Editor’s note: Roy Eduardo Kokoyachuk is co-founder and principal of Burbank, Calif., firm ThinkNow Research.

Big data, social listening, machine learning and AI are all affecting the market research industry in new and exciting ways. The mind-bogglingly large data sets generated by the digitization of our lives are presenting market researchers with the richest data trove ever created by humanity. The sheer abundance of data has prompted many to question whether we’ve entered a “post-survey era” where it no longer makes sense to field quantitative surveys with hundreds or thousands of respondents when data is available on millions. We at ThinkNow Research like big data sets as much as anyone else and our company collects lots of data through opt-in passive mobile tracking, social media monitoring and organic search studies. Those tools are terrific for taking the temperature of a population or topic but don’t always arrive at concrete, usable answers. Simply making an observation and arriving at a conclusion as to why it happened without asking the observed population whether that conclusion is correct can be dangerous. We all have biases based on our personal experiences so assuming that we know why a population is or is not doing something online can be misleading.

Online transactional data is good but not every question is transactional

Observing behavior is good and sometimes ideal, especially if it’s transactional data. Amazon became the juggernaut that it is today by harnessing the power of past purchases and online behavior. The data needed to get people to do more of the same thing by uncovering the triggers that make us purchase items online can be found on Amazon’s servers. This type of transactional data is highly predictive for future purchases of the same or similar types of products. However, getting people to engage in new activities or buy previously unpurchased products is less pre...