How to judge a calorie app's food database

An app is a search box over a food database, and the database is what determines whether the numbers you log mean anything. Here is how to evaluate one in about ten minutes.

Check one: search a dish you actually eat

Not "chicken breast" — every database handles that. Search the home-cooked dish you eat most often. Dal, rajma, poha, chicken curry.

Good: a handful of entries, clearly labelled, each with a source and a serving weight.

Poor: forty entries for the same dish, values from 90 to 400 kcal per serving, none stating where the number came from.

The second result is not a minor inconvenience. It means every logging decision is a choice between strangers' estimates, and it is also the slowest part of logging — the reason many people abandon tracking, as covered in why two apps give the same roti different calories.

Check two: does the entry name its source

The single most informative attribute of any food entry.

An entry citing a national composition table or a manufacturer can be verified. An entry citing nothing is an assertion. A good app shows the source on the entry itself rather than burying it.

The sources worth trusting, and what each is good for, are compared in IFCT or USDA: which numbers should you trust.

Check three: raw or cooked, stated

Rice roughly triples in weight when boiled and meat loses a quarter when roasted. An entry that does not say which state it describes is ambiguous by a factor of up to three.

Search "rice" and look at whether the results distinguish raw from cooked explicitly. Many do not, and that single ambiguity is the largest source of logging error there is — should you weigh food raw or cooked.

Check four: serving definitions

"1 serving" defines nothing and appears on an enormous number of entries.

Usable entries give a weight, and good ones give household measures alongside it — a katori, a roti, a cup, a tablespoon — because that is how people actually serve food. The reasoning is in household serving.

Check five: coverage of what you eat

Most databases cover American and European packaged foods thoroughly and Indian home cooking poorly, because that is where the data came from.

If a third of your meals require you to improvise an entry, the database does not cover your diet, however large it is. Test with five dishes from your own week rather than with whatever the app's demo shows.

Check six: entry count is not quality

A database advertising eleven million foods is usually advertising eleven million user submissions.

The useful question is not how many entries exist but how many verified ones exist for the foods you eat. A smaller, curated database built from composition tables and manufacturer data beats a vast unverified one for everything except obscure branded products.

Why user entries are biased, not just noisy

Worth stating plainly. People creating an entry for their own meal are estimating their own food, often while dieting.

The estimates skew low, and once created, those entries are logged by thousands of other people. So the error is not random — it has a direction, and the direction is the one that makes a food log read lower than reality.

That compounds the general under-reporting problem described in how accurate are calorie counts, really.

What a good database looks like

Ingredients from national composition tables. Analysed in a laboratory with documented sampling.

Packaged products from manufacturer data. Reliable within the label tolerance.

User entries clearly marked as such, and ranked below verified ones in search results.

State and serving stated on every entry. Raw or cooked, grams and a household measure.

Deduplicated. One entry per food, not forty.

That combination is uncommon, which is why the ten-minute test above is worth running before you commit to an app, as part of the broader evaluation in how to choose a calorie tracking app that sticks.

The thing a database cannot fix

Even a perfect database cannot price the dal you made on Tuesday, because that dal is your recipe with your quantity of oil.

For home-cooked food the answer is to calculate your own dishes once and save them, which turns the database into a source of ingredient values rather than of finished dishes — the method in how to calculate the calories in a recipe you cooked.