How to build a perfume accord

An accord sits between a material and a perfume, and it is where most of the craft actually happens. A material is a fact. A perfume is an argument. An accord is one idea, stated in the smallest number of materials that can state it, and learning to build one is most of learning to compose.

What is a perfume accord?

One idea. Not a small perfume, not a pyramid in miniature, one idea: warm amber, salt on stone, tea with jasmine still in it. The working test is that you can name the idea in a few words before you weigh anything. If you cannot name it, you are not building an accord yet, you are still auditioning materials.

The second half of the definition carries as much weight as the first: the smallest material count that states the idea. An accord is finished when removing any line changes the idea and adding any line does not improve it. That is a harder discipline than it sounds, because adding a material always feels like progress and almost never is.

How many materials should an accord have?

Most working accords live at three to five materials, and the reason is structural rather than aesthetic. Each line has to be audible, and each line has to have a job you can say out loud:

• One line carries the idea. This is the material you would keep if you could keep only one. It usually takes the largest share.
• One or two lines shade it. They push the centre toward the version of the idea you actually want: greener, saltier, older.
• One line anchors it in time. Materials leave a strip at very different speeds, and without something that stays, the idea evaporates before it has finished being said.

Two materials is usually a pairing rather than an idea. Past five or six, you stop being able to hear what each line is doing, and an accord you cannot hear line by line is an accord you cannot fix.

Why do fewer materials beat more?

The commonest beginner accord is not too strange, it is too crowded. Three materials doing one job do it at a third of the conviction each. Put three woods in to be sure the wood reads, and none of the three is dosed high enough to have a character, so what reads is not wood, it is the average of three woods, which is mud. The blend gets smoother and smoother and says less and less.

The fix is a decision, not a dose change. Audition the candidates for a job separately, pick the one that says it best, and cut the others entirely. A material at a committed dose has a personality; the same material at a hedge dose is noise. If you genuinely cannot choose, that is a sign the idea itself is not named sharply enough yet.

How do I dose an accord into a formula?

An accord is built at 100% within itself: its lines are ratios to each other, not to any finished perfume. That is what makes it reusable. When it goes into a concentrate it travels as one decision, at one dose, with its internal ratios held fixed.

1. Settle the internal ratios first, on strips, before the accord meets anything else.
2. Dose it into the concentrate as a single unit. Trial more than one dose; an accord that is the heart of one perfume can be a supporting shade in another.
3. When something is wrong, change the dose of the whole accord before you touch its insides. If the idea is right but too loud, that is a dose problem, not a formula problem.

One piece of arithmetic will quietly ruin this if it is done by hand: dilutions. If your only linalool is a 10% solution, a gram of it is a tenth of a gram of idea, and an accord ratio built without correcting for that is not the ratio you think it is. PerfuMate corrects every dose for the strength of the bottle it is actually weighed from, so the ratio you settled on strips is the ratio that lands in the concentrate.

How do I test an accord over time?

An accord approved on a fresh strip is only half approved, because its materials do not leave together. The idea you smell at minute one and the idea left at hour eight are made by different subsets of the same lines, and both have to be the idea you named. So the test is a clock, not a sniff: label the strip, date it, and come back at intervals through the day and again the next morning. Note when each line drops out, and whether what remains still says the thing.

If the accord falls apart in the middle hours, the usual cause is a gap in speed: everything is either fast or slow and nothing bridges them. That is a case where a fourth or fifth line earns its place.

How do I build an accord in PerfuMate?

PerfuMate treats the accord as a first-class build, and the doctrine above is wired into the tools rather than left as advice.

• Ask for an accord, capped. AI Studio's brief-to-formula build composes an accord as readily as a perfume: the brief carries a perfume-or-accord switch and an ingredient cap, so you can ask for the idea in a handful of lines and the cap travels with the request. Every returned line is a material from your own shelf with a percentage and a role, shares are renormalised to exactly 100% in code, and the answer is validated against your shelf before you ever see it.
• Pick candidates from described rows. Every one of the 1,581 catalog materials arrives with a written odour description and an impact figure, the odorants among them with olfactive families in dominance order, and every row resolves to a substantivity estimate, which is exactly the information the audition step needs. The catalog is free to browse without an account.
• Hear the accord over time before the strip does. The drydown is a 0 to 24 hour slider drawn in impact rather than quantity, and the chart is absolute: notes genuinely fade instead of being renormalised to keep the picture full. A computed confidence prints beside the prediction, and it never prints 100, because a strip in your own air is still the referee. How the model works and how far to trust it.
• Iterate without losing the version that worked. The Scent Equalizer turns "more mossy, less smoky" into concrete gram moves from bottles you own and applies them as a new branch. Snapshots record their parent in the evolution tree, any node diffs against the version it was born from, and Compare puts two formulas side by side with shared lines aligned, which is precisely what a week of strip notes needs.
• Check the accord against reality. After a wearing you can record what actually happened, how long it really lasted and whether each phase came in weaker or stronger, and the app shows its own error on your own formulas. Nothing is adjusted automatically and nothing is uploaded.

When should an accord run to more than five lines?

Three to five is a home range, not a law. A brief that truly needs more gets more: a bridge between fast and slow lines, a naturalistic idea that only reads with several facets stated separately. Nature itself is not minimal, which is why the catalog carries hundreds of estimated constituent tables: a single essential oil is already an accord of many molecules. The rule is not "never six lines", it is "never a line without a job". If you can name the job, the line stays; if the job is "to be safe", it goes.

The way to earn the exceptions is to build inside the range until the range feels tight. Every deterministic tool named on this page opens with a free account, and the bench itself opens with no account at all; AI Studio runs are the one thing that is paid per run, and the paid version is a one-time purchase, not a subscription. Pick an idea you can name in three words and state it in four lines: Open the bench.