16% of search queries every day — new
In one of his presentations, Google has published an interesting figure. It turns out that every day search website handles 16% new search queries that have never met before.
It would seem that where they come from, if all the words from dictionary with all the possible typos and all combinations have been exhausted? But if you think about it, it's easy to understand where they come from. First, contribute a couple of hundred languages other than English. Second, users very often look for specific information in a long unique search queries (addresses of specific establishments, names and biographies, quotes from songs, literary works and pieces of code much more). In the end, it's pointless shortcuts that accidentally hit the search engine and also unfinished requests from a dynamic search that updates results as you type user.
Maybe in the statistics taken into account not the total volume of queries and individual queries. For example, if one million people searched for [facebook], and another person was looking for [dfg8734kjv3d], such statistics 50% requests will be "completely new". That is, of the average daily volume of 3 billion search queries that Google handles, not 480 billion are brand new, and much less.
Anyway, it turns out that for a certain period of time (several months?) the proportion of unique search queries, which was found only once, may already exceed the number of "popular" search queries, which met two or more times.
If we assume that the proportion of unique queries in total amounts to only 1%, with 90% of them never occur again in the future, it turns out that unique "one-off" requests reach the 50% level in the list of requests already for 55 days and reach level 99,(9)% request list for 110 days of operation of a search engine.
Even if we reduce the proportion of new queries to 0.1%, it is still turn 99,(9)% comes remarkably quickly.
Article based on information from habrahabr.ru
It would seem that where they come from, if all the words from dictionary with all the possible typos and all combinations have been exhausted? But if you think about it, it's easy to understand where they come from. First, contribute a couple of hundred languages other than English. Second, users very often look for specific information in a long unique search queries (addresses of specific establishments, names and biographies, quotes from songs, literary works and pieces of code much more). In the end, it's pointless shortcuts that accidentally hit the search engine and also unfinished requests from a dynamic search that updates results as you type user.
Maybe in the statistics taken into account not the total volume of queries and individual queries. For example, if one million people searched for [facebook], and another person was looking for [dfg8734kjv3d], such statistics 50% requests will be "completely new". That is, of the average daily volume of 3 billion search queries that Google handles, not 480 billion are brand new, and much less.
Anyway, it turns out that for a certain period of time (several months?) the proportion of unique search queries, which was found only once, may already exceed the number of "popular" search queries, which met two or more times.
If we assume that the proportion of unique queries in total amounts to only 1%, with 90% of them never occur again in the future, it turns out that unique "one-off" requests reach the 50% level in the list of requests already for 55 days and reach level 99,(9)% request list for 110 days of operation of a search engine.
Even if we reduce the proportion of new queries to 0.1%, it is still turn 99,(9)% comes remarkably quickly.
Комментарии
Отправить комментарий