Recent studies had shown that the volume ofInternet searchesit can “predict the present” – that is, it can be used to track and quantify unemployment levels, home sales, or the incidence of diseases like influenza A in real time. Now, an investigation byYahoo! Researchand published in the magazinePNAS reveals that the Internet also allowspredict collective consumer behavior within days or even weeks in advance.
Specifically, the researchers used the volume of Internet searches to try to predict the collection of certainmovies at the box office, the sales figures ofvideo game premieresduring the first month, and the charts. The actual results largely matched the predictions in the first two cases, though not for music.
However, comparing the predictions with those based on traditional information (critics’ ratings, income from audiovisual advances, etc.), the researchers found that the degree of accuracy of these “classic” methods was higher than the predictions offered by the searchonline. The authors suggest that while this demonstrates the search data could onlymoderately improve predictions on large-scale trends, sudden changes in search volume would allow us to identify “turning points” that cannot be detected in other ways. And you could even make a profit in the case of applications such asfinancial analysis, where minimal variation can be valuable.
