The users of online sample should be able to source from as many sample providers as they choose to achieve preferred pricing and diversification. In these times of changing sample sources, we provide blending models. As a third party independent auditor, we monitor those sources to achieve optimum consistency: that is, no matter what sources you choose to input, we must make best efforts to achieve a consistent output.
What we provide:
| ||1.|| As online samples bring in new sources of their own, they change, thus we provide a service that tracks consistency over time.|
| ||2.|| Regardless of where sample is needed, we measure or blend the sample to assure consistency, thus providing a global solution.|
| ||3.|| It is necessary to have standards, both behavioral and demographic, built to provide references for comparison around the world, thus providing stable references globally.|
| ||4.|| In addition to digital fingerprinting, engagement measures and outside verification, we profile respondents by ten segmentations, measure survey participation and accuracy of responses, thus providing pre-profiled sample on a respondent level.|
| ||5.|| The Grand Mean Project® represents a first step in transparency covering over 200 waves of 500 interviews, representing panel companies in 35 countries.|
Our business model is straightforward. We work with the users of sample to create customized metrics and standards for their use. The sample sources work with us to create these metrics and standards. By using standard methods and metrics, the sample sources are able to use them with multiple end users to amortize the costs that are incurred. For both sides, it is a win-win.
Although diversification has its virtues, we prefer blending solutions that are targeted to represent a standard. At times, the standard is an historic one that represents the characteristics of the sample source(s) as it existed at some point in time. Often a better alternative is to re-configure the output of the sample stream so that it is representative of an external standard or census. Unfortunately, the census is demographically and not behaviorally anchored, thus we prefer to use standards that are multi-mode and representative of behavior and demography, we call this standard The Grand Mean Standard™. We use optimum blends of the segments from a battery of segmentations to achieve a demographic and behavioral blend.
The optimum blend is determined by varying the weight of each of the elements in a sample stream. We use a Buying Behavior segmentation scheme, characterizing a consumer on purchases within the sample sources. The segments themselves comprise the words in a common language that allows for interchangeability. Consumers are fitted to the demographic cell to which they belong in proportion to the distribution of the behavioral segments found within that demographic cell.
In the example above, by optimizing each of the four sample sources (SS#1-#4), we create a blend that is close to The Grand Mean. For example, the first column represents "Broad Range/Credit". When we blend the four sources together using optimization, we approach the Grand Mean (35% vs 36%). This is true for the other three segments as well: Price sensitive/shoppers = 15% vs 14% Credit/Environment = 22% vs 23% and Domestic/Coupons = 29% vs 27%. The final result is the RMS equals 3%.
Periodic audits are key to the use of optimization. It is also necessary to have a reliable reference, the Grand Mean. To create a stable Grand Mean, we use a blend of reference data with highly granular nested quotas. We use this as our standard in each country.
CASRO Online Conference 2011: Optimum Blending of Panels and Social Network Respondents