External websites we've worked on: Dr Andy Moorhouse on piano ... Prof Trevor Cox
Jury test analysis
The complete analysis is attached in a spreadsheet here. To understand the description below, you will need to view this spreadsheet.
Below the analysis is outlined for the case of washing machines, although the techniques can be applied to most products.
In terms of statistical analysis, more can be made of the second part of the study, where people made judgements on scales:
Where people placed a mark on the scale, you measure the distance along the scale (say from the left most marker) and note this down in a spreadsheet. In the spreadsheet of washing machine results, for the worksheet "Subject 1" you will see the following in columns AZ-BE:
|Washing Machine||What it does||pleasant||Loud||robust||High quality||Purchase influence|
"The overall sound tells me what the product does"
A lot - No
"The product sounds pleasant"
Pleasant - unpleasant
"The product sounds loud"
Quiet - Loud
"The product sounds robust"
"The product sounds like a high quality product"
Expensive - Inexpensive
"This sound will influence my purchase for that product"
Positively - Negatively
These are the distances measured along the scales in mm.
In the washing machine test, 1.1, 1.2, 1.3, 1.4 all refer to the same washing machine and test method, but different parts of the washing cycle:
The purchase influence question (see right most column in Table 1), was only asked for the overall impression
Reducing subject bias
Some subjects tend to use different parts of scales, on a simplistic scale some people score meaner than others, others are naturally more generous! To overcome this fact, it is necessary to apply a normalisation to the judgements. This is done by making the judgements from a particular subject on a particular scale are made to have a mean of zero and a standard deviation of 1. For example, for subject 1, and the first column of data headed "What it does":
|Washing Machine||What it does||After normalisation|
If the original scores are x, and the mean of x is mx, and the standard deviation stdx, then the normalised values are: (x-mx)/stdx