Corrections: Every Number We Got Wrong and Fixed

Our estimate for 31 packaged products assumed more seed oil than the product's own label allows. Each label prints the percentage of its ingredients, so we recalculated from those printed percentages.

Why: The label's own printed percentages are better evidence than our guess. The rows still carry the Estimated tag.

The Egg and bacon muffin had been scaled twice by mistake and went from 5.5 g to 4.1 g. The Very Berry smoothie went from 2 g to 1.3 g. The three sizes of the Cherry Pomegranate Refresha now use the 0.1 g of fat the chain prints.

Why: A re-read of the chain's own figures found our arithmetic was wrong on the muffin and too high on the drinks.

Where a chain left the saturated fat cell blank or printed it as zero, our ceiling treated nearly all of the item's fat as polyunsaturated. Those 350 rows are now capped at 80% of the fat the chain prints, and the receipt says so.

Why: No real cooking fat is more than about 80% polyunsaturated, so anything above that was physically impossible.

Eleven rows showed 0 g because the chain's table prints saturated fat equal to total fat, which left no room for anything else. They now show our estimate from the item's total fat and are labeled Estimated.

Why: A zero built on a misprinted cell tells people an item is clean when we do not know that.

We checked the home cooking table against USDA figures. Salmon went from 3 g to 8 g. Pork, lamb, eggs, wings and tofu went up. Shrimp, raspberries and sunflower butter went down.

Why: USDA has measured these foods. A measurement beats our estimate in either direction.

The Keg's BBQ Rib Dinner and 17 Cinnabon Canada rows were marked as coming from the chain. They are estimates. Both chains came off our fully documented list.

Why: A row only counts as documented when the chain itself publishes the figures behind it.

Plain black coffee, espresso, Americano and unsweetened tea with no milk now show 0 g, using USDA's measured figures. The receipt says the number is USDA's and not the chain's.

Why: USDA measures these drinks at under 0.1 g of polyunsaturated fat a serving, which rounds to zero.

Five UK chains were listed twice under two spellings, such as Gail's and Gail's Bakery. Each pair is now one entry under the name the chain uses, keeping the newer data. Crussh (UK, closed April 2025) and Relish Gourmet Burgers (Canada, closed) were retired.

Why: Two entries for one chain gave two different answers to the same question.

2,377 grocery products showed 0 g only because the label left the figure blank. One store-brand mayonnaise sat at 0 g next to the same brand's mayonnaise at 6.9 g. A zero is now ranked only when the product's name says fat free, and servings under 3 g are never ranked.

Why: A blank on a label is not a zero. Ranking those products as the cleanest was wrong.

45 whole pizza rows carried the figure for one slice, 0.8 to 3.2 g. Using the slice counts in the chain's own guide, they now show the whole pie, 11 to 53.7 g.

Why: The row names a whole pizza, so the number has to be for a whole pizza.

Four sauce receipts cited Pei Wei's own nutrition guide, but the figures had come from a third party nutrition website. The receipts were removed and the rows went back to Estimated.

Why: A third party site is not the chain. We only call a row documented when the chain publishes it.

The medium fries were calculated on a generic vegetable oil while the small and large used canola, the fryer oil Chick-fil-A itself names for its fries. That made the medium read higher than the large. All three sizes now use canola.

Why: The oil a chain states for its own fryer beats a generic assumption.

Eight items were stored twice with two different numbers. The 20 piece Boneless Wings showed both 32 g and 16 g. Each pair is now one row, recalculated from verified nutrition figures. Small orders came down and large orders went up.

Why: Two answers for one item means at least one of them was wrong.