Google Trends: Mere perception not reality

Yesterday I came across and used the Google trends software for the first time. The software calculates trends and compares between two key word searches provided by the user. The “so-called” popularity is calculated based on the searches for the word on the Google search engines over time specific time. Some of the pitfalls of the applications include:
1. I think (correct me if I am wrong) that the Google trend software does not consider the connotation and semantics of the word/phrase searches. For example, I searched “Marathi’ and ‘Hindi’ (languages In India) and got spikes in the news record at some points. When I saw the news at the spikes they were as:
“President gives away 'Hindi Sevi Samman Awards 2002'”
“Chinese Premier revives Hindi-Chini Bhai Bhai”
“Prasar Bharati invites Hindi film right holders for telecast on DD”
“Assamese help Hindi-speaking people in distress”
Now I was interested in comparing the popularity of Hindi versus the Marathi language but the results I obtained were completely misleading.

2.. Then I am not sure how Google trends is handling searches for words like VIP. Is it a user interested in the bags company VIP, abbreviation VIP (Very Important Person) or Visual Information Processing. Another example will be of Apple (fruit or company) or windows (software or house window). One can come up with numerous such examples. However, I guess I am clear on the point I am stressing.

3. The application considers only English news. With more than a billion people each in China and India, having their mother tongue anything other than English the punch line “See what the “world” is searching for” is certainly misleading.

4. One more thing to keep in mind is that the search results are just a part of Google only and does not consider other engines like Yahoo, MSN or AltaVista (I agree that Google is the most used search engine worldwide. Personally, I use it for every search).

5. The next thing, which bugged me, is the searches for phrases and concatenated words. Consider a search like “Life Insurance” here the importance for the word "life" is very less. The user is interested in insurance policies for life. I am again not sure how is the Google trends software handling such searches.

6. Google declares that “The results are based upon just a portion of our searches, and several approximations are used when computing your results”. The degree of approximations remains a point of discussion and validity of the results.

7. Searching a word on the search engine should be differentiated from actually clicking and browsing on the search results. The information of how many searches lead to actual browsing of the search details is not furnished or available.

8. Finally, I am also not sure whether the data is collected only on “” or it also contains “Google scholar”, “Google image” and other such specific search mechanisms.

1 comment:

  1. I have not experimented a lot with google trends and yes, it may give ambiguos results for ambivalent words/phrases but I was impressed by the tool. I tried the following searches : 'h1 quota', 'golf masters' 'daytona'. The featured regions and timeline tells you a lot. :D