Sensory Evaluation and Product Development
This domain tests how sensory tests are selected, designed, and interpreted: panel types, discriminative and descriptive methods, hedonic testing, and the controls that keep responses unbiased. Expect questions asking which test answers a specific product-development question.
Core rules
- Trained descriptive panels profile attributes and intensities using methods such as QDA or Spectrum. Discriminative tests determine whether a detectable difference exists. Affective tests measure liking or preference with consumers.
- Triangle tests present three coded samples, two identical and one different; the chance of guessing is one in three. Duo-trio presents a reference plus two samples. Paired comparison presents two samples and asks for a specified attribute or preference. Tetrad tests use four samples in two pairs.
- Hedonic testing typically uses a nine-point scale from dislike extremely to like extremely. Just-about-right scales identify whether an attribute is too low, too high, or appropriate, but they are not liking scales.
- Controls: randomize serving order, use three-digit codes, serve at controlled temperature and light, use booths to prevent interaction, and provide water and crackers between samples. Panelists should not smoke or eat strongly flavored food before testing.
- Threshold testing determines detection or recognition thresholds. Difference tests require sufficient panelists and the correct binomial or tables for significance.
Worked example
A supplier changes a stabilizer and the team asks only whether the change is detectable. A triangle test is appropriate because it detects small differences with relatively few panelists. If the triangle result is significant, it does not say which sample is preferred or how they differ; a follow-up paired comparison or descriptive panel answers that. If the question were consumer liking, a hedonic test with representative consumers would be required instead.
Traps that cost marks
Using a triangle test to determine direction or preference. Using untrained consumers for descriptive profiling. Ignoring carryover and serving order. Treating a just-about-right result as a liking score.