Playing with rankings

Written By: - Date published: 2:32 am, December 8th, 2008 - 10 comments
Categories: admin, blogs - Tags: , , ,

Sysop

Sysop

Since the 16th of November, I’ve been running an experiment with the Alexa ranking system to look at the site sensitivity in rankings from what is hopefully a single new person using it. This is one of the factors used by Tumeke and Halfdone‘s rankings. So I was interested in how much it would affect the Alexa ranking. It has been enlightening

None of the other admins at this site use Alexa’s Sparky because of its tracking behaviors. A number use a browser that it doesn’t support it like Safari.

The Standard’s score on this ranking had remained moderately static – from Tumuke nz blog rankings the traffic was moving around the 220k to 236k mark for a 3 month average.

October 226k nz451
September 221k nz338
August 236k nz388

I added Alexa Sparky to all of my Firefox’s on multiple machines on the evening of the 15th at home including my work laptop. At the time the alexa traffic was something like 220k despite a massive spike in the page views in election week.

I spend a considerable amount of time pulling pages from The Standard because of the nature of the moderation task in the evenings and weekends, and during breaks at work. The vast majority of which are admin comment pages. I also have two other page view measures in wordpress.stats and google analytics. Both exclude the admin pages in wordpress.

The google analytics and wordpress.stats showed a steadily decreasing number of page views, as would be expected post-election and as we move into summer.

After 3 weeks, Alexa now shows a traffic rank of 192k and nz266 rank. As far as I’m aware there have been no other readers starting to use this who read The Standard. It is possible, but the rapid decrease started when I started to do my usual operations with Alexa running after months of comparative stability. The average page views per person moved from a 3 month average of 2.9 to a week average of 10. This has to almost entirely be due to my activities.

It appears to me, as if a single person using this can significantly change the NZ traffic ranks at Alexa.  They would also change rankings that use this as a factor. Looks like a good technique to rort the ranking systems, especially for a admin of a site who spends a lot of time on their own site.

It does explain a lot of the stepwise changes in the ranking that I’ve seen that do not appear to relate to measured page views. It looks like there are so few Alexa users in NZ, that even one reading the site can change the Alexa ranks of a site that they read quite considerably.

I suspect that there will be a lot more Alexa users shortly as part of the struggle from blog ratings. It is something that I have a rather detached viewpoint on because I’m far more concerned about loadings on The Standard’s server. However I’m sure this observation will be useful for some discussion 😈

Lynn.

10 comments on “Playing with rankings ”

  • Comments are now closed

  1. peteremcc 1
    8 December 2008 at 2:55 am

    I expect there won’t be.

    Pretty much everyone around the internet has given up on Alexa for website rankings, the issues that you highlighted being only a small example of why Alexa can be so inaccurate.

  2. lprent 2
    8 December 2008 at 8:07 am

    Ah that does not surprise me. I haven’t bothered much with any SEO or ranking systems (I’m more of a c++ programmer). But I’d noticed some curious anomalies with the rankings of this site, so I decided to do an experiment.

    I saw that the wiki article on it isn’t complimentary. Made me a bit worried about what in the hell it was doing to my system. However Kaspersky hasn’t gotten worried.

  3. Madeleine 3
    8 December 2008 at 5:36 pm

    I can’t run the Alexa toolbar on my computer as my AVG scanner goes mental and attacks it for being adware. No matter how I try to tell it to ignore, to add it to the list of exceptions, it makes no difference.

    That said, MandM was steadily climbing the stats before the election and is still steadily climbing the stats post-election.

    The most effective way of pushing stats up is quality blogging.

  4. lprent 4
    8 December 2008 at 6:28 pm

    M: I’d agree. Like the opinions here or dislike them, our traffic keeps increasing month by month because people want to read and comment on them. Now I have time, I was interested in exactly how the traffic stats were calced externally.

    This blog has been coming down post election. But in the two weeks prior to election day our page views climbed up by 69% (and got me really worried about load). So that is hardly surprising that it has fallen back a bit. I’m anticipating (based on last year) a slowdown starting mid Dec to mid Jan with an almost dead period between xmas and new year dropping to half of our usual levels.

  5. Russell Brown 5
    9 December 2008 at 8:22 am

    Thanks Lynn — very interesting. The use of Alexa numbers makes a nonsense of these rankings, but you risk looking churlish if you say so.

    Public Address peaked in the week of the two elections in November; 153,515 impressions from Nov 5 to 11 — nearly triple our background level — with 33,000 pages on Nov 10. Both bloggers and readers are operating at a more modest level now …

  6. lprent 6
    9 December 2008 at 12:00 pm

    Russell,

    Yeah I’d often suspected that the Alexa stats were susceptible to manipulation. Post-election felt like a good time to find out. I was really surprised at how much they were susceptible since I was not doing more than my usual admin and commenting. The change in page-views per person is the real kicker.

    I added the firefox plugin http://www.quirk.biz/searchstatus/ to show me the google pagerank yesterday. That was interesting because it shows the PageRank and Alexa side by side. Looking at sites often shows significant differences. It is probably a good way to identify people who are using alexa to play with rankings. Since that also updates Alexa, it probably means that I’m now double hitting the stats.

    Apart from the external rankings, I’m deeply suspicous of all of these ranking systems. Off hand I can think of at least 4 or 5 ways to simply adjust the site layout to increase the number of page impressions. But those techniques at least still rely on getting people to the site – the Alexa one doesn’t, it relies on the refresh button.

    The problem with Alexa is that the statistical number of people using it is obviously far to low for its intended purpose. Problem with having a science background is that you want to check and verify – that tends to over-ride churliness. It does make me feel like I just joined the skeptics society though – dispelling myths.

    Anyway got to head to work. Been at an interview, so I stopped at home to get changed.

  7. lprent 7
    9 December 2008 at 1:33 pm

    In fact you don’t even need to do much to change the site ranking. For instance last night I was working on a plugin for some header changes to prevent client side caching of the pages.

    I need it because some people are having terrible client side caching with proxy servers etc and not seeing the comments updating on the RHS (or the posts updating on the LHS). The underlying problem actually appears to be a problem in the date/time differences between the client and the server. Because the .htaccess is already setting the expires in the http transport headers. So I’d guess that it is something dumb with a cacher that only reads html.

    In theory a few <meta http-equiv ‘s (especially cache: 0 and probably expires) should provide enough info for any non-moronic http cache to re-request the front page and single post pages

    However a side effect is that it will also cause browsers to re-request the page more frequently. I have no idea how that will affect the ranking systems, but I’ll take a bet that it does.

  8. lprent 8
    14 December 2008 at 6:42 pm

    Ok it is a week later. The rankings on this flawed system still getting better

    Alexa traffic ranking
    178,313  3mos average
    112,660 1 week average
    nz242

    avg pages per viewer
    3.6 3mos
    8.7 1week

Links to post