Reliability of Google Trends
8/20/08 / David Kennedy
I recently wrote a post on Google’s new service, Google Insights, which is an evolution of Google Trends. As a result of that post, I ran across this post discussing if Google Trends is reliable.
One of the examples used compared the term “market research” to “advertising” and showed that both terms declined (as a percentage of total searches), but advertising even more so than research. Of course, as noted, this doesn’t necessarily mean advertising is losing popularity, especially considering the growth in online advertising. But, as Geoff wrote in another Radiance post, you have to be careful what you’re asking for. In this case, comparing online advertising (growing) to print advertising (declining) would have been more appropriate.
As the amount of information on the web grows, users have become more sophisticated in their searches. Therefore, the search terms we test must be more detailed and we must test more of them. After all, it wasn’t long ago that “Yahoo” was a top search on Google (though there are numerous reasons for that too).
We agree that Google Trends and Insights should be used knowing the limitations of what it can tell you (and the methodology behind the numbers), and as with any methodology, those limitations should always be considered when interpreting results. For one quick illustration of what we mean, look at the middle chart: the red line is “online advertising”, the blue line is “print advertising”. Compare that to the top chart where the lines are for “online ads” and “print ads”. These trends are for searches for the specific terms you enter, not the concepts you are interested in. They are indexed by the average traffic for whichever term you enter first.
The bottom chart is fun from a linguistic standpoint, and adds a whole layer of interest to the interpretation of the top two charts: the red line is “online advertising” and the blue line is “online ads” — in 2004 and 2005 people typed out “online advertising” in their searches more often than they typed the abbreviated “online ads”, but in later years those usages have started to even out. What effect does that change in usage have on your interpretation of the results in the first two graphs? What if I also tell you that you see a different pattern in “print ads” vs. “print advertising”? See the challenges?
The bottom line is, as always, with data analysis and interpretation: proceed with caution.
For another (map based) example of this applied to interest in emerging media & technology, check out this post on Techcrunch.