English, British subtitles

← Why might the estimated number of results change when going from page 1 to page 2?

Get Embed Code
40 Languages

Showing Revision 3 created 10/10/2014 by Sherry Chen.

  1. Today's question comes from
    San Francisco, California.
  2. Blind Five Year Old asks, "How reliable is the site colon,
  3. query, and why does the total count sometimes change from
  4. page one to page two?"
  5. Well, I wouldn't really put it in terms of
  6. the site colon query!
  7. Instead, I would pivot and narrow and focus down on the
  8. results estimates.
  9. So there's a few things to know.
  10. Number one, if you haven't noticed, you should notice
  11. that results estimates are only accurate to about three
  12. significant digits.
  13. So they are estimates.
  14. As we're looking for queries, at some point we say, OK, we
  15. have enough to constitute the first page, and we see how far
  16. we've gone through our posting list - the
  17. amount of data we have--
  18. therefore, we can estimate there's about this many
  19. results left.
  20. Now when you click from page one to page two, suddenly
  21. you're going deeper.
  22. And so you have a better estimate or you have a
  23. slightly different estimate, because you're starting to
  24. look not just a little bit of the way through all those
  25. documents, you're actually starting to
  26. go deeper and deeper.
  27. So it's relatively common for our results estimates to
  28. Not Synced
    change going from page one to page two, typically becoming
  29. Not Synced
    more accurate.
  30. Not Synced
    So it's not usually a factor of the site colon query.
  31. Not Synced
    It's usually a factor of us digging deeper through our
  32. Not Synced
    data, and as a result, we have a better or at least a
  33. Not Synced
    different estimate for how many results there are.
  34. Not Synced
  35. Not Synced
  36. Not Synced
  37. Not Synced