Methodology & data sources

Cups to grams — ingredient densities

A cup is a measure of volume, so converting it to grams needs a density for each ingredient. Our default is the King Arthur Baking ingredient weight chart — the reference most US test kitchens treat as standard. Flours assume the spoon-and-level method (all-purpose flour = 120 g per cup); brown sugar is measured packed, as recipes intend. Volume math uses the US customary cup of 236.588 ml — what American recipes mean. (The 240 ml "legal" cup is a nutrition-labelling convention, under 21 CFR 101.9.) Volume↔volume and weight↔weight conversions are ingredient-independent; only volume↔weight uses the density.

How you fill the cup matters more than any rounding here. King Arthur's measuring guide puts a fluffed, spooned and leveled cup of all-purpose flour at about 120 g, and a compacted cup scooped straight from the bag at up to 160 g — up to a third more flour from the same written recipe. Weighing removes that error entirely.

Where we deliberately differ. King Arthur lists every liquid at 227 g per cup, which comes from reading 8 fluid ounces as 8 weight ounces. That is convenient but physically wrong — a cup of water weighs about 237 g, and whole milk, being denser than water, about 244 g. For eight of our 44 ingredients (honey, water, milk, buttermilk, heavy cream, baking soda, instant yeast, vanilla extract) we use a better-grounded figure than the chart's — such as a physical density, a USDA FoodData Central value, or the weight of a standard 7 g yeast packet — and each one names its source on its page. A test asserts that every other value matches the chart exactly, so nothing can drift from its source unnoticed.

Cake pan converter — capacities & batter depth

Pan capacities are brim-full volumes from Joy of Baking's pan size chart, which lists approximate brim-full volumes; to check your own pan, it says to pour water in by the cupful until the pan is full to the brim. Three sizes that chart doesn't list are taken from its nearest rows and labelled on their pages: the 8×3″ round is its 8×2″ round (6 cups) at 1½× the depth, the 9×3″ round matches its 9×3″ springform (12 cups), and the 13×18″ half sheet is its 12½×17½×1″ pan (12 cups) scaled up by bake area. Custom sizes use geometric volume (1 US cup = 14.4375 cubic inches). The ingredient multiplier is the ratio of the two pans' capacities — a volume comparison, not a flat area one, which is why a deeper pan isn't treated the same as a wider one. Time and temperature guidance is driven by batter depth (volume ÷ area) assuming the standard two-thirds fill: deeper batter bakes longer at a lower temperature, shallower bakes faster. Always confirm with a toothpick.

Cake serving calculator — the Wilton chart

Serving counts are transcribed from Wilton's Cake Baking & Serving Guide: party servings are a 1½ × 2 in slice, wedding servings a 1 × 2 in slice. Wilton's counts are for a two-layer cake about 4 inches tall. Height changes them less than people expect: per Wilton, cakes 3 to 6 inches tall baked in the same pan serve the same number, because slices are measured across the top — a taller cake gives taller slices, not more of them. A cake under 3 inches tall — one 2-inch layer, which is what most home sheet cakes are — yields about half. Where we show a single-layer figure, it is the chart count halved and rounded down. Wilton is our source for servings, batter per layer and icing amounts only; pan capacities come from the chart above.

Frosting calculator — a stated model, cross-checked

Frosting is computed, not copied: the cake's frosted surface (top, sides and the fillings between layers) × a coat thickness you can see — thin ⅛″, standard ¼″, generous ⅜″ — plus 25% when you add piping. Wilton's guide also publishes approximate cups of icing to ice and decorate each size, which it describes as very general; we use that column as the cross-check. For a two-layer cake at the standard coat with no piping:

Ours is rounded up to the next quarter cup, because running out mid-crumb-coat is the expensive mistake. Turning piping on adds the 25% allowance, which puts the model above Wilton's single figure.

Recipe scaler — fractions & safe limits

The scaler multiplies each ingredient by your factor and rewrites the result in fractions you can actually measure (it understands mixed numbers like "2 1/2", unicode fractions like "¾", and decimals). Lines with no quantity ("pinch of salt") pass through untouched, and fractional eggs are flagged with a practical fix. It warns beyond 3× because big multiples stop behaving like the original: one bowl mixes less evenly, and salt, strong spices and leavening can come out stronger than straight multiplication predicts. Mix in rounds of 3× or less where you can; if you scale in one go, bake a test batch and adjust leavening and seasoning to taste. We don't print a fixed reduction factor, because we haven't found a primary source for one.

Oven temperature converter

Fahrenheit↔Celsius use the exact formulas (°C = (°F − 32) × 5/9). Gas marks follow the standard UK gas-mark table (gas mark 4 = 350 °F / 180 °C, "moderate"). The fan column applies the UK convention of 20 °C lower than a conventional oven — the Which? oven conversion table gives fan temperatures "typically around 20°C less". A 20 °C difference is 36 °F (a temperature difference converts at × 9/5, with no + 32). A fan oven isn't hotter than its dial: it cooks faster at the same setting because moving air transfers heat into the food faster. US convection ovens follow a separate, smaller rule — LG, for example, says to reduce the recipe temperature by 25 °F (14 °C) for Convection Bake, and many US ovens make that conversion automatically. Your oven's manual overrides both.

High altitude calculator — Colorado State Extension

Adjustments are transcribed from Colorado State University Extension's High Altitude Food Preparation. Its adjustment table is for cakes ("Cake-recipe adjustment for high altitude"), in three bands from 3,500 to 10,000 ft, and we keep every figure as the range CSU publishes. CSU says most cake recipes perfected for sea level need no modification up to 3,000 ft; between that and the table's first band we point you to the first band and its smaller adjustment. Yeast breads, biscuits and muffins follow CSU's separate advice, not the cake table, and above 10,000 ft we say the table stops rather than extrapolating it. Water's boiling point comes from CSU's Table 1 (0 ft 212 °F, 2,000 ft 208 °F, 5,000 ft 203 °F, 7,500 ft 198 °F, 10,000 ft 193 °F); for altitudes between those published points we interpolate in a straight line, so those in-between values are our estimate, not CSU's.

What these tools deliberately don't cover

CrumbCalc does one thing: the measurement and proportion maths of baking — weights, volumes, pan capacities, ratios, and temperature scale conversions. Being clear about the edges of that matters as much as the figures inside it, so:

• We publish no food-safety thresholds. Nothing here tells you a safe internal temperature for meat, poultry, eggs or leftovers. For those, use FoodSafety.gov's safe minimum temperature charts — a government source that is maintained, which a baking calculator should not pretend to be. Where a tool mentions an internal temperature it is a doneness cue for a bake (a cake is set around 200–210 °F), never a safety threshold.
• We publish no canning or preserving guidance. Home canning is the one kitchen task where getting a number wrong can be genuinely dangerous rather than merely disappointing, and its processing times depend on jar size, acidity and altitude together. Our high-altitude pages deliberately stop at baking and stove-top work; for preserving, follow the National Center for Home Food Preservation.
• Our numbers can't harm your oven. Every oven temperature we publish sits inside the normal baking range (225–500 °F / 110–260 °C), well below the temperatures an oven reaches during its own self-clean cycle, and every oven has its own thermostat and thermal cut-outs. A conversion that's off by a few degrees costs you a disappointing bake, not a damaged appliance. We give no electrical, appliance-repair or equipment-modification advice of any kind.
• Estimates say so on the page. Where a figure is our own model rather than a transcription — frosting amounts, geometric pan capacities, interpolated boiling points — we show the assumptions, label it as an estimate, and cross-check it against a published figure wherever one exists, rather than copying an unsourced chart or inventing a citation.

One honest caveat that outweighs all our rounding: your oven is probably the least accurate instrument in this chain. Home ovens often run hot or cold and swing as the thermostat cycles. Published gas-mark charts also disagree with each other by 5–10 °C, because gas marks are defined in Fahrenheit (gas 1 = 275 °F, then +25 °F per mark) and every publisher rounds the Celsius column differently — ours stays within 5 °C of the exact conversion, and a test enforces it. An oven thermometer settles all of it for the price of a coffee.

How we keep it accurate

The figures these tools rely on are pinned by an automated suite of over 340 checks that runs before anything ships. Every one of the 44 ingredient densities is asserted against the King Arthur chart value, and the eight that deliberately differ must carry a written reason — a value cannot quietly drift from its source, and a departure cannot go undocumented. Beyond straight transcription, the checks test the derivations: that Wilton's wedding counts for round and square cakes really are the cake's area divided by a 1 × 2 in slice (exactly for squares, to within a couple of servings for rounds; Wilton's sheet-cake counts run below that (Wilton doesn't say why), so they are checked against the chart value by value instead), that a party slice never feeds more people than a smaller wedding slice, that pan geometry stays within a fifth of published capacity, and that a cup of water weighs what physics says it does. Testing the rule rather than re-stating the number is what catches a typo; simply asserting the values we shipped would not.

Between you and that reference there is no server and no AI guessing a number — the engine runs entirely in your browser, so what you see is the arithmetic on the cited data, nothing more. Spot a figure that looks off? Tell us and we'll check it against the reference and correct it.