How to improve a perfume formula

A formula improves by diagnosis, not by adding things. Smell it at three fixed times, name the one thing that is wrong at the time it goes wrong, change one line, and keep the version you changed from.

Most formulas that stall do so because the perfumer is reacting to the whole rather than to a moment. "It is not quite right" is not a diagnosis. "The heart is empty at two hours" is, and it tells you where to look.

How do I diagnose a perfume formula?

Put the trial on a blotter and on skin, write the time on both, and read them at three points: ten minutes, two hours and eight hours. Those three readings are three different compositions made by different subsets of the same lines, and a fault almost always lives in one of them.

• Ten minutes. Is the opening the idea, or just the loudest material? Is anything harsh, sharp or chemical that should not be? An opening fault is usually a dose problem on a top note or a strong aldehyde.
• Two hours. The tops have gone. Is there still a shape, or a gap where the heart should be carrying the idea? This is where most amateur formulas fail, because the heart is dosed as if it were still competing with the tops.
• Eight hours. What is left, and does it still say the thing you named? A base that is only "warm" is a base with no idea in it.

Write one sentence per reading. If you cannot write it, smell again the next morning; a two-day-old strip is often clearer than a fresh one because the nose has forgotten what it expected.

Why change only one line per trial?

Because two changes at once give you a result you cannot attribute. If you raise the vetiver and lower the bergamot and the trial is better, you do not know which move did it, and the next trial is a guess again. One change, one trial, one sentence about what it did. It feels slow and it is the fastest method there is, because every trial teaches something true.

The change should match the diagnosis. An empty heart is fixed in the heart, not by adding more base. A harsh opening is fixed by lowering the harsh line, not by adding something sweet on top of it.

What order should the fixes come in?

1. Cut before you add. Remove any line whose job you cannot say in a few words. Most formulas get better with fewer lines before they get better with different ones.
2. Fix the structure. A hole at two hours, a base that arrives too late, an opening that promises what the drydown will not keep. These are about which materials and how much, and they come before shading.
3. Then shade. Greener, saltier, drier. Small doses, and only once the shape holds.
4. Then the finishing questions. Concentration, maceration time, whether it needs to last longer or throw further, each of which has its own page.

A worked example

A citrus-woody sketch: Bergamot Essential Oil 30, Linalool 10, Hedione 25, Iso E Super 30, Ambroxan at 10% dilution 5, in parts.

At ten minutes it is a good bergamot. At two hours it is faint and vague: the bergamot has left, and Hedione and Iso E Super are both quiet materials, so the heart has volume but no voice. At eight hours it is a pleasant ambery wood. The diagnosis is written in one line: the heart is empty at two hours.

The fix goes into the heart. One change: raise Hedione from 25 to 40 parts. Trial, three readings, one sentence. If the heart now has lift but still no character, the next trial is not more Hedione, it is a heart material with an actual opinion, a small dose of Geraniol perhaps, and again only that one change. Three or four trials in, the two-hour reading is a composition rather than a residue, and each version explains the next.

Why keep every version?

Because the version that was better is only useful if it still exists. Edit a formula in place and trial four, which was the best one, is gone, and the memory that it was best is attached to nothing. Number the versions, keep the sentence that says what changed and why, and keep the ones that were worse. A formula is a family tree, and the branch that was abandoned at trial six is often the one worth returning to at trial twelve.

What does a drydown chart add to the strip?

A strip tells you what happened. A drydown chart tells you what is likely to happen before you weigh anything, and it points at the hour to look at. PerfuMate draws the formula in impact, how loud each material reads rather than how much of it is in the bottle, along a slider from hour zero to hour twenty-four; each material decays against its own strip life, and the chart shows which families still carry the formula at each stage. Drag to the heart stage and a hole shows as the picture thinning, before a blotter has to tell you. It is a model, and it prints its own confidence beside the prediction rather than pretending to be a nose.

PerfuMate does exactly this loop with the materials you own: the scent profile and drydown chart show the formula at every hour, the evolution tree keeps every version with what changed between them, the substitutes engine finds a replacement on your own shelf, the IFRA check runs live as you edit, and the AI is handed your catalog and shelf for each run, never trained on them. See the bench.