The Grand Mean Standard --- Achieving Representative Samples using the Real ID® Method.
Sourcing is one of the greatest influences driving behavior variability between online samples. The chart below examines the differences in buyer behavior within four different sample sources and represents the composition of the US Grand Mean.
The primary difficulty in accounting for differences is that the differences transcend simple demographics. Even when weighted to account for rarely considered characteristics such as marital status and education, the online only and general populations differ in their consumer behavior.
Our solution is to quota behaviors in addition to demographic quotas!
Though this is not so easy to execute as it is to imagine. One needs, first, some idea of what a respondent's behaviors are, and second, he or she needs a reference point in order to quota control these behaviors. Between the Grand Mean Standard and Real ID, we provide both.
By interviewing respondents using a standard seventeen minute questionnaire through phone, as well as multiple online methods, such as panel, social networks, and river, we have done our best to reach all respondents who can be profiled. By comparing these profiles against consumer benchmarks from a number of trusted sources, we optimize these profiles to produce a blend that is our best approximation of consumer behavior in the general population.
In order to combine the many variables collected through this method into an easily summarized form, we rely on our three structural and seven market segmentations. These are, respectively, general and market specific typologies of consumer behavior formed through cluster analysis.
These typologies in turn serve as behavioral quotas by which we can sample our database of online respondents to deploy a representative sample for any study in the field. In addition to consumer behavior, we also have survey taking behavior, which allow for us to filter invitations by the quality of the individual's responses, as well as hyperactivity and panel tenure.
At the end of the day, the client receives a survey that is population balanced and high quality, with a great deal of extra information via each respondent's profile that will allow in-depth analysis of the data.
The Grand Mean Standard exists as an average of multiple sample sources within a market and serves as reference points by which sample providers can compare their panelist’s behavioral profiles in those markets. By comparing these profiles against consumer benchmarks from a number of trusted sources, we optimize these profiles to produce a blend that is our best approximation of consumer behavior in the general population. Currently, our best approximation of a representative sample contains a split between phone and online respondents (as shown below), with social networks and river sample excluded from the blend.