Google Trends is Behind the Curve
Google Trends is sometime heralded as a tool that can help to predict the outcome of future events. I have said before that I think this is unlikely to work, and from what I found today, I believe this even more so.
Today, I stumbled over an article that Anne Zelenka wrote some 9 months ago at GigaOm. In this article, Anne refers to a CNET article that described how Marissa Mayer stated that based to Google Trends, Hillary Clinton was likely to win the Democratic Party’s nomination. We all know how accurate that turned out to be…
Here is the Google Trends graph from that article in GigaOm:
So, today, I went back into Google Trends to extract a more granular graph from January 2008:
As you can see, until January 3rd, the date of the Iowa Caucus, Clinton and Obama had the approximate same number of Google searches. Once Obama had won in Iowa, his searches skyrocketed and he left Clinton behind for some five days. However, on the day that Clinton won the New Hampshire Primary, Janury 8th, they were neck to neck again, until the 26th Janury, when Obama won South Carolina, at which point he started to build up a lead over Clinton again.
In all three cases, there would have been no way to use Google Trends as an accurate tool to try and gauge who will win the next primary/caucus. However, in retrospect, you can actually see who the winner was from looking at Google Trends.
What this means is the following: This means that Google Trends is always behind the curve, not ahead of it. The reason for this is that people can only search for terms, when they already know of them. They can only know them, when they have heard of them from somewhere else.
A much more meaningful tool to predict what will happen is fundamental data that you take from websites themselves. Have a look at the visitor stats of Obama’s and Clinton’s websites:
You can easily see how Obama was separating from Clinton in December 2007, he seems to have had roughly twice her web traffic and he maintained that 2x lead for several months.
I suggest that if you want to see how a candidate is doing in the online world, have a look at basic data, such as website traffic. Using basic data from websites enables you much better to understand who is going up and who is going down than Google Trends.