Google Trends in pictures:
Note: Until September 2012, this tool was called Google Insights for Search. It was then renamed Google Trends and the old Google Trends (keyword research) was removed.
Here’s what this tool does:
– keyword search trends (timeline) with comparison between keywords
– list of most popular related keywords
– related keywords on the rise right now
– filtering of trend curves by categories (for example with Apple we see the distinction between high tech and fruits)
– filtering these related keywords by categories
– map of countries, regions and cities where queries are most popular
Data scaling and normalization
Be careful in interpreting the figures, it seems that Google is based on queries in the broad sense. This means, for example, that by comparing “air conditioning” and “air conditioner”, we actually compare the set of queries using the word “air conditioning” with the set of queries using the word “air conditioner”.
Google Trends data is updated daily! Although the analyzed data concerns the whole world in all the languages managed by Google, the interface is currently only available in English. It will no doubt be translated into French in the coming months.
Scaling: the values indicated are calibrated between 0 and 100. 100 corresponds to the daily traffic record of a query, the other values are therefore an index representing the daily search volume. When you compare several keywords, the values shown are always scaled by taking the index 100 for the maximum reached by the keyword with the highest traffic.
Normalization: All values are normalized, i.e. actual search volume numbers are changed to make comparisons easier. For example, the actual data per region is modified based on the total traffic in that region.
Search by entities
In December 2013, Google introduced the ability to specify the entity to be analyzed. For example, if you type “bordeaux” in Google Trends, are these searches related to wine or the city (or something else or both)? You now have the option of specifying the theme of your search (let’s call it an “entity”), or else to stay in the global sense.
It is very useful for:
- not be polluted by queries that use the same word but have nothing to do with it (for example the island of Java or the Java programming language)
- let Google expand the search to other spellings of the entity you are referring to (for example for the city of Tokyo, Google will not only base it on how you typed it, but also on variations like 東京, Токио, Tokyyo, Tokkyo, Japan Capital , etc. Google even indicates that its algorithm could take into account a query like “the main actress in Iron Man” when looking for the entity corresponding to the actress Gwyneth Paltrow.
This new feature is still in development, at the “beta” stage. Google announces that it has made available around 700,000 themes (or rather entities), which can then be analyzed more finely by filtering by country (but only for Brazil, France, Germany, India, Italy, the UK and the USA.
It’s a nice demonstration from Google that now its algorithm is based on the notion of entities, much more than on words or even syntagms. This notion of entity processing is at the heart of the Hummingbird algorithm in place at Google since the summer of 2013, as well as in its “Knowledge Graph”.
Moreover, coincidentally, apart from Russia, the 7 countries for which it is possible to refine searches by entities in Google Trends are also the countries for which the Google Knowledge Graph is available.
Vocabulary used by Google
– Categories: division into sectors of activity
– Rising Searches: refers to related keywords on the increase, calculated between the selected analysis period and the one preceding it (if the selected period is the year 2007, the comparison is made with the year 2006; if the period is June 2007, the comparison is made with May 2007).
– Breakout: is said of related keywords having a very strong increase (greater than 5000%)
Also about Google Trends…
Google Flu Trends, flu tracking
For years, Google has offered Google Flu Trends, a tool that tries to predict the arrival of the flu in a region of the world, based on queries from Internet users.
Certain search terms seem to be good indicators of the spread of the flu. In order to provide you with an estimate of the spread of the virus, Google “Flu Tracking” therefore collects data relating to searches launched on Google.
It takes about 2 months before this interface also has official data, now taken into account throughout the season in the new model developed by Google.
As in Google Trends, Google makes believe that Limousin and Corsica do not have access to the Internet (in any case geolocation does not really seem to work well for these regions)…
Unfortunately, this service is now closed (details here).