JMP platforms and statistical methods

Graph Builder:
Heat map

Distribution:
Bar charts, frequency table, confidence intervals and hypothesis tests of proportions

Objective

Use visualization methods and statistical inference for proportions (e.g., hypothesis tests and confidence intervals) to assess if consumers can discriminate between two different ingredient formulations of a soft drink.

Problem statement

A manufacturer of soft drinks has been experimenting with a new ingredient formulation for one of its drinks that is popular with teenagers. Though this new formulation would be slightly more expensive to produce, it uses a healthier substitute for the current sweetener that the company believes could help increase sales by appealing to parents. If consumers of the drink are unable to tell the difference in taste, it could be more advantageous to use this new ingredient formulation.

The triangle test is a common method used in sensory studies to determine if consumers can discriminate between two formulations. In this type of test, assessors (taste testers) are given three samples to taste. Two of them are the same and one is different. Each assessor tries all three and is asked to identify which one is different. In this study, the assessors are not asked to rate which one they think tastes better or which characteristics they perceive to be different, but merely to identify which one they believe is different.

Each assessor gets up to one minute to taste the three samples, as often as they need and in any order. Each must decide which one they believe is different even if they aren’t sure they can tell the difference and simply must guess. There are two possible test sets one can receive: two of the samples are the current formulation and one new (Current, Current, New), or two are the new formulation and one the current (New, New, Current).

The study recruited 120 teenagers (60 males and 60 females) ranging in age from 13-17 (24 for each age), all of whom were consumers of the soft drink’s current formulation at least once a week. A balanced experimental design for age and gender and test set was created resulting in six assessors in each of the 5 x 2 x 2 = 20 groups.