Although it’s 2012, most public opinion pollsters continue to report poll results as though it were 1962, and the only tool available to them was a Royal typewriter.
Consider the display-quality of these poll results, just released by one of the nation’s most prestigious national polling firms:
That is exactly how poll results looked like 40 or 50 years ago. Not a pica-pole’s worth of improvements in a half a century. Terrible.
By contrast, Google has developed a robust, elegant, beautiful back-end, which allows DIY researchers to visualize instantly what their results are saying.
Note: We have taken issue in the past with the very real danger that Google’s decimal-point precision (showing results to the tenth of a percentage point rather than rounded to a whole number) and Google’s extra precise margin of sampling error (showing a margin of sampling error for each and every answer choice, with differences, often, in the plus side vs the minus side of the “plus or minus”), will imply a precision that simply is not knowable given the methodology of these surveys. But that is a subject for another day.
Today we focus solely on what Google has done on the “back end,” and presuppose that the question wording and the sampling (“the front-end” of the research project) were all done perfectly. Examined from that perspective, what Google has done with Google Consumer Surveys is a model for what every opinion research firm should be doing with its survey results.
The following 3 examples illustrate what we are talking about:
- Example #1, left column, shows how the survey results are shown to you by Google, as a default. Both genders, all ages, all regions, and all incomes are selected by Google as a default. If you leave those “defaults” alone, the number of respondents shown to you will equal the number of respondents interviewed. So: if your DIY Google Consumer Survey interviewed 1,000 adults, there will be 1,000 adults in the default view.
- Example #2, middle column, shows the power of being able to click once and customize the report. In the example above, we have clicked on “men” (only; excluding women), and as such, the results that Google displays will automatically show the results of just men. This is just an illustration, and just as easily, we could have clicked on just 18 to 24 years olds, or just those with incomes above $150,000 a year. But the beauty of what Google has built, is that it is not just an “or” select. Meaning: you do not have to choose whether you want the results filtered by gender or age or income. What Google has built is an and/or select, which means you can select as many filters for the results as you would like. Which brings us to example #3.
- Example #3, right column, shows what happens when we custom select every available variable. In this case, we ask Google to show us just the results of those men who are age 18 to 24, and who live in an urban area in the state of California, and whose income is between $50,000 and $74,000. This is likely more granular than anyone would need to example research results, but we go through the exercise just to demonstrate that in fact, such customization and such granularity is possible.
Let’s see how this might work in the “real world” …
As we have previously explained, Google encourages those who use the DIY Google Consumer Surveys research tool to make the results of their surveys public in return for a 10% discount on the cost of the survey. For this reason, we are able to look at the results of actual surveys conducted using the Google Consumer Surveys tool.
Let’s take as an example, this question:
If you were shopping for a wedding photographer, what is most important to you?
When you leave the default filters in place (so you are looking at the answers from all respondents), you see results that look like this:
And you would conclude from looking at the graph that “Cost” narrowly edges “Attention to Detail” as the most important consideration when hiring a wedding photographer. But the results are actually more nuanced than that, and tell a wonderful story, related to whose paying for the wedding. If you examine the results from just young women, you see that “Attention to Detail” jumps to most important, “Artistry” jumps to 2nd most important, and “Cost” drops to 3rd place. That’s because by-and-large, young women do not pay for their own weddings; they are the brides.
But then, if you examine the results from just those who are midle-aged, middle-income, meaning: the folks who pay for most of the weddings in this country, the balance tips almost 2:1 in favor of cost, and poor “Artistry” falls to last place.
Two things are important to understand here:
- With most research companies, an analysis like the one we just completed, above, would not be possible at all. The results might be analyzable on a single dimension, but not on multiple dimensions at once.
- Even if the research company did make multi-dimensional analysis available, the time involved to conduct the analysis would take the lay person hours or days.
By contrast, everything that we show you above, using Google Consumer Surveys, took seconds. Not minutes; seconds. And you do not have to be a PhD statistician to run such an analysis with Google Consumer Surveys. Just a person willing to experiment, and willing to pay reasonable attention to the inherent limitations of increasingly small sample sizes.
Hat’s off to Google for developing this most-powerful back-end reporting engine. Every research firm should offer the same kind of horsepower to their lay clients.
To see everything we have written about Google Consumer Surveys, type “google” into the search box at the top right corner of this page, or click here.