Editor's note: Steve Gittelman is president, and Elaine Trimarchi is executive vice president, of Sample Solutions, an East Islip, N.Y., research firm.
Researchers must know if the changes they see in their data are real or artifacts of an inconsistent sample frame. The quality of respondents as measured by their levels of engagement is hypothesized to be correlated with the consistency of data obtained from commercial online panels around the world. To that end, identical tracking studies were conducted among 21 panel companies in 13 countries, represented by 10 companies and a total of 22 panels. A correlation was found to exist between respondent disengagement and the inability of a panel to generate replicable, consistent data using the same survey vehicle.
Consistency of online samples has become a core issue for market researchers. After all, much of the value we provide is in the tracking studies we perform, but even one-time studies should relate to some reference and not float in a sea of variability between panels. If your data changes, it is essential to know if the changes are real or the inadvertent product of sample inconsistency.
In the past, we had no reason to fret over sample consistency. At the core of every research career there is a fundamental reliance on probability. Toss a coin, any coin. It will reliably come up half heads and half tails. There is no magic in it; in the coin toss exercise we are matching to known characteristics.
No one would expect to toss a coin a million times in order to prove its “fairness.” Market researchers have drawn samples from known commodities for decades always relying on the fairness of a coin toss. Households could reliably be reached by telephone almost 99 percent of the time and the small fraction of non-phone homes mattered little. Yes, we had to adhere to strict calling regimens, callbacks, refusal conversions and most of all recove...