House Formulation Doctrine

How a recipe gets designed here — the ratios that have worked, why they work, and the rails a new formula stays inside. This is what formulate reasons from. It is taste plus six batches of evidence, not chemistry law; a departure is allowed, but it's a decision, not a default.

The backbone is Bastille — a soap that is mostly olive oil (a true Castile is 100% olive). Olive makes a mild, conditioning, hard-when-cured bar that's gentle enough for daily use. The rest of the formula is small levers on top of that.

The two house blends

Everything to date is one of these two, and they cover the whole use-case range so far.

Body bar — 72 / 18 / 10 at 5% superfat

Olive 72% / coconut 18% / castor 10%. The default for a bar that isn't washing hands ten times a day. High olive for mildness, 18% coconut for enough lather and cleansing without stripping, 10% castor for creamy bubbles. Batches #1–#4.

Hand bar — 62 / 28 / 10 at 6% superfat

Olive 62% / coconut 28% / castor 10%, superfat bumped to 6%. A frequently-used hand bar needs more cleansing and a bubblier lather, which is the 28% coconut — but 28% coconut alone would start to dry skin, so the extra superfat cushions it back. The two moves are a pair; don't make one without the other. Batches #5–#6.

The levers, and their rails

• Olive is the backbone — 62–72% here. Down toward 62 makes room for more coconut; up toward 72 makes the mildest bar. This is the balance you actually turn.
• Coconut is the cleansing/lather dial — 18–28%. Below 18, lather goes flat; above ~30% it reads as a degreaser on skin. 28% is the house ceiling and it's the reason the hand bar runs 6% superfat.
• Castor stays pinned at 10%. It's a small lever with real effect on lather creaminess; above ~15% bars turn sticky. No batch has had a reason to move it.
• Superfat is the free-oil safety margin — 5% for body, 6% for high-coconut. Higher is milder but softer and shortens shelf life; lower is harder and more cleansing but less forgiving of any lye error. On a 1g scale, 5–6% also keeps a comfortable margin against rounding.
• Water at 38% of oils is the house constant. More water is a fluider batter and a longer cure; less accelerates trace and firms faster. It also moves the fill math, so it isn't a free knob — tools/lye.py re-checks fit whenever it changes.

Fragrance

Load rates, all as a percentage of oil weight, from what's actually been poured:

• Light ~2% — a whisper of scent, or when the fragrance browns and you want to keep the bar pale. (Batch #5.)
• Moderate ~3% — essential-oil batches, where more is expensive and fades anyway. (Batch #1.)
• Strong ~6.5% — fragrance oils where the scent is the point. (Batches #2, #3, #4.) This is about the ceiling for these; most FOs list a max around 5–6% for leave-on-adjacent use.

formulate pulls the real per-oil data from inventory/fragrances.toml. Three things there change a design:

• Vanillin browns the bar. High-vanillin fragrances (Vanilla Bean, the gourmands) turn a bar tan-to-dark-brown over cure. That's not a defect to avoid — it's a look to design around, and it pairs naturally with a kraft-paper wrap. But it rules out a pale or white bar.
• Acceleration steals your working time. An accelerating fragrance (Coconut + Praline) means clay and scent fully staged before trace, a hand stir instead of the stick blender, and a fast pour. Don't pair it with an intricate technique.
• Amount on hand caps the batch. A nearly-spent bottle (right now: Vanilla Bean ~9g, Sensuous Sandalwood ~7g, Coconut + Praline ~5g) either sets the batch size or gets blended with something else. formulate reads the ledger so it won't spec 44g of a 7g bottle.

How formulate uses this

1. Mold + use case → blend and superfat. Hand-washing picks the hand bar; a body/shower bar picks the body bar. A new use case is a real design conversation, not a default.
2. Mold + target fill → oil weight, via tools/lye.py --fit. Aim for ~92–95% of a mold's ceiling — full bars, real margin against overflow. Never design to 100%.
3. Fragrance choice → check the ledger for amount, vanillin, and behavior; set the load rate to taste and to what the bottle holds.
4. Everything → tools/lye.py for the actual weights, fill, and yield. Never by hand.
5. Brian cross-checks on SoapCalc. Only then is it Ready.

Room to grow

Directions the palette could take, noted so they're on the radar — none tried yet, each a real design conversation when it comes up:

• A new oil (shea, cocoa butter, sweet almond) means adding a validated SAP value to reference/sap-values.toml first — verified against SoapCalc, never guessed — and probably its own small run before it joins a real batch.
• Coffee grounds as an exfoliant/deodorizer are staged in the staples but have never been used; the first batch that includes them is doing something genuinely new.
• A distinct kitchen bar (higher coconut, coffee grounds, maybe a lower superfat since it's washing dishes not skin) would be a third blend worth formalizing once it's proven.