LucaFor packaged food, scan the barcode. It leads to the manufacturer's nutrition declaration, which is produced under labeling rules, while a photo can only guess from appearance. Label data still has limits: values are rounded, regulators accept some tolerance, and open, volunteer-built food databases can contain errors. Check the entry against the pack and log the amount you ate.
A photo estimate and a barcode scan answer different questions. The photo asks “what does this look like?”. The barcode asks “which product is this?”. For anything that comes in a package, the second question has a much better answer, because the manufacturer has already declared what is inside.
Why a label beats a visual guess
A photo sees a surface. It cannot tell a full-fat yogurt from a low-fat one, a sweetened cereal from a plain one, or a protein bar from a chocolate bar with similar wrapping removed. The differences that matter are in the recipe, and the recipe is invisible. The limits of photos are covered in how photo calorie estimation works.
A nutrition label is a declaration made under rules. In the EU, the European Commission explains that Regulation (EU) No 1169/2011 requires the vast majority of pre-packed foods to carry a nutrition declaration with the energy value and the amounts of fat, saturates, carbohydrate, sugars, protein and salt. In the US, the Nutrition Facts label is governed by 21 CFR 101.9.
| Photo estimate | Barcode and label | |
|---|---|---|
| Identifies | Foods by appearance | The specific product |
| Nutrition values come from | Typical values for similar foods | The manufacturer’s declaration |
| Sees the recipe | No | Yes, as declared |
| Knows how much you ate | Estimates it | No, you enter it |
| Works for | Home-cooked and plated meals | Packaged food |
The last two rows matter. A barcode knows nothing about your portion, and it does not exist for a homemade stew.
What is an open food database?
A barcode is only a number. Something has to connect it to a label. Luca uses an open, volunteer-built food database for that.
The project describes itself as a database of food products with ingredients, allergens, nutrition facts and other information found on product labels, made by a non-profit association of volunteers. Contributors use an app or their camera to scan barcodes and upload pictures of products and their labels. Its terms of use say that product information is collected in a collaborative way, and that users of the website can add, complete or correct product data.
The license
The project’s data page states that the database is available under the Open Database License (ODbL). In the license’s own summary, you are free to share, create from and adapt the database, as long as you attribute it, share adapted versions under the same license, and keep it open.
Data can contain errors
Open, crowd-sourced data is powerful and imperfect. The project is direct about it. Its terms state that the information and data are provided as-is, and that it makes no promise of their accuracy, completeness or comprehensiveness.
In practice, the errors you may meet are ordinary ones:
- A typo made when someone copied the label.
- An old entry for a product whose recipe or pack size has since changed.
- Values entered per serving in a field meant for per 100 g, or the reverse.
- A missing product, or one with some fields empty.
So check the entry against the pack. It takes a glance: do the calories and the serving size on screen match the label in your hand? If not, trust the pack and edit the entry.
How close is a label to the real value?
Label values are the most direct data you will get for packaged food. They are still not laboratory measurements of your particular pack.
Rounding
Under 21 CFR 101.9, US labels express calories per serving to the nearest 5-calorie increment up to and including 50 calories, and to the nearest 10-calorie increment above 50. Amounts under 5 calories may be expressed as zero. Total carbohydrate and protein are expressed to the nearest gram.
The European Commission’s 2012 guidance suggests rounding energy to the nearest 1 kJ or kcal, and fat, carbohydrate and protein to the nearest 1 g at 10 g or more per 100 g.
Tolerances
Foods vary. The same Commission guidance explains that foods cannot always contain the levels of energy and nutrients printed on the label, “due to natural variations and variations from production and during storage”.
| Rule | What it says |
|---|---|
| US, 21 CFR 101.9(g)(5) | A food is deemed misbranded if a nutrient such as calories, total sugars, total fat, saturated fat, cholesterol or sodium measures more than 20% above the declared value |
| US, 21 CFR 101.9(g)(4)(ii) | Naturally occurring protein, total carbohydrate and fiber must be at least 80% of the declared value |
| EU guidance, fat | Under 10 g per 100 g: plus or minus 1.5 g. From 10 to 40 g: plus or minus 20%. Over 40 g: plus or minus 8 g |
| EU guidance, carbohydrate, sugars, protein, fiber | Under 10 g per 100 g: plus or minus 2 g. From 10 to 40 g: plus or minus 20%. Over 40 g: plus or minus 8 g |
Two notes on the EU rows. The document is guidance for the authorities that check labels, and it states that it has no formal legal status. Its tolerances also include measurement uncertainty.
None of this is a reason to distrust labels. It is a reason not to chase single calories. A label is a close, regulated description of a product, and that is far more than a camera can offer.
Serving size vs per 100 g
The most common barcode mistake has nothing to do with the database. It is logging the wrong amount.
- US labels are per serving. The FDA notes that the serving size is not a recommendation of how much to eat or drink, and that one package may contain more than one serving. Some packages show two columns, per serving and per package.
- EU and UK labels are per 100 g or 100 ml. The European Commission says all the information must be expressed per 100 g or per 100 ml, and may also be given per portion. The NHS adds that the manufacturer’s idea of a portion may be different from yours.
Per 100 g is handy for comparing two products. Per serving is handy when you eat one serving. For everything in between, you need the weight you ate. Check the pack weight, estimate the share you had, and log that. How to estimate portions has references that help.
When to use which
- Packaged, with a barcode: scan it, check it against the pack, enter your amount.
- Home-cooked or plated: photograph it or describe it, then adjust the estimate.
- A mix, such as cereal with milk and fruit: scan the packaged parts and estimate the fresh parts.
How Luca does it
Luca is an AI calorie tracker for iPhone, coming to the App Store. You can log by scanning a barcode, by photo, by voice or by typing a description. Barcode and manual entry are unlimited on the free plan. See logging packaged food by barcode and features.
This article is general information, not medical or dietary advice.
Sources
- The open food database (home page)
- The open food database: terms of use, contribution and re-use
- The open food database: data and conditions for reuse
- The open food database: about the project
- Open Data Commons, Open Database License (ODbL) summary
- 21 CFR 101.9, Nutrition labeling of food (US Code of Federal Regulations, 2024 edition)
- European Commission, Guidance document on tolerances for nutrient values declared on a label (December 2012)
- European Commission, Nutrition labelling
- FDA, Serving Size on the Nutrition Facts Label
- NHS, How to read food labels