT or USDA: which numbers should you trust?

Every calorie number that can be checked traces back to a food composition table: a national reference built by analysing samples of food in a laboratory and publishing the results with the method attached.

Two of them matter for most people reading this, and they are good at different things.

What each one is

IFCT 2017 is the Indian Food Composition Tables, published by the ICMR–National Institute of Nutrition. It replaced a reference whose core analytical work dated back decades. It covers Indian cereals and millets, pulses in the forms actually eaten, regional vegetables and fruit, and a set of commonly prepared dishes.

USDA FoodData Central is the American reference and, because it is comprehensive, free and available in bulk, the base layer of most calorie apps built outside India. It contains several distinct datasets: newly analysed Foundation Foods, the older SR Legacy reference set, Branded Foods from manufacturer labels, and survey data covering prepared dishes.

Both are serious pieces of work. Neither is universal.

Why they disagree about "the same" food

Four reasons, all legitimate.

The food is different. Indian atta is milled at a different extraction rate from American whole wheat flour, with different fibre and moisture. Buffalo milk and cow milk are not interchangeable. Toor dal is not the same pulse as the "pigeon pea" entry in a Western table, grown and processed the same way.

Recipes are different. A composition table's "vegetable curry" was analysed from a representative recipe for its country. The representative Indian recipe and the representative American one are not the same dish.

Analytical method. Carbohydrate is sometimes measured directly and sometimes calculated by difference — subtracting water, protein, fat and ash from 100. The two approaches produce slightly different totals for the same food.

Fibre conventions. Whether fibre is inside the carbohydrate figure, and what energy value it is assigned, differs between systems. This changes the computed calorie total for high-fibre foods.

Which to use, by food

Indian ingredients and dishes: IFCT. Atta, all the dals, millets, regional vegetables, paneer, ghee, Indian preparations. This is what it was built for.

Generic and internationally traded items: USDA. Chicken breast, eggs, oats, almonds, olive oil, bananas. Foods that are broadly the same everywhere, and USDA's coverage and analytical depth are excellent.

Packaged products: the label, or a database of labels such as Open Food Facts. Manufacturer data beats any general table for a specific branded item.

Home-cooked dishes: neither. Calculate your own from ingredient values, because your recipe is not the representative recipe. That is the whole argument in how to calculate the calories in a recipe you cooked.

Where both are weak

Prepared dishes. A table can analyse a representative sample of "chicken curry", and the result describes one recipe with one quantity of oil.

Your version differs, usually in the fat, and cooking fat is the dominant variable in Indian cooking — as demonstrated in how much oil is actually in your curry. No composition table can solve this, and no app built on one can either. It has to be handled by counting your own ingredients.

Why the source should be visible

An app that shows a number without saying where it came from is asking for trust it has not earned.

The practical test: search a common dish and see whether the results distinguish a laboratory-analysed entry from a user-typed one. If they all look the same, most of them are user entries and you are choosing between strangers' guesses — the situation described in why two apps give the same roti different calories.

This is also the main criterion in how to judge a calorie app's food database.

How Forkbell uses them

Indian dishes and ingredients come from IFCT 2017. Generic items come from USDA FoodData Central. Packaged products come from Open Food Facts and manufacturer data. Every figure on a page names which one it came from, and estimates are labelled as estimates.

That is not a differentiator so much as a minimum standard, and it is uncommon enough to be worth stating.

What to do with all this

Prefer entries that name a table. Prefer IFCT for Indian food and USDA for generic ingredients. Calculate home dishes yourself. And when two sources disagree by 10 or 15 per cent, stop optimising and pick one — consistency across your own log beats accuracy in any single entry, for the reasons in how accurate are calorie counts, really.