A few months back I decided to test something I really could not find a huge amount of information on in regards to accuracy, the Alexa rankings.

I have seen too many people place a lot of decision weight into a sites Alexa rank, sometimes being very disappointed, other times pleasantly surprised. I wanted to see how good the Alexa guesstimate really was.

Some people see it as the prime tool for site valuation while others dismiss it as fluff. I had to know for myself. Through my own sites and those of my clients I have a good deal of data for many different sites and I decided to track 4 sites and see how the Alexa data compared to the real metrics.

From what I know, Alexa uses ISP data as well as some user data to rate a sites traffic. I would imagine the data they collect serves as a sample group. The problem with sample groups is that unless the group is a true sample of the worlds internet user base, the results could be heavily skewed.

Now I am not sure if they base stats on samples, but, It would be kinda difficult to get data from every single user, so I assume its based on some type of sampling.

Alexa does state they use ISP data, I do not know which ISP’s, but there can be many issues with using an ISP as a sample of any population of internet users. For example, if the ISP data used is from a dial-up provider, you can expect that media based sites would suffer and your sample group would be occasional internet surfers, as opposed to daily die hards. If it was a dedicated line provider, it would be heavily business oriented. Sounds logical… right? anyway, I had to see how the 4 sites I picked did.

I used 2 sites that were well established, and 2 fairly new. The reason I chose these was because 1 new one and 1 old made a pair that received close to the same amount of real traffic (site 1 & 3, 2 & 4) and I wanted to see how accurate the rankings were.

The first established site had a traffic base between 1k and 2k daily, all visitors going to the homepage with 50% of the traffic coming from PPC.

The other well established site was an old single word domain that goes back to 1998, most traffic is organic SE and it consistently receives 800 to 1.3k daily.

The third site, was a newer blog that would receive 1k to 3k visitors to many different pages of which 1/3 to 1/2 would be the home page.

The last site was an informational site that was at about 600 daily with a growth that would put it in the 1k to 2k range within a few months.

At the beginning of the test period I logged the Alexa rank of all 4 sites.

I tracked this over 3 months because Alexa uses a 3 month average and I used these sites because they had close to the same page views and traffic

Here is how it looked when I started and I am rounding off the numbers

Site 1 traffic 1.2k daily / alexa 725k / page view 2

Site 2 traffic 900 daily / alexa 160k / page views 2

Site 3 traffic 1k daily / alexa 800k  / page views 2

Site 3 traffic 700 daily / alexa 2M+ / page views 2

1 month later

Site 1 traffic 1.4k daily / alexa 660k / page views 2

Site 2 traffic 800 daily / alexa 165k / page views 2

Site 3 traffic 2k daily / alexa 620k  / page views 1

Site 3 traffic 900 daily / alexa 1.5M+ / page views 2

2 months later

Site 1 traffic 1.3k daily / alexa 585k / page view 2

Site 2 traffic 900 daily / alexa 160k / page views 2

Site 3 traffic 3k daily / alexa 260k  / page views 2

Site 3 traffic 1.2k daily / alexa 670k / page views 2

Well, you can look at those numbers and make your own judgments on accuracy. The first thing that really amazed me is that site 1 and 2 are both established sites with single word domain names in the same type of traffic field (the same surfers would most likely visit both) Site 2 has always received less traffic than site 1 but 1/2 of site 1 traffic is ppc and site 2 Alexa score is incredibly better than site 1 when both sites are in the same niche and both are over 5 years old.

At this point I have 2 sites with little growth and that have not shown growth in years, also 2 sites that are growing. I did a few things to the first 2 sites to see how it tracked also site 3 had a roller coaster of a 3rd month.

Since its a bit long, I will save it for another post in a few days.