Google released a research paper yesterday that takes an in-depth look at Google Trends and Google Insights for Search and compares its trend results against more mainstream industry reports based on actual sales data in several different categories like auto sales and home sales. Several statistical models are explained and compared, and relative accuracy is determined for each model.
As interesting as that sounds, the crucial difference is that, for Google, the statistics have already been generated — sometimes for weeks before actual retail numbers come out. Google’s trends are much closer to the present day economic situation, meaning that analysis can be made earlier, and market turning points can be identified sooner.
All this forecasting power isn’t available just to Google, either. On both the Google Insights for Search page and the Google Trends page, any query result can be downloaded as a CSV (comma separated value) format file, suitable for import into any spreadsheet or database application. If you are in the trend business, this is great news as you can immediately start using Google result datasets against other, more traditional reports.
An important point that is made at the beginning of the report is that Google Trends pulls from all Google properties using localized, anonymous search usage data. As Google Blogoscoped points out,
Whenever you search in Google, send a Gmail email, call Google’s 411 service, enter something in Chrome’s URL bar and so on, you’re adding to the data which Google can mine for other purposes than the ones you might have been immediately thinking of.
The fact that Google is sitting on an information treasure trove is one thing. Knowing how to take advantage of that fact is another. Google goes a step further and offers their aggregate statistics for others to use, too. For example, they have a special page that just tracks flu trends across the US to help people make informed decisions on if they should get a flu shot, based on what region of the country they live in.