witching sensory test systems has allowed General Mills, Inc. to improve testing accuracy and save on costs in product reformulation, said John Cowden, senior sensory scientist for the Minneapolis-based company. He spoke Feb. 27 at the Institute of Food Technologists’ Wellness 13 in Rosemont.
For years General Mills used triangle tests to determine whether reformulated products, such as those with reduced fat or sugar, had similar enough sensory qualities to the previous products. The triangle test involves people sampling three products, such as three cookies, and giving their opinion on which one is different from the other two, such as in texture or flavor.
Now, General Mills has switched to a tetrad sensory test that involves four samples. After tasting all the samples, people pair them up, such as saying cookie No. 1 and cookie No. 3 taste the same while cookie No. 2 and cookie No. 4 taste the same. The change came after General Mills tried both triangle tests and tetrad tests on nine products: three cereals, two baked foods, two dairy items and two spicy meals.
Switching to the tetrad test from the triangle test allowed General Mills to better avoid two types of sampling errors, Mr. Cowden said.
Type I sensory testing errors happen when companies wrongly conclude that a difference exists in a product. For example, a company through sensory testing may conclude a reduced sugar ice cream is less sweet than the conventional ice cream when the actual sweetness is similar. Type 1 errors may send product developers back to reformulating when no reformulation is necessary, Mr. Cowden.
Type II sensory testing errors happen when companies wrongly conclude no difference exists in products. For example, sensory testing may show a reduced sugar ice cream has similar sweetness to the conventional ice cream when there is actually a big difference in sweetness. The type II error thus may lead to a failed product once it hits the market.
Besides reducing type I and type II errors, the tetrad system requires fewer sensory panelists, fewer samples to make and less testing time when compared to the triangle system, Mr. Cowden said. The tetrad system also is less complex for the laboratory to execute even though the tetrad system has four samples and the triangle system has three samples.
Benoit Rousseau, Ph.D., senior vice-president of The Institute of Perception, Richmond, Va., also spoke at the Feb. 27 session on sensory tests in reformulation. He said companies should consider five factors in sensory testing. The first two are type I and type II errors.
For the third factor, companies need to accurately find how big a difference exists between the reformulated product and the previous product. No two products will ever be exactly equal, but the size of difference — say the sweetness in ice cream — may not be enough to affect consumer acceptability.
Sample size is the fourth factor. A greater sample size will provide greater statistical power and reduce the chance of errors. In other words, a sample size of 100 people is better than a sample size of 10 people. The testing protocol, such as triangle or tetrad, is the fifth factor.
