Who really owns the family farm on your cereal box?

Method

How this was built

How brands were chosen

A brand is in the dataset if its name, its packaging, or its own "our story" page leans on a farm, a family, a homestead, a place of origin, or a named founder. The signal has to be something the brand itself puts forward. A company that happens to be family-owned but never says so does not qualify, and a company that says "family-owned since" does qualify whether or not it still is.

The rule is deliberately blind to who owns the brand. The obvious alternative, picking "brands that look small but are not", would build the answer into the selection by excluding every brand that looks small and is small. Those brands are in here, and they are the reason the finding is worth anything.

Scope is the breakfast aisle only: cold cereal, granola, oatmeal and hot cereal, breakfast bars, and pancake and waffle mix. Syrup, yoghurt, coffee, and frozen waffles are breakfast, but they are not this aisle.

How ownership was verified

• Two independent sources per claim. Where possible one of them is the parent company's own material: its brands page, its investor page, or its annual report. The second is independent of the first: a deal announcement, trade press, a filing, or Wikidata.
• Nothing from memory. Brand ownership moves through acquisitions, spinoffs, and private equity sales. Every row was looked up.
• A Wikidata cross-check. A script queries Wikidata for each brand's "owned by" and "parent organization" properties and compares them with the dataset. A disagreement is not treated as an error and is never auto-corrected. Wikidata is volunteer-edited and goes stale, so a disagreement is a prompt to recheck the row by hand, and the decision is written into that row's notes.
• Failures stay visible. A row that could not get everything it needed is kept and flagged, rather than dropped. Dropping it would hide the weakest part of the dataset.
• Ownership is as of the verified date on each row, shown in the table and summarised in the footer.

How the page works

The dataset is a hand-curated CSV in the repository. A Python script validates it and cross-checks it against Wikidata. A second script turns it into one JSON file, which is committed. This page reads that file and makes no API calls of its own, so it keeps working regardless of what happens to any service it was built from.

Limits

What this does not show

• Ownership changes. These rows are a snapshot on the dates shown. A brand can be sold the week after it is checked, and private equity holdings in particular turn over fast.
• A big owner does not mean a worse product. Nothing here measures quality, ingredients, farming practice, or how workers are treated. Ownership is a fact about a corporate structure, and that is all it is.
• Brand origin is the brand's own claim. The origin city is what the brand says about itself on its own page. It is not a check on where anything is grown, milled, or made today.
• This is one aisle, not the food system. Around fifty brands in five breakfast categories is a sample chosen by a branding rule. The concentration measured here is concentration within this sample, not a market share and not a claim about groceries in general.
• The ownership chain is simplified. Real structures include intermediate holding companies, minority stakes, licensing deals, and regional rights that can differ by country. The dataset records a direct owner and an ultimate parent, and the tree uses the ultimate parent.