There’s lot of hype about Alexa Rank, especially among bloggers. A simple Google search will give you 100+ tips on how one can increase Alexa rank. Even I wrote a guide to 20 Quick ways to increase Alexa Rank, long back.
But does it really matter? Yes, for people who have a mindset of rating a site based upon Alexa rank. No, for the rest of the web.
Image Credit: Alexa sucks
I was amazed by the popularity, and even some quick tips helped raising the rank. But what is the secret? How does it calculate? I wanted to dig-out the Reality out of curiosity.
One of the similar recent Curiosity lead me to Discover What Architecture powers YouTube, the Video site that handles billions of views everyday.
I had been monitoring and analyzing the pattern of how Alexa ranking works over a period of 1 year. It involved data capture and analysis and lot of maths. I was surprised to know, it wasn’t relevant enough. It’s implementation lacks consideration of certain facts that can make it’s Ranking wise deal.
Ranking Algorithm Isn’t Perfect
The ranking algorithm isn’t perfect. Firstly, all analytics data is based on last 3 months Only. This is something really bad. Suppose site X received a million hits-a-day 95 days back and unfortunately today they get half million, site gets no credit of achievements in the past. Same holds good for other ranking parameters.
The rank is calculated based upon certain Parameters. And some of them are wrong ways of doing it. Let’s cover them one by one:
Traffic Rank and Pageviews: Pageviews looks at “how much traffic a site had been receiving over last 3 months”. The value is compared against the web trends and ranked. Looks fair – More traffic on a site, better it is.
However, TrafficRanking, which is supposed to be driven by Pageviews, isn’t exactly so. e.g. When two sites have same Pageviews, TrafficRanking can be totally different. This difference can, Max., be of the order of 30x times (explained later).
This is so because TrafficRank is nothing but the Alexa Rank. PageViews is NOT the only factor that influences it. Okay, it’s fair, we need to have more parameters.
Whatever the parameters may be, PageViews should take highest precedence, but it`s not the case here . The rank is weighted/divided inappropriately among the other parameters, as discussed further.
Reach: Specifies how much on the breadth, is a particular site popular (worldwide).
Pageviews/User: This has to do with returning users. Does a site’s user return again? Simply stating, does a site make permanent user experience? If yes, site is ranked better.
Bounce %: Bouncing is defined by users exiting a site. If users go out from a site more frequently, via hyperlinks or whatever, the site is awarded with a lower Rank!
Time on Site: This is self explanatory. If average of time users spent on a site is high, it’s ranked higher. Ahem.. Think of this Gmail, users keep it open all day and Google get’s opened on need basis. It means that Gmail will be ranked higher than Google.
Search %: How much of site’s traffic comes from Search engines.
Sites linking In: This is the Most Influencing factor. How many backlinks a particular site/blog has, impacts the rank heavily. I agree, even Google does the same. But wait a minute, it’s updated every 3 months. Hey Alexa, are you kidding me?
If I get better traffic and pageviews, better search%, better time on site, by 2x factor, Alexa will still rank the other site better which has better backlinks that too updated 3 months back. Ahh Snap! I am more than impressed.
Now look at it all together:
Site Score = ( A x Views + B x Reach + C x PerUser + D x Bounce + E x Time + F x Search + G x Links ) / Some constant
Here G has very high value of the order of 3. A = 2, B=1.5, C = 2, D = 1.6, E = 1.9, F = 2, G= 3.
You can clearly makeout these values aren’t perfect for ranking a Site.
Values like A = 3, B=2, C = 3, D = 0, E=1.9, F = 3.5, G = 3 makes more sense for me. (d=0 means it shouldn’t depend on bounce rate as we are considering Loyality with PerUser already)
How I calculated this Formula:
I won`t explain everything, since it’s complex, here is a brief on it:
All values were calculated by studying alexa’s behavior over a period of 1 year. During this period, data was collected in 4 sectors (each of 3 months), making the formula more accurate.
An automated PHP script captured all the parameters by parsing HTML/AJAX objects on daily basis for Taranfx’s Alexa Stats to conclude to the approx. mathematical formula. The actual formula came out to be much more complex but the above one gives 90 – 95% accuracy, not bad at the cost of ease.
If you are good at mathematics, you can do this too and share your formula.
I`ll be happy discuss any comments coz this is interesting. 🙂 Or if you wish to see some analytics or Research on any (Tech) topic, feel free to suggest in comments below.