How the Seed Oil Tracker numbers are made

1. Estimating PUFA per menu item

Seed oils are a concern because of one fat in particular: polyunsaturated fat (PUFA), mostly linoleic acid. Almost no restaurant or label reports PUFA directly, but they do report total fat, and the cooking oil is known. We estimate a food's PUFA as:

PUFA grams = total fat grams × the oil's PUFA share

The PUFA share of each fat comes from USDA oil-composition data. When a food could use a regular or a high-oleic variant of an oil, or names several fats, we use the higher-PUFA value. That keeps every estimate conservative and matches the app's purpose: flagging seed oils rather than underplaying them.

Representative values; the full table lives in the app's open-source estimator. High-oleic varieties (bred to be mostly the stable fat in olive oil) can be far lower and are noted per chain where known.

2. Grading a chain

A chain's grade is the average estimated PUFA per menu item across the items we have scored for it. Lower is better:

Averaging across the menu means a chain is not judged by its single worst or best item, but by the typical choice a customer faces.

3. Sourcing the cooking-oil facts

The oil a chain fries and cooks in is the biggest driver of its score, so we track it per chain. Those facts come, in order of preference, from: the chain's own published ingredient and allergen documents; official customer-service answers; and, where a chain does not disclose, the documented US industry default (soybean or canola) marked plainly as an assumption. Each chain's oil is shown on its report card and in the open dataset. When a chain uses different oils for different items (for example peanut oil for chicken but canola for fries), we record that split rather than flattening it.

4. Validation and honesty

Two ways we keep ourselves honest:

• Where a food carries a measured PUFA value (from USDA FoodData Central or Open Food Facts), we use the measured number instead of the estimate.
• Our estimator is checked against USDA oil-composition data, and the whole dataset runs an automated audit that flags implausible values before publication.

We also keep a balanced Evidence page that includes the studies which disagree with the seed-oil-harm case, because a source you can cite has to show both sides.

5. Limits

These are estimates, not lab assays. Preparation, portion size, and cooking oil vary by location and change over time; a chain can be clean on the fryer and not on the sauces. Treat the numbers as directional guidance for eating out, and tell us when something looks wrong so we can fix it.