Back to Journal

Where Do Food Scanner Apps Get Their Product Data?

Published · Updated  · Sláinte Editorial

Scanning a barcode is easy.

Maintaining the information behind millions of barcodes is the difficult part.

Illustration for Where Do Food Scanner Apps Get Their Product Data?

A useful food-scanning database needs to know what product the barcode represents and then populate fields such as product name, brand, nutrition, ingredients, allergens, pack size and market.

That information can come from several places.

Manufacturer information

The manufacturer is an obvious source.

Official product pages can provide ingredients, nutrition, pack size, allergens and product descriptions.

But even an official webpage may lag behind a physical recipe change.

That is why the exact product and date still matter.

Retailer listings

Irish supermarket websites provide useful product records.

Tesco Ireland, SuperValu, Dunnes, Aldi and Lidl all expose different levels of information online.

One retailer may show the complete nutrition panel. Another may show only basic product details.

A retailer listing is useful evidence. It is not automatically complete evidence.

Public product databases

Open Food Facts is a collaborative open product database.

Its openness means the data can be inspected and reused. That is valuable for the wider food-tech ecosystem.

Crowdsourced data still needs quality checks. The presence of a product does not guarantee every field is current.

Pack photographs

The physical product can be one of the most useful sources.

A clear pack photograph can establish:

  • exact barcode
  • exact ingredients
  • current nutrition panel
  • warnings
  • pack size

The limitation is that the photo itself also needs context.

When was it taken? Which market was it from? Is the pack still current?

Automated extraction

Software can help turn webpages or photographs into structured data.

That speeds up database building. It also creates another reason for validation.

A parser can associate a number with the wrong heading. OCR can misread small text.

Automation is useful precisely because it allows scale. It should not remove uncertainty from the process.

Human review

Some product records need manual review, especially when:

  • two sources disagree
  • ingredients have changed
  • a value looks implausible
  • the pack is unclear
  • two products share similar names

Why can two sources disagree?

Because product information changes.

The possibilities include:

  • one source is older
  • different markets use different formulations
  • one is a different pack size
  • a transcription error occurred
  • a website has stale data

The right response is to investigate, not to silently choose whichever number produces the preferred score.

Updated record vs updated recipe

A database record edited yesterday could still have copied information from an old package.

Those dates mean different things.

A useful provenance system should distinguish between:

  • source date
  • review date
  • product market
  • pack version where known

What should “verified” mean?

It should mean something specific.

“Verified” should not imply “We sent this food to a laboratory” unless that literally happened.

For a product database, verification may mean the fields were checked against a source.

The interface should make that distinction understandable.

How Sláinte helps

Frequently asked questions

Is manufacturer information always current?

Not necessarily.

Are retailer websites reliable?

They can be useful first-party retail sources, but the current physical pack remains important.

Is Open Food Facts crowdsourced?

Yes. It is a collaborative open product database.

Does a recently edited record mean the recipe is recent?

No.

What should I do if the scanner disagrees with the pack?

Treat it as a possible outdated or incorrect record and use the current pack as important evidence.

The short version

  • Food scanners need databases behind the barcode.
  • Data can come from manufacturers.
  • It can come from retailers.
  • It can come from open databases.
  • Pack photographs can help verify formulations.
  • Automation speeds up data collection but can make mistakes.
  • Recently edited does not mean recently formulated.
  • Source provenance matters.