Technical Report: Making Count-Based Claims from Sample Data

ABSTRACT

Numerical superiority claims such as “2 out of 3” or “4 out of 5 people prefer…” are powerful and commonly used to promote products. The National Advertising Division (NAD®) is the advertising industry’s self-regulatory body that adjudicates advertising claims in the USA. Examples of numerical superiority claims that were challenged at the NAD include a claim by General Mills that “In a national taste test, nearly 2 out of
3 Americans agree that Yoplait® Greek tastes better than Chobani®. Another example is the claim that “Four out of five consumers prefer LG Cinema 3D over ... Samsung Active 3D in head-to-head comparison. These claims involved performance numbers based simply on the actual results of the experiments. These cases raise a number of issues in using sample data to support advertising claims. One issue is whether it is justified to report sample data results in a claim where clearly the advertiser does not wish to limit the message to the particular consumers tested on a particular occasion. Even if it is truthful to say that particular results were obtain in a particular experiment, the implication to a consumer is that he/she will experience the benefit at a similar level. In reality the advertiser wishes to communicate a benefit message that applies to a target population at large, rather than reporting what amounts to an interesting factoid about an experiment that they conducted. A second issue is whether there should be consistency in the application of statistical principles to superiority tests that are not numerical and those that are. In this report we consider these two issues using a scenario for illustration of the points we will make.

This technical report appears as:

Ennis, D. M. and Rousseau, B. (2019). Making Count-Based Claims from Sample Data. IFPress, 22(1) 3-4.

Download this technical report here:

Making Count-Based Claims from Sample Data