Tips: attribute importance

You can ‘derive’ the relative importance of different factors using correlation or regression.  It is standard practice in research, but using regression can cause problems.

 

Derived importance identifies which attributes are most important by measuring the strength of the relationship between ratings for each attribute to the ratings for an ‘overall’ attribute (e.g. overall satisfaction).  The advantages are that:

  • it avoids the interview time involved in asking people to rate the importance of each attribute

  • what people say is important does not always reflect their actions.  For example, a person may say that price is very important but he/she gives a low price rating for the brand they most often purchase – indicating that price isn’t really a driver of purchase to them.

Derived importance is done using regression or correlation analysis.  The problem with regression is that it ignores a record with missing data on any of the attributes. So if a respondent answers “don’t know/ no opinion” to even one of the attributes, they will be not be included in the regression calculations. Or, if an attribute was only rated by a sub-group of respondents, then anyone not answering that attribute will be ignored. Correlation does not have this issue and is consequently better to use.