I was developing pagination on big data, and tried to see the UX of Google search results. After number of test tries, I came to following conclusion that Google displays fake total results found.

If you try to search anything, say "entity", it would show

About 224,000,000 results (0.60 seconds)

But, when you navigate to other results page, by jumping (clicking on highest page number shown), you will note that there are not even 100 pages :( , that is 10 results per page * 100, is equals to 1000 results, means there are not even thousand results.

In above case the last page is page 37. On page 36 it shows:

Page 36 of about 224,000,000 results (0.71 seconds)

On page 37 it changes its declaration to as follows:

Page 37 of 367 results (1.09 seconds)

Am I doing something wrong that we don't get what is reported, or are we reported fake figure ?

We use these giants (Google/Facebook/Twitter) as yardstick for our big data applications, does the above experience means that they also just work around issues instead of finding solutions.

Thank you.

  • 2
    I don't really see what your question is here. Are you just asking if Google results count is accurate? Because that's not really a UX issue. Can you describe what it is that you want to do yourself and we'll advise on whether or not your approach is optimal. – JonW Nov 18 '14 at 13:24

There is no practical reason for them to serve up the least relevant results.

There may well be millions of results, but Google also knows that there is no way a human could look a them all since people tend to give up or try a different search after the first page or two. So they cut corners to make things faster.

On page 37 you'll see this at the bottom:

Search results omitted.

If you choose to repeat the search you can in fact get farther than page 37:

Page 71

However, if you manipulate the URL to jump ahead to page 100/result 1000 you will run into this road block:

enter image description here

So no matter how many millions of results there may be in reality, Google only serves up the 1000 most relevant ones (and may default to less than that if there are lots of duplicates).

I don't see a compelling use case for showing the least relevant results, since they are, by definition, not what the user is looking for. As long as they are confidant in their ordering for results that are indeed the most relevant, this approach probably saves lots of time and money.

Not the answer you're looking for? Browse other questions tagged or ask your own question.