Every formula, and where it comes from
The calculator is a thin layer over a small, open formula library. This page is that library written out. Conventions are labelled as conventions; constants carry their source.
- Saponification and lye
- Water, three ways, and master batches
- Acids and fatty liquids
- Deriving a saponification value from composition
- How well the published data pins the lye down
- Fatty-acid profile, soap qualities, iodine and INS
- Fragrance load, IFRA and allergens
- Mould sizing
- Liquid soap dilution
- Where a derived saponification value comes from
- Why the lye carries a range
- Conventions we label as conventions
On this page
- Saponification and lye
- Water, three ways, and master batches
- Acids and fatty liquids
- Deriving a saponification value from composition
- How well the published data pins the lye down
- Fatty-acid profile, soap qualities, iodine and INS
- Fragrance load, IFRA and allergens
- Mould sizing
- Liquid soap dilution
- Where a derived saponification value comes from
- Why the lye carries a range
- Conventions we label as conventions
Saponification and lye
Water, three ways, and master batches
Acids and fatty liquids
Deriving a saponification value from composition
How well the published data pins the lye down
Fatty-acid profile, soap qualities, iodine and INS
Fragrance load, IFRA and allergens
Mould sizing
Liquid soap dilution
Where a derived saponification value comes from
40 oils in the database have no saponification value in any primary document we could use. Rather than leave them out of the calculator or invent a number, the figure is calculated from the oil's own measured fatty-acid composition. That is not a guess: three moles of hydroxide saponify one mole of triglyceride, and the molar mass of the triglyceride follows from the chain lengths of the acids in it. Composition determines the value.
The question is how accurate it is, and that is measurable. The fair test is the Codex subset of the database, 36 named oils where the saponification range and the fatty-acid ranges were published by the same body from the same material. Against those, the method's median absolute error is 0.84 %, its ninetieth percentile 2.44 % and its worst case 3.57 %. It runs slightly high, by 0.55 % on average: it assumes pure fat, while a crude oil carries a little unsaponifiable matter that lowers the measured value. High is the lye-heavy direction, and the band a derived value carries covers it. The median published range in the same database is ±2.9 % of its own centre, so a derived value is about as well determined as one read off a standard.
Measured against the rest of the database the apparent error is larger: a median of 1.7 % and a worst case of 22 %. That is not the method failing. The largest disagreements are the same records the consistency check already flags as internally contradictory, where a published saponification value sits beside a composition that cannot produce it. Comparing against those conflates method error with data error, which is why the validation uses the Codex subset alone.
Every derived value is labelled as derived wherever it appears, carries a ±3 % band into the uncertainty panel, and is set aside entirely the moment you enter a measured value from your own lot.
Why the lye carries a range
A saponification value is not a constant. Codex gives coconut oil 248–265 mg KOH/g because the fat genuinely varies with origin, season and refining. Every soap calculator hides that by printing one number to one decimal place, which is false precision: the figure looks like a measurement of your oil and is really the midpoint of somebody else's survey.
Carried through, the band lands on the number that matters. Superfat is the margin between a mild bar and a caustic one, and it is exactly what the uncertainty eats. A recipe asking for 3 % superfat on ordinary olive and coconut oil can really be anywhere from −0.2 % to 6 %, and the negative end is a lye-heavy bar. That is not a reason to distrust the arithmetic; it is the reason the craft settled on about 5 % superfat, and as far as we can tell no calculator has ever shown the working behind that habit.
Taking every oil at the same end of its range at once is a bound, not a confidence interval. Real lots are not perfectly correlated, so the achievable spread is narrower, and the arithmetic says nothing about how likely either end is. The only thing that collapses it is a measurement: enter the saponification value from your lot's certificate of analysis and that oil stops contributing any uncertainty at all.
Conventions we label as conventions
The five "soap qualities" are sums of fatty-acid groups compared with bands that hobby calculators have used since the mid-2000s. They are useful for comparing recipes and useless for predicting an actual bar: they ignore superfat, water, cure, additives and process, and treat every acid in a group as equal. The 38 % "full water" default and the "× 0.4" mould rule are habits, not chemistry. We implement all of them exactly because makers compare in that language, and we say what they are.
Hear when something new ships
A short note whenever something new ships. No newsletter, nothing sold, one click to stop.