Early this winter Google Flu Trends predicted that up to 11% of the US population would in have influenza by mid-January – the peak of the flu season.
The actual number was roughly 6% contracted influence during that peak period in January.
Why did Google Flu Trends get it so wrong?
Because Google algorithms looked at numbers only. Some researchers suggest that wide-spread media coverage inflated the flu searches by people who were actually not ill.
Looking only at numbers may produce patchy results. For a proper analysis you need context.