How will you combine data sources?

Mixing of panels has been suggested as an obvious means to stabilize data sources. The result of an average would always be better than the disparity of the most unusual panel. On the one hand, there is no assurance that simple combinations of panels will produce the desired result. On the other hand, optimum blends of panels are available through our research.

The optimum blend is determined by varying the weight of each of the panels in a group. While any number of panels can be used for the group, usually it is desirable, for purposes of practical execution, to keep the number small. We use Buyer Behavior segment distributions, which characterize the population of the data sources in respect to issues of consumer purchasing. In this case, three panel sets are often used. Note, however, that the eventual optimum may have less than three if any of the weights are close to zero.

The optimum is the point where the distribution of segments most closely approaches the Grand Mean. The size of the segments of the resulting blended panel is the weighted average of the component panel values. The optimum can be demonstrated graphically as shown on this chart. The bottom-most point on the surface represents the optimum value. The dark orange zone area represents the lowest level, the center of which would be the optimum. The third panel is determinate.

Over 25% of the optima we calculate globally, indicate less than a 1% deviation from the local Grand Mean. The following figure shows the optimum solutions we obtained from each set of three panels or sources from our database of 17 US panels. The Grand Mean was estimated from these data sources. This resulted in an ensemble of 680 possible optimum sets.

We can optimize to the Grand Mean. The example below shows the expected standard error from the Grand Mean based on the average of all random choices (8.31%). Based on equal weighting of three panels selected by optimization to the Grand Mean (2.36%) and the same three panels blended in proportions to optimize to the Grand Mean (0.40%). Optimization gives us precision.



Optimum panel mixes are not all created equal. Not all blending of panels can generate improved optima. In some cases, two or even one panel made up the optimum solution.

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, a sufficient number of data sources have to be audited. Optimum Blending Solutions® is a new service available to the industry, researchers and panel companies alike; there is no need to put all your eggs in one basket.

We have developed optimum solutions in countries all around the world. Panels are welcome to join The Grand Mean Project and be part of blending solutions. Speak to us today about insuring the future of your next project.