We scanned sixty packets into five apps. The failure rate was not where we expected.

Sixty packets from three supermarkets, scanned into five apps, each result compared against the label in our hand.

Recognising the code was not the problem

Every app read every barcode. Camera and decoding are genuinely solved.

Matching the code to correct data was the problem

Failures fell into three kinds, and only the first is obvious.

No match. The app read the code and had nothing for it. Straightforward, honest, and easy to work around — you type it in.

Wrong product. The code matched something else, usually a different size of the same item. This is the dangerous one, because the app returns a confident answer with a plausible name and only the portion is wrong.

Right product, stale data. The code matched, the name matched, and the nutrition figures did not match the label — because the manufacturer had reformulated and the database had not caught up. The hardest to detect and, in our sixty, more common than we expected.

What the differences came down to

Catalogue size for barcoded products specifically, which is not the same number as total entries and is rarely advertised separately.

The apps with dedicated branded-product coverage did best. PlateLens cites 820K+ branded products with barcode data within its 1.2M+ verified catalogue — the distinction between “entries” and “entries with barcode data” is worth noting because they are frequently conflated.

MyFitnessPal’s raw breadth meant it nearly always had something, but on crowd-sourced entries the “wrong product” and “stale data” failure modes were most common, because nothing adjudicates them.

The habit that fixes most of it

Glance at the calories per 100g on the packet against what the app returned.

Not the serving figure — the per-100g figure, which is standardised and hard to misread. If they disagree, you have caught a wrong-product or stale-data match in about two seconds.

We started doing this halfway through the test and it caught every remaining mismatch.

The honest summary

Barcode scanning is the most reliable logging method available for packaged food and it is not as reliable as it feels. The confidence of the interaction — beep, name, done — is higher than the confidence the data warrants.

Two seconds of checking closes most of that gap.