In short
A fragrance experiment turns a creative hunch into a test: one stated change, a control made at the same time, coded samples and a rule for deciding, written before you look. It need not be elaborate. Most learners already make versions, and what this adds is the ability to say that a change caused the difference, and to catch yourself when you merely liked your idea. The Research and Innovation track builds on this in far more depth.
What you will learn
- Write a hypothesis you can test
- Design a controlled comparison with one variable
- Complete an experimental composition with a written rationale
Before this lesson: AI in Perfume Research and Formulation: Uses and Limits
From hunch to hypothesis
| Hunch | Testable version |
|---|---|
| This is too heavy. | Removing 5 parts of benzoin will reduce "heavy" ratings at 1 hour. |
| A musk would help. | Adding 3 parts of musk X will increase how long two assessors detect it on strips. |
| Metal might be interesting here. | A trace of a metallic note will change people's description, without being rated unpleasant by most. |
Parts of a sound design
- One variable. Change one thing, or one planned set. Everything else stays the same.
- A control made alongside. Make both at the same time from the same stocks. A version from last month has aged differently.
- Split batches. Make one batch and divide it, then dose only one portion. This removes weighing differences.
- Coding. Have someone code the samples, or use a shuffle-and-cover method so you can't tell. See blind smelling and the triangle test.
- A stated medium. Strips, fabric or skin, each with the appropriate safety gate.
- Timed checks. Fixed times, chosen in advance.
- Replication. Repeat on a second day, or with a second batch, before you trust it.
- A decision rule. For example: "Keep the change if I identify it correctly in two sessions, and if at least two of three assessors prefer it."
A worked plan
This example is illustrative. Its results would be yours to find.
Question: does a trace of a metallic-mineral material improve a clean musk draft? Control: 50 g batch split into two 25 g portions. Change: add the material to one, as a stock, and add the same weight of solvent to the other, so the dilution stays equal. Coded: a friend labels them X and Y. Checks: strips at 15 minutes, 2 hours, 8 hours and 24 hours, on two days. Assessors: you plus two others, describing first and then stating a preference. Rule: keep it if you can tell them apart in both sessions, and at least two out of three prefer X or Y at 2 hours and 8 hours.
Notice the small but important detail of adding the same weight of solvent to the control. Without it, you have changed two things.
What a small test can and cannot show
A small test can tell you that a change is detectable, for you and your few assessors. It cannot show what the market would prefer, what is true for people in general, or whether something is safe. Report it accordingly: "Three people, strips, one batch." If that is all you have, say so.
Level 9 project: an experimental composition
Bring together the level's themes in one composition with a written rationale.
- Choose a question that interests you, using a technique or material from this level.
- Write the rationale first: the idea, the hypothesis and the decision rule.
- Build a working draft of up to 25 materials. Check restrictions and record your sources with dates.
- Run at least two controlled experiments, each changing one thing.
- Have at least two others respond before explaining the concept.
- Write up the rationale, the record, the results and the final decision in no more than two pages.
Assess yourself. Your project is complete if your hypothesis was written before you started, each experiment had a real control, the record would let someone else remake the formula, you report what did not work, and you make no claim beyond your data. This formula is a study composition. It has not been assessed for skin use, and passing this project does not change that.
Practical exercise: Level 9 project: an experimental composition
Carry out the Level 9 project above. Set aside several sessions over at least a week, so that your 24-hour checks and second-day repetition can take place.
You have done this well if a reader who has never met you could tell what you tested, repeat it, and see why you decided as you did.
Common mistakes
- Changing several things in one version. You cannot attribute the effect.
- Moving the goal after you see the result. Write the rule in advance.
- Comparing against an old version. Age itself could be the difference.
Key terms
- Control
- A version made the same way but without the change being tested, which gives you something to compare with.
- Hypothesis
- A specific statement that a test could show to be wrong.
Knowledge check
Written and reviewed by the RUŌOD Lab team. This is general education, not legal, regulatory, medical or safety advice. Example dilutions and exercises are for learning, are not validated products, and have not been safety-assessed. Completing a lesson confers no qualification. How we check these lessons.