I thought I would re-run the interview here, since he included my punch line without the set up. 🙂
Q: Are these new “ads” from Google really “research,” or are they a front for data gathering for advertisers anywhere Google does business?
A: Yes, they are really research. The hidden agenda — and it is not that hidden — is they don’t want publishers who make information freely available on their sites to go out of business: the rise of paywalls threatens the quality of content their search engine points people to. So with this initiative they are more interested in helping publishers (by providing an additional revenue source through their “survey wall”) than in helping advertisers.
Q: Has Google officially crossed over the line of knowing too much about us?
A: Whenever multiple people use a computer, Google becomes confused about who is who. Communispace has some great statistics about how inaccurate Google’s inferred demographics actually are.
If you check out Google Ads Settings, you will find out how accurate Google is for you. It has my age and gender correct but has a ludicrous list of my interests: it lists among my interests bicycles, dance and electronic music, dating, face and body care, rap & hip hop, and shooter games — I have absolutely no interest in any of those things. And this is on a computer that only I use.
Google Consumer Surveys provide a model of opinion on particular topics; like any other model, the results need to be treated tentatively. Does it have value for the research firms and corporate researchers that use it? Absolutely. Is it all-seeing and all-knowing? Ask my favorite hip hop star.
Ray Poynter of Vision Critical, who Terry also interviewed, believes the people who take these surveys are fundamentally different from the people who decline them (see the article for his quote). For the record I believe that Ray is generous to a fault in his support of the market research industry and its ongoing evolution, but he and I often disagree about sample quality issues (and he is kind enough to let me voice that disagreement on NewMR webinars and radiocasts). I disagree that nonresponse bias is currently the primary issue with Google Consumer Surveys – Google tested different microtasks as alternatives to paywalls and found surveys were the most popular. I believe all the microsurvey platforms – Google Consumer Surveys, RIWI, and Civic Science – are reaching wider audiences than traditional survey panels do, and I think that offers the possibility of improved representativeness.
I do worry that Google’s shift to multiquestion surveys (its surveys to-date have been limited to two questions) may reduce the quality of results it receives over time by increasing nonresponse bias. But my bigger concern about its representativeness is about its audience of readers. I would like to see Google provide a greater profile of its audience across psychographic, attitudinal, and behavioral measures: Google surveys an online audience of readers, and I am keenly interested in how readers may differ from the wider online audience (which includes people who primarily use the Internet to watch video or play games or use email or access social networking).
Back to Terry’s article: Terry threatens to lie when answering surveys, and that does no one any favors, in any type of research. Yes, sites may often have hidden agendas when it comes to how and why they provide online services (whether free or paid). While Google’s policy has been not to link survey answers to the advertising profiles it maintains of website visitors, Google certainly is free to change this policy at any time.
For me, Google Consumer Surveys remains a useful tool in the toolbox – fit for some types of research, but not for others.