Methodology

This page explains how every figure on this site is sourced, checked and either published, range-published or rejected. It is written so the process is auditable: a reader can pick any ingredient, follow its sources back to the original records, and verify that the rules were applied correctly.

Sourcing rules

Every published figure requires at least two independent sources. For an entry to reach high confidence, three sources must agree within the category tolerance and at least one of them must be a standards body (Codex Alimentarius, FDA identity standards, or equivalent) rather than another baking chart.

The primary source for most entries is USDA FoodData Central, cited by FDC number. Where an FDC record provides a cup weight directly, that figure is used. Where it provides only a per-100g or per-tablespoon figure, the cup weight is derived by extrapolation and the derivation is noted. The secondary source is typically a published culinary reference (King Arthur Baking Ingredient Weight Chart, America’s Test Kitchen), and where possible a standards body figure is added as an independent check.

A figure is only published if it matches one source exactly. Averages are never published. Where two sources agree within the category tolerance, the entry publishes whichever is the primary source’s figure and names both. Where they disagree beyond the tolerance, the entry publishes the range they cover and names the method behind each.

Category tolerances

Category	Tolerance	Why
Fats and oils	5%	Continuous-phase substances that flow. Two sources measuring the same oil should agree closely.
Sugars and salts	5%	Crystalline solids pack reproducibly by scoop.
Milled powders	10%	Powders clump, bridge and pack differently by fill method. Narrower tolerance would mean publishing on easy substances only.
Liquids	5%	Self-levelling removes the fill-method variable.
Syrups	5%	Dense concentrated sugar solutions. Density is uniform across brands.

Physical rejection

Some source figures fail on physics rather than on tolerance. Where a substance’s composition dictates a density (a mostly-water product with dissolved solids denser than water must itself be denser than water), a source figure implying a lighter density is rejected rather than range-published. Where an independent physical standard exists (Codex Alimentarius for oils, FDA identity standards for dairy), a source figure outside that standard is rejected on the standard.

Rejected sources are recorded, not deleted. Every ingredient page shows every source consulted, including rejected ones, with the reason for rejection. The corrections page lists every rejection currently in the dataset. This transparency is what distinguishes rejection from silent deletion: a reader can check the rejection reasoning and challenge it if they think the physics is wrong.

Confidence grades

Grade	Requirements
High	Three independent sources agree within tolerance, one of which is a standards body (Codex, FDA) rather than another baking chart.
Medium	Two sources agree within tolerance, or one source is accepted after another was rejected on physics.
Low	Range published because sources disagree beyond tolerance, or only one source is available, or method-based state splits exist without a clear default.

Method splits

Some ingredients are measured differently in different contexts and both methods produce genuinely different weights. All-purpose flour dipped-and-swept weighs about 20% more than spooned-and-swept. Cake flour follows the same pattern. Where a source explicitly names the method behind its figure, the entry is split into states, and each state carries its own sources and figure. Where no source names a method, the figure sits in an unspecified state and is treated as a single measurement.

The measurement cup

All conversions on this site use the US customary cup of exactly 236.5882365 mL. This is derived from the definition of the US gallon as 231 cubic inches and the inch as 25.4 mm, so it is an exact decimal rather than a measurement. The tablespoon and teaspoon are derived from it: exactly 1/16 and 1/48 of a cup respectively.

This is not the same as the FDA labelling cup (240 mL), which appears on nutrition facts panels and is used by some sources without saying so. A cup weight computed against 240 mL will read about 1.4% high compared to a US customary measuring cup. Where a source figure appears to be off by roughly 1.4%, this is often the reason, and the entry notes it.

Review dates

Every source on every entry carries a last_reviewed date, updated when the figure is re-checked against its source. The full dataset is at data.json under a permissive CC BY 4.0 licence.

Corrections and suggestions

If a figure here is wrong, if you know a source that would settle one of the open entries, or if a rejection reasoning is faulty, the About page has contact details. A source not yet consulted is worth more to this project than another page is.