Presented at the 2013 Pangborn Symposium in Rio de Janeiro, Brazil.
The decision to accept or reject a product reformulation following a sensory discrimination test is often based solely on a statistical test against a null-hypothesis of “no difference”. But a non-significant difference with a given number of assessors can become significant if the sample size is increased. This fact illustrates the need for a missing piece of information: the size of the difference that the scientist does not want to miss. An attempt to provide this information has previously been made using the proportion of detectors. Unfortunately, the proportion of detectors depends on the protocol used and ultimately cannot be used to represent meaningful sensory differences.
In this presentation we discuss several alternate methods for establishing consumer-relevant action standards above which sensory differences will be considered to be ‘too large’ to represent an acceptable business risk. These methods are all consumer-driven, and include: linking consumer preferences to underlying sensory differences, using the same-different or degree of difference methodologies to determine the criterion for two products to be considered similar ‘enough’, connecting trained panellist and consumer sensory acuities, and investigating the variation already present in the market due to natural or production runs variability.
Once the criterion has been established, suitable power and sample size calculations can be conducted, which in turn provides increase stability and confidence in decisions reached within a sensory discrimination program.
Colleagues can download the poster here:
Not a Colleague? Click here to join for free!