Flu Trends was released in 2008 and was supposed to predict flu breakouts using people’s search data. The collected information was based on specific flu-related keywords.
The theory was that people with the flu would search more about its symptoms. That way, Google would have relevant data about flu trends well before official health organizations like the CDC (Centers for Disease Control and Prevention).
The plan was good on paper, but in reality, Flu Trends turned out to be a total failure. It completely missed massive outbreaks such as the 2009’s Swine Flu. Furthermore, its predictions for 2011 till 2013 were almost entirely wrong.
According to a study by David Lazer from Northeastern University, Boston, and his colleagues, Flu Trends managed to get right only 8 out of 108 weeks’ worth of flu spread-out during that period.
The main reason for the tool’s failure seems to be both the data it worked with and how it analyzed it. Even though Google Search provided a staggering amount of information, not all was relevant. For example, most people with flu-like symptoms didn’t actually have the flu, but some other illness. As a result, organizations such as CDC made better predictions since they worked with actual patients’ information.
Additionally, a service such as Flu Trends needed a constant update of what keywords people used and how they used them. This way, the tool could potentially collect the right kind of information. Google never revealed what keywords they tracked, but the results were speaking for themselves.