Search@Mail.Ru: search preferences ratings

Since the audience of search engines has become quite popular and varied in terms of gender, age, geography and other parameters, we have received a new interesting research tool.

Query analysis as a sociological tool has several unique advantages. First of all, this is the universality of the topics studied - people are looking for information on a variety of subjects. Secondly, in contrast to the polls here, we see one hundred percent "honesty" of the respondents. Thirdly, mass character - it is not so easy in other conditions to collect information about your favorite cat breeds from millions of people.
')
In the case of Poisk@Mail.Ru , we have another important advantage - the ability to find out the preferences of users based on their gender and age. After all, almost all users of our search have a search box on Mail.Ru, when registering which indicated their data in the questionnaire. This allows you to reveal some interesting or just funny facts.

A few examples:

• The most "male" political parties (with the maximum share of requests from the stronger sex) are the Communist Party of the Russian Federation and the Liberal Democratic Party. Women's attention is maximal for the parties "United Russia" and "Fair Russia". At the same time, Vladimir Zhirinovsky is particularly popular with the age group of 13-18 years (see picture above).
• The most "male" hobbies are casino, poker and football. The most "women's" - shopping, tanning beds and crafts.
• The maximum proportion of men's requests among all countries is from Israel. The interest of women is distributed more evenly and only Tunisia stands out somewhat.
• The most "female" interest from the "stars" of the actor Robert Pattinson, and the male audience is looking for the information about Anna Semenovich on the Web.
• The most feminine TV channel is Domashniy, the most masculine one is Vesti. If we talk about specific programs, then the largest share of the fair sex in the show "Let's Get Married", while men leave the absolute majority of requests for the transfer of "Main Road".
• And finally, drinks. Mostly women seek information about tea and coffee, while men dominate on-demand with the words "vodka" and "beer."

We periodically calculate both the general fluctuations in the interests of users on various topics, and the preferences of groups of a particular gender and age. To get an objective picture that does not depend on the features of the portal’s audience (Mail.Ru is dominated by young people and most of the users are women), we use a special indicator that is well known to the advertising industry workers - this is an affinity index or a correspondence index.

It shows how the measured topic is interesting to the selected group of users. If the values ​​are in the range of 90-110 - this, as a rule, is not considered to be object-emitting indicators, but going beyond this framework already indicates a significant interest among the selected audience.

Example: Shopping

For example, we tried to determine which purchases on the Internet are most characteristic for men and women, and for this we analyzed Top30 popular queries with the word “buy”. As it turned out, the most "male" purchase was air conditioning. Apparently, home improvement considers it the prerogative of the stronger sex.


Frequency - the proportion of the total number of requests with the word "buy"

The most "female" purchase is a clutch. But further more interesting, it turns out, women are more likely to go somewhere by train and plane, not averse to acquire a summer residence and are ready to purchase a fake sick-list in order to have a little rest in an out-of-order manner.



As for age distribution, here is a list of the most characteristic purchases as follows:

13-18 years old - scooter (frequency - 1.13%)
19-25 years old - driving license (frequency - 0.97%)
26-34 years - air conditioning (frequency - 1.65%)
35-49 years old - air ticket (frequency - 2.72%)
50-65 years old - shower cabins (frequency - 1.76%)

We are now planning to calculate user preferences (both taking into account the social modem and popularity) for some other topics (something ordered by the media, something for ourselves), so we will be happy with new suggestions for analysis.

To be continued