Sensory Difference Test: Triangle and Paired Comparison Online
What Is a Sensory Difference Test?
Controlled tasting or smelling experiments judge whether two products have a perceptible difference, for scenarios such as formula changes, raw material substitution and shelf life comparison. Common methods are the triangle test (three samples, two identical and one different; pick the odd one out, with a guessing probability of 1/3), the duo-trio test (a reference is given first, then two samples; pick the one matching the reference, guessing probability 1/2) and paired comparison (choose one of two samples, such as sweeter or better, guessing 1/2). The three methods differ in statistical efficiency: the triangle test is more sensitive to differences but needs more panelists.
How to Design the Experiment
Code samples randomly with three-digit codes and balance order and position effects by swapping left and right or first and second. The number of panelists is determined from tables by the expected detectable difference and significance level (at alpha = 0.05, detecting a general difference with the triangle test needs about 30 or more panelists). Panelists evaluate independently without discussion. The tool includes critical value tables for each method and can work backwards to the required number of panelists; brief the panelists on the procedure first and record the whole process for traceability.
How to Judge the Results
Count the number of panelists who correctly identify or select the target and compare with the critical value: reaching or exceeding it means a significant difference exists (p less than alpha). The triangle test decision table gives the critical correct count for the total number n; enter n and the correct count and the tool gives the conclusion directly, stating the significance level, and also shows critical counts at two significance levels (such as 0.05 and 0.01) so you can choose flexibly.
Precautions
A difference test only answers whether a difference exists, not preference (preference needs a paired preference test). Control sample temperature, presentation order and evaluation environment strictly; a significant result means a perceptible difference exists, and whether it affects acceptance requires consumer testing. Record samples, codes, environment and panelist count in the report so the experiment is reproducible, and keep original evaluation sheets when results drive product change decisions.