My Google SERP testing

I have seen many discussions here where members have expressed concern over what position their website comes up in the SERPs.
I had the thought that maybe one part of Google’s algorithm might be randomness and that websites may not be displayed in an order based on merit alone. That is, if there were a lot of “OK” results, might some of them be included in the results to help determine if they should gain a more permanent prominence?

For any experiment to yield useful results it is important to have all things equal except for the one variable being tested.
It is also importantant to have as large a data set as possible.
I met neither of these to my satisfaction and I consider the results inconclusive.
I did however make some interesting observations, and now have a few more questions!

To hold the data I created four tables:

  • terms (term, number_found, seconds, date_searched)
  • domains (term, domain, date_searched)
  • backlinks (term, number_found, seconds, date_searched)
  • results (term, p_title, p_url, position, date_searched)

Two objectices were to

  • reduce the possibilty of search history affecting subsequent searches
  • I made sure I was not logged in to Google Wemaster Tools and I Deleted browser cookies prior to every search
  • make copying and pasting easy to do
  • For the majority of tests I turned the default Instant Results off, and set to 100 results per page.
  • Google Search Preferences
  • I used custom CSS to display: none everything except the displayed page titles and URLs. i.e. no advertisements, no image results, no page content snippets, etc.

instead of


this instead


The Terms Table
I didn’t find anything all that useful from this table’s data.
What was interesting is how often the number of results and the time it took varied for the same search term.
I saw no coorelation between the two, and the amount of variance was generally not that great.
Differences in how many results were found, and how fast or how slow the results were returned had no apparent effect on page positions.

The Domains Table
Because I used only the displayed page URLs, and only those with protocols (I used FILTER_VALIDATE_URL) were entered into the database, this table’s usefulness is limited.
What I did find interesting is that of 42 searches totaling 4133 found pages, the domain found most often (albeit the page positions were most often in the 90’s and 80’s) was books.google.com i.e.

books.google.com		121 
developer.wordpress.org	53 
teamtreehouse.com	50 
premium.wpmudev.org	43 
wordpress.org		42 

The Backlinks Table
This table was not used because searches using the link: search modifier proved to be completely unreliable to the point of being useless.
Questions about the Google search link modifier

The Results Table

By far, this table proved to be the most useful.
By using different queries it was possible to compare search results many different ways.

Instant Results vs. Never and 100 per page

Because this Google Help page has

The possible searches that you see are based on what other people are searching for and the content of web pages indexed by Google.

I wanted to see if there was any noticeable difference between the default Instant Results 10 per page SERPs and the 100 per page SERPs.

For the most part, I did not see any differences in page positions based on the setting.

An interesting exception being that searches for the term wordpress returned results for twitter.com/WordPress (with upper case) in the number two position, but when using the default Instant Results, twitter.com/wordpress (lower case) was also returned in the number twelve position.

Reversed Search Terms

Trying different search term order eg. wordpress codex vs. codex wordpress, had the most effect on what and where pages were returned in the search results.

I now understand why in the past many “stuffed” all possible variations into the now unused meta keywod tag.

Observations

  • Most changes in both position, and inclusion or omission within the first 100 results, occur in the later positioned pages.
    That is, pages that occur earlier in the results tend to be more “stable”.
  • Which search term is used can have great effect on both position and whether or not a page is included within the first 100 results.
  • In addition to page position, site density can vary.
  • The number of results per page setting has little effect on the returned results.
  • If randomness has any part in position and whether or not a page is included within the first 100 results, it does not happen within a short period of time and happens mostly with later positioned pages.

results.html (16.4 KB)

2 Likes

This topic was automatically closed 91 days after the last reply. New replies are no longer allowed.