857,655 products analyzed. What your calorie scanner isn't telling you.

Why this study

When you scan a barcode in a calorie tracker in France, MyFitnessPal, Yazio, Lifesum or Lean, the data comes from more or less the same source: OpenFoodFacts, a remarkable French collaborative database under the ODbL open license. Millions of product entries, filled in by volunteer contributors.

A collaborative database is both its strength and its limit: anyone can create or edit an entry, and nobody checks systematically. Nobody had ever quantified the actual reliability of what your scan returns. So we downloaded the complete public dump of August 23, 2026 and tested every product sold in France.

The method: every entry tested against itself

No lab needed to detect a wrong entry: just confront it with itself. A food's calories can be derived from its macronutrients with the Atwater factors, the very ones the European INCO regulation imposes on manufacturers: 4 kcal per gram of protein, 4 per gram of carbs, 9 per gram of fat.

If an entry displays 100 kcal but its own macros add up to 250, one of the two lines is wrong. And your app shows you one of the two without blinking.

Result 1: almost 1 entry in 3 is incomplete

391,707 products sold in France have no calories filled in, or are missing their three macronutrients. Concretely: you scan, and the app displays 0 kcal, or calories with no protein or carbs. You feel like you're tracking; you're logging emptiness.

It's the sneakiest error, because it doesn't look like an error: the entry displays, the product has a name, a photo. Only the essential part is missing.

Result 2: 46,757 products with inconsistent calories

Across the 857,655 complete entries, 46,757 display calories incompatible with their own macros, by at least 10 % and 30 kcal per 100 g. Every dot in the chart below is a real product: on the x-axis what its entry declares, on the y-axis what its macros yield. An honest entry lands on the diagonal.

The distribution of gaps shows the database is massively correct, then trails off into a long tail: wrong entries are not slightly wrong, they are very wrong.

18,251 products, two material inconsistencies out of five, are off by a factor of two or more. At that level, it's no longer an imprecision, it's a data point that flips your daily calorie balance.

Adding incomplete and inconsistent entries together: more than one scan in three in France returns missing, incomplete or inconsistent data.

Products sitting in your cupboard

Every entry below was re-checked online on publication day. On the left what the OpenFoodFacts entry displays, on the right what its own macros yield:

The textbook case: Raynal & Roquelaure's Auvergne-style cooked lentils exist under two different entries. One displays 48 kcal/100g, the other 207. The real value is around 99. Depending on which barcode your app catches, you believe you're eating half or twice the reality.

A user spreading 30 g of the butter above undercounts 110 kcal. Every single day. While believing they're doing things right.

Lean checks entries for consistency and can also scan your plate from a photo, without depending on a barcode.

The most affected categories

The inconsistency is not randomly distributed, and the ranking's irony is cruel: the most affected category in the entire database is the one athletes track the most. More than one protein bar in five (20.8 %) has an inconsistent entry.

Complex recipes and fortified products concentrate the errors: protein bars, confectionery, candies, cheeses. Simple products fare better.

A note of honesty: candies, confectionery and bars are partly explained by sweeteners. When an entry doesn't declare its polyols, the recomputation overestimates the gap. We corrected everything that was declared; the residue also reflects that declaration gap, which fools your app in exactly the same way.

What this changes for your tracking

First, what this does not mean: you shouldn't stop scanning, and OpenFoodFacts remains a remarkable project that lives and corrects itself continuously. The curve below shows it: entries created in 2020-2021 are the least reliable (up to 7.5 % inconsistency), and quality has clearly recovered since 2023.

Four of the six worst entries in the dump had already been corrected online by the time we checked. But your app may have synced the wrong version, and the hundreds of thousands of entries created in the worst years are still in the database.

Two reflexes to remember:

1. Scanning is a speed tool, not a truth tool. After a scan, a glance at the macros is enough: if protein, carbs and fat don't match the displayed calories, the entry is wrong. That's exactly this study's test, and it takes two seconds.

2. The invisible error isn't on your plate, it's in your expenditure. A wrong entry off by 110 kcal takes ten seconds to fix once spotted. A daily expenditure that's wrong by 300 kcal is never visible: it's buried in the « activity multiplier » your app made you pick at sign-up. That's where Lean concentrates its precision: TDEE = BMR + NEAT + EAT + TEF, with a BMR computed on your real bodyfat and a NEAT based on your real steps, not on a ticked box.

Get Lean, stay Lean: the only app that computes your expenditure from measurements, not checkboxes.

Frequently asked questions