Content Length and Rankings

Discussion in 'Case Studies and Journals' started by Bender Bending Rodríguez, Feb 6, 2017.

Share This Page

  1. Bender Bending Rodríguez

    Bender Bending Rodríguez Senior Member

    Joined:
    Mar 10, 2016
    Messages:
    507
    Likes Received:
    338
    I analyzed the top 10 rankings for 400 keywords. Here are the rankings, average word count for pages at that ranking, and standard deviation:

    1. 3,287 words - 3303 STD
    2. 3,004 words - 3774 STD
    3. 2608 words - 2713 STD
    4. 2861 words - 4428 STD
    5. 2582 words - 2575 STD
    6. 2434 words - 2401 STD
    7. 2729 words - 3051 STD
    8. 2899 words - 4076 STD
    9. 2615 words - 2307 STD
    10. 2607 words - 2430 STD
    On the whole, the average word count for the top 10 was 2,749. The STD was 3207. The total URLs analyzed was 4,000.

    The Pearson Correlation Coefficient for Word Count and Rankings was -.5768, which shows a high correlation between Word Count and Rankings.

    [​IMG]

    I then analyzed what percentile a 500 word, 1,000 word, 1,500 word, 2,000 word, 2,500 word, and 3,000 word article would be for that rank (see table above). This was to see what might happen if I cut costs in half by going from a 2,000 word article to a 1,000 word article (or any other combination of words). Basically, a drop of 500 words is a drop in 5 to 7 percentiles (and an increase in 500 words is an increase in 5 to 7 percentiles).

    I then compared the article's word count with total links to the article and total links to the site. The Pearson score for Word Count and Backlinks was -.03 (no correlation). The Pearson score for Word Count and Total Domain Backlinks was -.12 (no/little correlation).

    So, longer articles are ranking better than shorter articles and it is not because longer articles got more backlinks than shorter articles or because longer articles are appearing on more authoritative sites. Its probably because Google has a preference for long articles. You can use the above table to figure out how many words to order based upon what percentile you want to be in.

    EDIT: I just ran the Pearson score for Backlinks and Rankings as well as Domain Backlinks and Rankings. The scores are .07 and -.03 respectively. These are medium/long tail keywords. For these 400 keywords, there's a much, much greater correlation between word count and rankings than backlinks and rankings!
     
    Last edited: Feb 6, 2017
    Mordor and Trevor like this.
  2. sites

    sites Active Member

    Joined:
    Jan 2, 2016
    Messages:
    89
    Likes Received:
    75
    Neat!

    Were the keywords informational or transactional?

    Did you exclude the non-essentials (text associated with sidebar links and footer links, information about comments, and the like) while counting the words?

    Guess you got the backlink information from one of the usual sources (AHrefs/Majestic/Alexa/...). (Or did you combine multiple sources?)
     
  3. Bender Bending Rodríguez

    Bender Bending Rodríguez Senior Member

    Joined:
    Mar 10, 2016
    Messages:
    507
    Likes Received:
    338
    The original test was on e-comm keywords.

    Word count is all text on page including header, menu, sidebar, and footer.

    Data was from AHrefs.

    I re-did the test on 3 other topics. Here were the results:
    1. ~40 super low competition keywords such as "Albania Embassy in Morocco" (In English and not Albanian). The top 10 were all around 1,500 words total. There was no correlation. I believe it was because all the pages had the same words.
    2. ~17 "$Brand review" keywords such as "Apex Forum Review". Here, the average word count at #1 was 3,000. #2-#10 was 1,500 to 1,000. The correlation was -27% (medium/low). I think it was lower since #2-10 were about the same word count. There was no correlation for ranking and backlinks, ranking and domain backlinks, words and backlinks, and word and domain backlinks. Again, hints that word count is being used in these results.
    3. ~60 competitive results. The word count of top 10 were very similar and there was no correlation.
    So, yeah, when the word count of the top 10 is different, there's a correlation with more words and higher rankings. When the word count of the top 10 is the same, there's no correlation between word count and rankings (but you should have the same amount of words or close to that!).
     
    sites likes this.
  4. emp

    emp Senior Member

    Joined:
    Dec 15, 2015
    Messages:
    376
    Likes Received:
    287
    ? I am not sure what you are doing there.

    Either you don't know what a standard deviation is, or you use ther acronym STD for something else.

    :p

    ::emp::
     
  5. Bender Bending Rodríguez

    Bender Bending Rodríguez Senior Member

    Joined:
    Mar 10, 2016
    Messages:
    507
    Likes Received:
    338
    My guess is that LSA algorithms prefer more words so that they can get a better guess at the topic. For example, feeding an LSA algorithm 10,000 books would have it learn English OK but feeding the same LSA algorithm 1,000,000 books would have it learn English much, much better.

    So, for Google, having a 3,000 word page gives Google a very, very good idea at the page's topic whereas a 500 word page gives Google a so-so guess at the page's topic. The better guess gets rewarded with higher rankings.

    @cardine Any inputs?

    I also run some adult sites and am curious how this plays out in adult. For example, I have tube sites with thousands of videos on them. They're scrapes from xHamster, YouPorn, etc. Same title, same description, same comments. No spinning. Google indexes these tubes faster than it indexes my mainstream sites (with actual, unique words). I'm curious if its because Google recognizes that its scraped content and is indexing it accordingly or that it has a different indexing algorithm for media heavy sites. I'm also curious how adult rankings work w/ word count since the top ranking sites are always xVideos, PornHub, xHamster, etc and they have 0 words (its either a search page or a category page).

    So, yeah, looks like a few more tests for adult are coming up!
     
  6. Bender Bending Rodríguez

    Bender Bending Rodríguez Senior Member

    Joined:
    Mar 10, 2016
    Messages:
    507
    Likes Received:
    338
    The correlation between words and rankings in adult is 19%, as in more words correlates with lower rankings in Google. Here's the data:

    Words and Rankings 19%
    1 643.4
    2 548.3
    3 1078.6
    4 1835.4
    5 835.1
    6 655.5
    7 1149.5
    8 861.1
    9 959.1
    10 5440.4

    IMO, the data is just an outlier. No one in the dataset is actually doing SEO and the ranking pages are search or category pages for xHamster, YouPorn, etc, who have millions of links to the domain. After the big sites, there might be a tumblr blog or a small tube. That's all.

    The larger word count pages are also search pages or category pages that are just showing more videos (and thus more titles). Its not real sentences. This must be giving the LSA algorithm a really hard time!