The Triangle Test: Can Anyone Smell the Difference Between Two Batches?

When to use it

• Batch consistency: is the new batch distinguishable from the reference? See repeatable perfume batches.
• Material changes: does a new lot or supplier change the perfume? See raw material supplier changes.
• Cost or restriction reformulation: can people tell the reformulated version from the original? See reducing a fragrance formula's cost.
• Process changes: a different maceration time, filter or alcohol grade.

It answers only "is there a perceptible difference?". It does not say which sample is better, how big the difference is, or what it is. For preferences, see consumer fragrance testing.

How to run one

Prepare the samples

1. Make both samples the same way in everything except the variable: same concentration, same maturation time, same bottle type, same temperature.
2. Present them on identical blotters dipped to the same depth at the same moment, or in identical coded vials. Blotters are easier to keep consistent.
3. Give each sample a random three-digit code. The person preparing the samples should not be the one asking the questions, if you can arrange it.

Balance the order

There are six possible arrangements of two samples, A and B, in a triangle: AAB, ABA, BAA, ABB, BAB, BBA. Use each equally often across assessors so that neither position nor which sample is the odd one biases the result.

Run it

• Each assessor smells the three samples in the order given, left to right, and must choose one as different — even if they have to guess. This is a forced choice; "no difference" is not an allowed answer.
• Smell at a fixed time after dipping; for perfumes, you may run separate triangles for the opening and for the drydown.
• Allow a short break between triangles to limit fatigue.

Reading the result

Count correct identifications and compare with the minimum needed. These figures come from the binomial distribution with a one-in-three chance of guessing correctly, at a 5% significance level:

Meeting the number means the difference is unlikely to be chance. Falling short does not prove the samples are the same; with few assessments, real but small differences are often missed. Testing specifically for similarity — "can we be confident nobody will notice?" — needs more assessments and is covered in the standard itself.

Small labs and repeated tests

Few small labs have 24 assessors. Options: repeat the test with the same few people on different days (each repetition counts as an assessment, though repeated judgments by the same person are not fully independent); recruit colleagues, friends or customers for an hour; and use the test for the decisions that matter most. Even six well-run assessments are better than one person's impression.

Recording a triangle test

• What was compared — formula versions, batch numbers or material lots — and why.
• How samples were prepared and presented, and the time point.
• The codes and orders used, and each assessor's answer.
• The count, the threshold, the conclusion, and the decision taken.

Keep the record with the batch or version it relates to. The broader quality routine of a lab is described in how to set up a perfume lab.

Sources

• International Organization for Standardization — ISO 4120:2021, Sensory analysis — Methodology — Triangle test, checked October 9, 2026 (catalogue entry; the standard itself is sold by ISO).

Frequently asked questions

Can I do a triangle test alone?

Only if someone else codes and arranges the samples so you do not know the layout. Repeating it on several days gives you some evidence, but your own sensitivity is only one data point.

Why not just ask "are these the same?"

Because people say "the same" or "different" for many reasons, and you cannot tell guessing from detection. The forced choice with a known chance level is what makes the result interpretable.