Search data "no more informative" than standard analysis

 

Gallery

By Reuters

Posted on 28 Sep 2010 at 10:37

Tracking web searches for new songs, video games and movies can predict which ones will be big hits, but often not much better than traditional methods, researchers at Yahoo have reported.

And they confirmed earlier findings that showed searches associated with diseases, such as Google's Google Flu Trends, were not any more effective than traditional methods for predicting the spread of infections.

Tracking web searches worked the best in predicting how a new video game would sell, Yahoo's Sharad Goel and Jake Hofman said.

"Here we show that what consumers are searching for online can also predict their collective future behavior days or even weeks in advance," they wrote in a report published in the Proceedings of the National Academy of Sciences.

"Specifically we use search query volume to forecast the opening weekend box-office revenue for feature films, first-month sales of video games, and the rank of songs on the Billboard Hot 100 chart, finding in all cases that search counts are highly predictive of future outcomes," they added.

Search data are, at least in some cases, no more informative than traditional data sources

"We considered four different classes of web activity. At least three of them we do a pretty good job of predicting outcomes," Goel said.

Counting the number of searches related to new songs was the least effective, they found. Goel said using Billboard Top 100 listings to predict whether a song would stay at the top of the charts worked as well as or better than counting web searches.

Critics' ratings for movies also worked just about as well for predicting box office performance as web searches, he said.

"Given the attention that search-based predictions have received recently, it may seem surprising that search data are, at least in some cases, no more informative than traditional data sources," they wrote.

To check their findings, they took another look at flu data. Google offers a free product for trying to predict the movement of influenza by tracking web searches for words such as "flu", "fever" and "Tamiflu".

The Yahoo team found the search engine method only worked a little better than simply tracking hospital and health department reports of flu activity.

Hofman and Goel said they could not comment on whether Yahoo was developing a search engine product for predicting sales.