Google Trends Update

Google labs recently upgraded their trends feature - a system which displays the volume of use of a query over time. The new feature mines this data set for query terms which are showing some interesting upward changes. At first, when I looked through the top few trends, they all seemed to look the same - like this (note to Google - put the title of the graph on the graph):

In other words, a flat - almost zero - level of activity followed by a sharp up swing. However, today, it appears that there is a better variety of examples:

There still seems to be a bias towards jumps from very low volume of some sort to - well, we don't know what the vertical scale of these graphs is, so we have no idea as to the significance of the trends. Providing a ranking function that attempts to optimize multiple variables is hard, and it appears that this algorithm discounts trends for terms which already have a reasonable volume. For example, if there was a sudden run on the term 'iraq', it may not appear as (I assume) the existing search volume for iraq is probably quite high.

Google trends - with the added hotness - are fun. One improvement they could make would be to include sparklines for each term so that one wouldn't have to click through to see the shape of the trend. 

I'm always on the look out for new and interesting consumer deployments of time series in the search space. It is interesting to note that just as Goolge updates this feature, Technorati has sunk their time series feature leaving BlogPulse as the only real tool for blog time series (and here, of course, I include the BlogPulse clone IceRocket).

Comments

[Zach Leatherman]

Looks like your 'Trends' link at the top of the page is borked. Take off the trailing slash for maximum click-through.

[Charlene]

The potential for longitudinal data analysis is intense. I've been recently thinking about this. I would love to take that data out of economic frameworks of analysis and pluck it into sociological and psychological ones.

[Sandro]

Google Trends is certainly a very powerful tool. However, I think it is really easy to misinterpret the graphs obtained. This is particularly true for the case of "regions". Google is normalizing the data with regard to the total search volume for a specific region. Therefore, it is very difficult to conclude that a certain term is more often searched in a specific region or city for example.