Depending on how you count it, we've either completed the first decade of the 21st century, or we are in its final year. Hard to believe.

 

A lot has happened in the last ten years, not the least of which (for me, anyway) has been the establishment of ADM Marketing & Research Consulting, just 8 years ago. We've done some interesting work, met some terrific people, and plan to do more of the same as we enter the "2010s."

 

Happy new year, and I hope it brings you success (however you define that) and satisfaction.

 

Just because you CAN do something, SHOULD you do it? This is not a new question, and it does not require a "Yes/No" answer. In marketing research, as in many other disciplines, it's really a matter of HOW new techniques and technologies are used, not WHETHER you use them.

 

A little history.  The modern field of marketing research started in the 1930s with George Gallup, who collected data in the only feasible way at that time: door-to-door. This allowed him to construct scientific random probability samples, which yielded projectable results. When telephones became ubiquitous, the industry developed random digit dialing to produce equally reliable samples at a fraction of the cost. Now we have online surveys,which some believe continues the path toward ever decreasing research costs while maintaining quality.

 

The good news is that sometimes it works well. When you want to survey a niche population (employees or customers; subscribers to a certain publication; online buyers or students), there is less variation than in the general population, so you can be pretty confident that a sample of a few hundred (or thousand) is reliable. The bad news is knowing whether the results are truly representative. Difficult to know for sure, but probably not. The point is: the more homogeneous the population, the less you need a strictly designed sample.

 

But for general population surveys, online is more "mushy." Think: snail mail surveys done electronically. You have the same problem, especially because of response rates. If you send out several thousand (or tens of thousands of) questionnaires and get 10% (or fewer) back, it is difficult to know WHICH 10%. Are they spread out representatively throughout the outgoing sample? You can run demographics and compare them with census data to see how well (or poorly) your sample matches up. If it's close, you are lucky. More likely, you will need to weight the data to make it LOOK representative. Either way, I always have a queasy feeling about it.

 

Bottom line: Online surveys are here to stay, especially because do-it-yourself products make it so easy for non-researchers to conduct surveys. Let's be honest about the meaning and use of these data.

 

Amen