Code bias occurs when the number codes on our samples, in themselves, contribute to the response or sample selected by the respondent. The example quoted is that from a Chinese research study. In China, the character used for the number 4 looks similar to that used for ‘death’ and as such the number is often avoided for example, for floor numbers in buildings and hotels. So in the Chinese study, when one of the products was coded as 444, it received a very low hedonic rating. In Western culture the number 13 may evoke a similar response as would others such as 007 and 666.
Less obviously however, the article goes onto discuss the impact of the magnitude of the code on bias. In a small study IFP found that respondents would tend to choose samples labelled with higher numbers over those, same samples, when labelled with a lower number code.
The recommendation? Use more than one code for every sample in product testing, one high number and one low number. In addition, make sure that the difference in magnitude of the codes is similar.
So how do you select codes? The IFP has a list available to download but as a guide avoid the hundreds (200, 300, 400); symmetrical numbers; numbers beginning with zero; numbers ending in 5; duplicate digits; triplicate digits; and sequences e.g. 494; 012; 765; 266, 888 or 468.
Other common sources of bias in product testing arise from sampling, position and questionnaire design.